Adhesive Joints

Formation, Characteristics,
and Testing
Adhesive Joints
Formation, Characteristics,
and Testing

Edited by
K. L. Mittal
IBM Corporation
Hopewell Junction, New York

PLENUM PRESS. NEW YORK AND LONDON

 Library of Congress Cataloging in Publication Data

International Symposium on Adhesive Joints: Formation, Characteristics, and Testing

(1982: Kansas City, Mo.)
Adhesive joints.

"Proceedings of the International Symposium on Adhesive Joints: Formation, Char-

acteristics, and Testing, held September 12-1:7, 1982, in Kansas City, Missouri"-
T.p. verso.
Includes bibliographical references and index.
1. Adhesive joints- Congresses. I. Mittal, K. L., 1945- . II. Title.
TA492.A3I58 1982 668.3 84-2132
ISBN-13: 978-1-4612-9702-4 e-ISBN-13: 978-1-4613-2749-3
DOl: 10.1007/978-1-4613-2749-3

Proceedings of the International Symposium on Adhesive Joints: Formation,

Characteristics, and Testing, held as a part of the American Chemical Society meeting,
September 12-17,1982, in Kansas City, Missouri

© 1984 Plenum Press, New York

Softcover reprint of the hardcover 15t edition 1984
A Division of Plenum Publishing Corporation
233 Spring Street, New York, N.Y. 10013

AIl rights reserved

No part of this book may be reproduced, stored in a retrieval system, or transmitted,

in any form or by any means, electronic, mechanical, photocopying, microfilming,
recording, or otherwise, without written permission from the Publisher
PREFACE

This volume documents the proceedings of the International

Symposium on Adhesive Joints: Formation, Characteristics and
Testing held under the auspices of the Division of Polymer Mater-
ials:Science and Engineering of the American Chemical Society in
Kansas City, MO, September 12-17, 1982.

There is a myriad of applications (ranging from aerospace to

surgery) where adhesives are used to join different materials, and
concomitantly the understanding of the behavior of adhesive joints
becomes very important. There are many factors which can influence
the behavior of adhesive joints, e.g., substrate preparation, in-
terfacial aspects, joint design, mode of stress, external environ-
ment, etc., and in order to understand the joint behavior in a
holistic manner, one must take due cognizance of all these germane
factors. So this symposium was planned to address not only how to
make acceptable bonds but their characterization, durability and
testing were also accorded due consideration.

The symposium was organized with the following objectives in

mind: to discuss the latest developments and activities anent
adhesive joints, to provide a forum for cross-pollination of ideas,
to bring together the researchers from various disciplines, and to
highlight the areas which need intensified efforts. The interest
and high tempo of research in adhesive joints is well manifested by
the size of the program which contained 51 papers from many parts
of the globe covering various ramifications of adhesive joints. So
this was truly an international event and if the comments from the
attendees are a measure of the success of a symposium, then it was
more than successful and fulfilled very well the objectives of the
symposium.

The present volume containing a total of 51 papers (921 pages)

by 94 authors from 12 countries is divided into five parts. The
topics covered include: substrate and interfacial aspects of ad-
hesive bonding; determination of the locus of failure; factors in-
fluencing behavior of adhesive joints; evaluation, characterization
and testing of adhesive joints; stress analysis and performance

v
vi PREFACE

aspects, fracture aspects, and durability or stability behavior of

adhesive joints. Apropos, the availability and widespread use of
sophisticated surface analytical tools has proven a boon in adhe-
sive joints research and in this volume are included many examples
illustrating the usefulness of these techniques.

The purpose of a symposium is to present the state of know-

ledge of the topic under consideration and it can best be accom-
plished by a blend of overviews and original research papers. This
is exactly what was done in this Symposium and as a result this
volume contains a number of overview papers and unpublished orig-
inal research contributions. It should be added here that all
papers were properly reviewed as comments from the peers are a
desideratum to maintain the standard of publications. As a result,
most manuscripts were returned to the respective authors for suit-
able revisions. It should also be pointed out that this volume
contains some papers which were not presented, and certain papers
which were presented are not included for a variety of reasons. As
for the discussion section, yours truly had hoped to include dis-
cussion at the end of each paper or group of allied papers, but the
number of written questions received did not vindicate undertaking
such endeavor. However, it should be recorded that there were many
spontaneous and lively discussions both formally in the auditorium
as well as in other (more relaxed) places.

Even a cursory look at the Table of Contents shows that many

facets and ramifications of adhesive joints are accorded due
coverage in this volume. It is hoped that this volume would be of
interest to both the veteran (as a reference source for contem-
porary research and development activities) and to the tyro as a
fountain of new ideas.

Acknowledgements: First of all I am thankful to the Division

of Polymer Materials:Science and Engineering for sponsoring this
event, and my thanks are due to H. R. Anderson, Jr. (IBM Corp.) for
permitting me to organize this symposium and to Steve Milkovich
(IBM Corp.) for his understanding and cooperation during the tenure
of editing.

On a more personal note, I am thankful to my wife, Usha, for

helping me in more ways than one, and to my children (Anita,
Rajesh, Nisha and Seema) for rendering home environment conducive
to work and for letting me use those hours which rightfully
belonged to them. Special thanks are due to Phil Alvarez (Plenum
Publishing Corp.) for his continued interest in this project. I
would like to extend my appreciation to Barbara Mutino (Office
Communications) for meeting various typing deadlines with a smile.
Special thanks are due to the unsung heroes (reviewers) for their
time and valuable comments. The generous support of the Petroleum
Research Fund of the American Chemical Society towards travel
PREFACE vii

expenses of certain invited overseas speakers is gratefully

acknowledged. Last, but not least, the cooperation and enthusiasm
of the contributors is sincerely appreciated without which this
book would not have seen the light of day.

K. L. Mittal
IBM Corporation
Hopewell Junction, New York 12533
CONTENTS

PART I. SUBSTRATE AND INTERFACIAL ASPECTS

The Role of Surface and Bulk Characterization in the

Evaluation of Adhesive Joints
W. L. Baun ••.••.••.•••••••.•••.••••••••.•••••••••••••• 3

Practical Adhesion Measurement in Adhering Systems: A Phase

Boundary Sensitive Test
A. A. Roche, M. J. Romand and F. Sidoroff ••••••••••••• 19

Effect of a Phenol Formaldehyde Resin on the Adhesion of a

Polysulfide Sealant
R. Ramaswamy and P. Sasidharan Achary ••••••••••••••••• 31

Polysulfide-Polyurethane Interfacial Aspects

A. M. Usmani •••••••••••••••••••••••••••••••••••••••••• 41

Epoxy Resin Wetting of E Glass Single Filaments as it

Relates to Shear Strength
E. J. Berger and Y. Eckstein •••••••••••••••••••••••••• 51

PART II. FACTORS INFLUENCING BEHAVIOR OF ADHESIVE JOINTS

Influence of Surface Roughness on Mechanical Properties of

Joints
Y. Gilibert and G. Verchery ••••••••••••••••••••••••••• 69

The Effects of Surface Conditioning of AISI-304 Stainless

Steel on the Interfacial Properties of Alloy/Epoxy Composite
Structure; Adherend Surface Characterization Using X-ray
Emission Spectroscopy (LEEIXS)
F. Gaillard, A. A. Roche and M. J. Romand ••••••••••••• 85

ix
x CONTENTS

Surface Characterization of Anodic Oxides on Aluminium

Alloys by Means of Surface Potential Difference, Surface
Impedance and Surface Morphology
A. Kwakernaak, R. Exalto and H. A. van Hoof •••••••••• 103

Investigation into the Effect of Surface Treatment on the

Wettability and the Bondability of Low Surface Energy
Materials
J. P. Jeandrau •••••••••••••••••••••••••••••••••••••••• 121

The Effect of Moisture on the Dimensional Stability of

Adhesively Bonded Joints
E. J. Hughes, J. Boutilier, and J. L. Rutherford 137

The Dimensional Stability of Epoxy Adhesive Joints

J. P. Sargent ••••••••••••••••••••••••••••••.•.•••••••• 151

Strength Characteristics of Mono and Multiple-Wire Steel to

Steel Joints Bonded with an Epoxy Adhesive
R. W. Hylands ••••••••••••••••••••••••••••••••••••••••• 165

A Study of Adhesive Joints Between Aliphatic Polyamides and

Metals
S. S. Pesetskii, V. E. Starzynskii, and
S. V. Shcherbakov ••••••••••••••••••••••••••••••••••••• 195

"Honeymoon" Phenolic Fast-Setting Adhesives for Exterior-

Grade Finger Joints
A. Pizzi •••••••••••••••••••••••••••••••••••••••••••••• 214

Ageing of Structural Film Adhesives - Changes in Chemical

and Physical Properties and the Effect on Joint Strength
C. E. M. Morris, P. J. Pearce and R. G. Davidson 231

Structural Adhesives Based on Diallyl Phthalate

.A. M. Usmani •••••••••••••••••••.••••••••••••••••••••••• 247

Possibilities of Integrating Surface Treatment of Bonding

Parts in the Adhesive Bonding Process
K. Ruhsland ••••••••.•.•.•.••••••.•.•••••••••••••••.••• 257

PART III: EVALUATION AND CHARACTERIZATION

Evaluation of Adhesive Test Methods

G. P. Anderson, K. L. DeVries and G. Sharon 269

A Critical Appraisal of Dental Adhesion Testing

W. J. O'Brien and S. T. Rasmussen ••••••••••••••••••••• 289
CONTENTS xi

Adhesive Joint Characterization by Ultrasonic Surface and

Interface Waves
s. I. Rokhlin ..••.•••.••.••.•••••••.•.•••••.•••••••••• 307

Nondestructive Evaluation of Some Bonded Joints

T. C. Ward, M. Sheridan and D. L. Kotzev •••••••••••••• 347

Ultrasonic Assessment of Cure Rate Effects in Bonded

Honeycomb Structures
R. A. Pike and R. S. Williams ......................... 369

The Three-Point Bend Test for Adhesive Joints

N. T. McDevitt and W. L. Baun ••••••••••••••••••••••••• 381

Accelerated Aging Procedures for Glue-Wood Bonds and Their

Use for In-Plant Quality Control
R. B. Jathar •••••••••••••••••••••••••••••••••••••••••• 395

Peel-Strength and Energy Dissipation

T. Igarashi ••••••••••••••••••••••••••••••••••••••••••• 419

The Estimation of Adhesion in Filled Polymer Systems

Yu.S. Lipatov. T. T. Todosiychuk, P. K. Tsarev and
L. M. Sergeeva .•••••••••••••••••••••••••••••••••••••••• 433

PART IV. DURABILITY OR STABILITY ASPECTS

Adhesion and Durability of Metal/Polymer Bonds

J. D. Venables ••••••••.••••••••••••••••••••••••••••••• 453

Failure Mechanisms in the Boundary Layer Zone of Metal/

Polymer Systems
W. Brockmann, a.-D. Hennemann and H. Kollek ••••••••••• 469

Joint Durability Studies with Abraded, Etched, Coated and

Anodized Aluminum Adherends
J. D. Minford ••••.••••••.••••••••••••.••••••••••••••.• 485

Comparative Study. of Aluminum Joint Strength and Durability

with Varying Thickness, Boehmite-Type Oxide Surfaces
J. D. Minford ••••••••••••••••••••••••••••••••••••••••• 503

Epoxy Adhesion to Copper, Part II: Electrochemical

Pretreatment
J. M. Park and J. P. Bell ••••••••••••••••••••••••••••• 523

Hydrothermal Stability of Titanium/Epoxy Adhesive Joints

F. J. Boerio and R. G. Dillingham ••••••••••••••••••••• 541
xii CONTENTS

A Study on Elastomer/Metal Bonds Applicable in Underwater

Sonar Systems
R. Y. Ting •..•.•.••.•.•.•.•.•••.•.......•....•.•..•.•. 555

PART V. STRESS ANALYSIS AND PERFORMANCE ASPECTS

Life Prediction Methodology for Adhesively Bonded Joints

J. Romanko, K. M. Liechti and W. G. Knauss •••••••••••• 567

Stress Analysis of Adhesively Bonded Joints

R. A. Kline •..•.•.•.•..•.•.••••.•.•....•.....•.•••...• 587

The Impact Strength of Adhesive Lap Joints

J. A. Harris and R. D. Adams •••••••••••••••••••••••••• 611

The Performance of Adhesive-Bonded Thin-Gauge Sheet Metal

Structures with Particular Reference to Box-Section Beams
A. Beevers and A. C. P. Kho ••••••••••••••••••••••••••• 627

Cyclic uebonding of Adhesively Bonded Composites

S. Mall, W. S. Johnson and R. A. Everett, Jr. 639

Effects of Low Cycle Loading on Shear Stressed Adhesive

Bondlines
W. Althof 659

Effect of Scrim Cloth on Adhesively Bonded Joints

E. C. Francis and D. Gutierrez-Lemini ••••••••••••••••• 679

PART VI. FRACTURE ASPECTS

Mechanical Measurement of Interatomic Bonding Energies at

Interfaces
E. H. Andrews •..•..•••..•••••••••••.•.••••••....•..••. 689

Review of Continuum Mechanics Factors in Adhesive Fracture

M. L. Williams •.••••••••••...•••••••••..••..••••••...• 703

Fracture Energetics of Adhesive Joints

L.-H. Lee ••..•.•••••••.•••.•••.••.•••.••.•••••.•.•.•.• 739

Fracture of Composite-Adhesive-Composite Systems

E. J. Ripling, J. S. Santner and P. B. Crosley 755

Characterization of the Fracture Behavior of Adhesive Joints

D. L. Hunston, A. J. Kinloch, S. J. Shaw and S. S. Wang 789
CONTENTS xiii

Structural Precursors to Fracture in Adhesive Joints

w. A. Jemian •.••••••••••••••••••••••••••••••••.••••••• 809

Fracture Toughness of Elastomer Modified Epoxy Adhesives

A. A. Donatelli, C. T. Mooney and J. C. Bolger •••••••• 829

A Three-Dimensional Analysis of a Butt Joint with a Flaw

R. S. Alwar and K. N. Ramachandran Nambisan ••••••••••• 839

The Influence of Layer Thickness and Internal Stresses on

the Bond Strength of Metal-to-Ceramic Joints
W. Diem> G. Elssner, T. Suga and G. Petzow •••••••••••• 855

Bond Strength Characterization of Metal-to-Ceramic and

Adhesive Joints by Critical Energy Release Rates
W. Diem, G. Elssner, T. Suga and G. Petzow •••••••••••• 871

Interfacial Properties of Filled Epoxide Resins

A. C. Moloney, H. H. Kausch and H. R. Stieger ••••••••• 883

About the Contributors 905

Index ...................................................... 923

 Part I
Substrate and Interfacial Aspects
THE ROLE OF SURFACE AND BULK CHARACTERIZATION IN THE EVALUATION OF

ADHESIVE JOINTS

W. L. Baun

Materials Laboratory, AFWAL/HLBM

Wright-Patterson AFB, Ohio 45433

Adhesive joints are often evaluated by a mechanical

test followed by exposure of a similar joint to elevated
temperature, humidity, and/or corrosive atmosphere.
After this exposure to a deleterious atmosphere, the
bonded joint is usually tested again using the same
geometry and conditions. The original load to failure
is usually taken as a measure of the quality of the
joint and performance during the accelerate.d test
determines how well the joint will hold up in service.
In the past only these numerical values were noted and
there was little diagnostic work on the failure
surfaces. More recently, however, there has been
greater emphasis on where and why a failure took place.
In order to fully evaluate failure surfaces it is
necessary to determine the locus of failure. Often this
task requires modern methods of surface
characterization, especially when the failure takes
place along a weak boundary layer (HBL). These surface
characterization probes use beams of ions, electrons or
photons and include ISS, SIMS, AES, and XPS. Examples
of the use of these techniques along with microscopy are
shown for adhesive bonding research on aluminum,
titanium and steel.

3
4 W.L. BAUN

INTRODUCTION

Numerous technologies in addition to adhesive bonding rely on

a system in which two materials are placed in intimate contact and
must remain so for the life of the resulting part. Joining
methods such as welding and soldering are examples of such
processes. Metallic and polymeric coatings on metals to provide
corrosion resistance, enhance appearance, or to take advantage of
some property of the couple, provide other examples. Following
fabrication, various mechanical tests are performed to determine
how well one material adheres to the other. Tests are also
performed under accelerated conditions to determine the durability
of the bonded materials in long time service. Often the only
information recorded from such tests is a numerical value of the
force or energy necessary to cause the bonded structure to fail.
Equally important is where the joint failed. Careful examination
of the failure surfaces by the methods to be dis 1ussed here allows
determination of the locus of failure. Good says "There is
great practical importance to the correct identification of the
failure locus. It is obvious that the measures which must be
taken to remedy an interfacial failure are different from those
which must be taken to remedy "cohesive" failure in either bulk
phase" .

DISCUSSION

Where Can Failure Occur?

In a simple two component system such as shown in Figure 11

failure may take place in one or more of the five regions. That
is, failure may propagate in either of the two bulk phases (1 and
5), at the interface (3) or in regions of A and B very near the
interface (22 and 4). It has been the long-standing opinion of
some workers that most failures which are called interfacial
(region 3) are in reality failures in a weak boundary layer (WBI,)
occurring very near the interface.

It has been shown earlier that it is not simple using visual

or even microscopic examination of adhesive bonds to determine
after testing whether an apparent adhesive failure occurred at the
interface due to improper wetting or at some new interface,
leaving behind a jhin layer of adhesive on the adherend or oxide
on the adhesive. There is a resolution limitation for most
scanning electron microscopes (SEMs) which makes very thin organic
films difficult to detect, especially when the adhesive is a pure
polymer containing no fillers of higher atomic number than the
polymer to increase contrast. Optical and staining methods have
been reported to determine the presence of adhesive films.
THE ROLE OF SURFACE AND BULK CHARACTERIZATION 5

lIm-rever, the optical technique uses the interference phenomenon,

which is applicable only to fairly thick films, certainly not to
films only a few molecules thick or boundary layers containing
both adhesive and adherend components. Staining techniques are
sensitive only to specific compounds present in the usually
complex adhesi ve systems. Several investigati ve techniques on
both sides of a joint failure are necessary to determine the locus
of failure conclusively.

It is also important to determine the locus of failure of

thin films subjected to many mechanical tests depending on peel,
scratch, abrasion, and deceleration as reviewed by Mittal •
Generally, thin polymer films on metals present the same problems
as encountered in adhesive bonding. On the other hand, evaporated
metal films, by nature of their simpler composition and more ideal
interfaces, present fewer difficulties. There are usually
differen.ces in color between the evaporated film and the substrate
which make it easier to determine the mode of failure. However,
when both metals are the same or nearly the same color or when a
metal is deposited on a thin oxide film or on a polymer, there can
be problems of interpretation of visual results.

In actual practice interfaces in adhesive bonding are not so

simple as shown in Figure 1. Many more materials and interfacial
regions exist in practical systems, as shown in Figure 2, adding
considerably to the complexity of determining the exact locus of
failure. To make it even more difficult, it has been suggested
that these interfaces are not sharply separate~ but rather, are
intermixed and diffuse. The term "interphase" probably better
describes these boundary areas. In order to determine precisely
where 'a bonded system failed, it is necessary to use sensitive
methods of analysis of chemistry and morphology on the failure
surfaces. In Figure 2 it can be seen that cohesive failure may
take place at A, B, C, or D and interfacial or near interfacial
failure may take place in the regions marked 1, 2, and 3 at each
interface.

A,FILM 1
1--------,::~-3
-~4
2
5
B,SOLID

Figure 1. Model of the interface formed in a simple two-component

system.
6 W.L. BAUN

adhesive , ___ lI',A

.. ,LV
~ ~ ....----;j
primer " , '
t
2 -'- .
- 0"'8 B
'oxide 3_.
3 .~
C

alloy roo

Figure 2. Model of the interfaces formed in a typical adhesive

bonding system.

One of the most common failures, especially when the adhering

phase is primarily organic, is depicted in Figure 3 and takes
place along a layer of trapped air or other vapor. This type of
weak boundary layer fa~ure is a failure of the first kind, as
described by Bikerman. a This type of failure is especially
prevalent in adhesive bonding where in addition to the trapped
gases, it is also possible that the curing process causes gases to
be released which cannot vent from the joint. Examples of such
failures are seen in Figure 4, where the glass reinforced polymer
Tefzel is shown from a failure specimen which had been bonded to
steel. A significant area of voids exceeding 50% of the total
area is seen in this scanning electron micrograph. Similar
effects are seen in some adhesive bonded failure surfaces,
especially where woven scrim cloth is used. Micrographs from such
wedge specimen failure surfaces using the commercial adhesive FM
400 are seen in Figure 5. In this case, the voids were caused by
improper application of tempertture and pressure cycles during
curing. Bascom and Cottington show results in which a thin
film of air is trapped and covers 50% of the interfacial area.
Incomplete displacement was found to occur if the resin does not
fully wet the adherend or if the resin does not become
sufficiently fluid during the cure cycle. These authors found
increases in bond strength of as much as 30% by proper vacuum
processing to insure removal of air.
THE ROLE OF SURFACE AND BULK CHARACTERIZATION 7

air or other
trappad gasses

Figure 3. Failure of the first kind along region of trapped air

or other vapor.

Figure 4. Failure surfaces from a steel lap shear specimen bonded

with Tefzel.

What Characterization Methods are Applicable?

Each characterization method is based on an intrinsic

property of the atoms and molecules on the surface. Such atomic
and molecular surface properties are listed in Table I along witg
a partial list of techniques which use these properties. Park
has reviewed these methods and others and discusses the
limitations as well as the advantages of each. He emphasizes in
his critical review that in a field as new as surface
characterization it is not surprising that much remains to be
learned about the effective use of these tools. In the present
paper we consider those techniques which show promise for
determining the locus of failure when an adhesive bond breaks or a
coating separates from a substrate. We have largely ignored
8 w.L. BAUN

Figure 5. Wedge failure surfaces from titanium bonded with

commercial FH 400 adhesive.

ION BEAI'I
FOR PROFILING
EXCITATION SOURCE DETECTOR

~ ~SP'CTRO",T'R FO"
I ~ ISS, AES, XPS

0+ S[ S

00 00
0000 SAMPLE
0000
Figure 6. Representation of surface characterization methods.
THE ROLE OF SURFACE AND BULK CHARACTERIZATION 9

techniques exploiting vibrational states and emphasized those

using atomic properties because the latter appear more sensitive
and easier to interpret. Figure 6 shows schematically how these
surface methods are used to determine composition of the outer
layers of material.
Ion Scattering Spectrometry. When an ion approaches a
surface~ it is most likely that the ion will become neutralized by
electrons at the surface. Some ions do survive, however, and are
back scattered from the surface after undergoing a simple elastic
binary collision. The masses of the atoms at the surface can be
determined by measuring the energy lost by the incoming ion beam.
The ISS technique is only one which analyzes the first atomic
layer at the surface. In each of the other techniques there is a
finite sampling depth.

Secondary Ion Mass Spectrometry. The momentum transfer from

the probing ion beam to the surface atom results in some atoms at
the surface being dislodged. Although many of these sputtered
species are neutral, some become ionized and can be analyzed by
conventional mass spectrometry.

Table I. Properties of Surface Atoms and Molecules and Examyles

of Characterization Techniques Utilizing These Properties •

Properties Technique

Atomic mass Ion Scattering Spectrometry

Secondary Ion Mass Spectrometry
Thermal, Electron, Ion Desorption

Vibrational States Infrared Absorption, EmIssion

Raman Spectrometry
Inelastic Electron Tunneling

Valence States X-Ray Emission

Auger Electron Spectroscopy
and
Core States Photoelectron Spectrometry
Soft X-Ray Appearance Spectrometry

Crystallography Electron Diffraction

Field Ion Microscopy

Auger Electron Spectrometry. If a material is irradiated

with an electron or photon beam of sufficient energy, an electron
is ejected from an inner level, placing the atom in an excited
state. When the atom reverts to the ground state, it may do so by
emi tting characteristic X-rays or by radiationless transitions,
the so-called Auger transitions. One major disadvantage of AES
for adhesive bonding research is that organic adhesives are very
10 W.L. BAUN

unstable under the electron beam due to localized heating.

Electrical charging is also a problem with adhesive surfaces.
Naturally AES does not suffer such limitations for thin metal
films. Information on chemical bonding is included in spectral
details and is useful as a "fingerprint" but is extremely
difficult to interpret from a theoretical point of view.

X-Ray Photoelectron Spectroscopy (Electron Spectroscopy for

Chemical Analysis. In X-ray excited photoelectron spectroscopy a
specimen is irradiated with a beam of X-rays denoted by hv in
Figure 6 (obtained by electron bombardment of, typically, aluminum
or magnesium) exciting electrons from the target that are energy
analyzed. X-ray photoelectron spectra contain chemical
information similar to that mentioned for Auger spectra, but it is
easier to interpret because it involves only one simple transition
rather than several. As the chemical environment of an atom
changes, the photoelectron spectrum undergoes changes in peak
shape, position or intenSity. Information such as this is
invaluable in determining if chemical reactions take place in an
adhesive joint following use in the field or after accelerated
testing. In thin-film technology, changes in electron spectra
allow the study of diffusion and alloying between thin films or
thin-film/substrate combinations. Inert gas sputtering may also
be used with XPS to provide elemental profile information.

What Do We Look For?

~~en we use these spectrochemical tools to determine species

on the surface, we probably use them slightly differently each
time, especially in adhesive bonding. Even when chemical and
morphological information has been collected, interpretation may
be difficult. Just how do we decide where failure has occurred?
In the typical complex adhesive bonded system, we have several
interfacial regions. Each of the materials coming together to
form these interfaces has its own individual chemical signature.
The substrate for instance usually contains alloying elements
which vary in content between the surface and the bulk. In
addition to alloying elements, surface treatments leave behind
elements characteristic of each treatment. For instance the
popular FPL (Forest Products Laboratory) etch for aluminum alloys
consists of sulfuric acid and sodium dichromate in distilled water
and leaves a detectable amount of chromium on the alloy surface.
Primers often contain anions and cations which can be followed by
spectrochemical methods. These additives (such as strontium
chromate) are usually placed in the primer to provide corrosion
protection in the coating. An example of using these elements as
tracers are illustrated in Figure 7 where an aluminum alloy
adherend which appeared to fail at the oxide/primer interface
shows appreciable chromium. Analysis of the opposite side
suggested a mixed mode failure near the suspected interface.
THE ROLE OF SURFACE AND BULK CHARACTERIZATION 11

In addition to determining what elements exist on failure

surfaces, it is most important to look at these surfaces using
optical microscopy and SEM. Dwight 8 has used the SEM
extensively, especially in all cohesive failures, . to attempt to
understand the mechanism by which failure occurs. Plastic and
brittle failure mechanisms are easily differentiated on the
polymer surface. Initial and final flaws and voids may be
determined and are of importance in evaluating joint performance.
An example of the use of the SEM was seen in Figures 4 and 5 where
micrographs show the failure to be a weak boundary layer of the
first kind, due to entrapped air or other vapor. Often we
consider WBLs to occur only during bond processing but in
actua1i ty WBLs may be created by numerous processes including
surface preparations to the original alloy or metal. An example
of such chemistry changes. is seen in Figure 8, the elemental
profile prepared from AES data from a 2024 aluminum alloy treated
with a sulfuric acid"-chromate etch. As can be seen in this
normalized profile, copper is concentrated in a band between the
aluminum oxide and the metal. ~milar results were shown by Sun
and co-workers also using AES. Another common weak boundary
layer produced on aluminum alloys containing magnesium (most
structural alLoys) occurs due to concentration of MgO on the
surface during heat treatment. ISS and SIMS spectra obtained from
such a surface on 2024 Al are seen in Figure 10. Although initial
bonda'bility is often not affected by such an oxide layer, long
time durability (especially under conditions of high humiMty) is
severely reduced.

K
Cr'

ISS
o AI

~· ·e -
+ S IMS

Figure 7. ISS and SIMS data from aluminum alloy failure surface.
12 W.l. BAUN

Surface characterization techniques have been applied to a

wide variety of failure surfaces obtained from several mechanical
testsllsuch as lap shear, peel, crack opening and three-point
bend.
Failure surfaces from thin and thick-thin adherend wedge
tests have been analyzed by the ISS-SIMS technique. The wedge
test provides information about adherend surface preparation in
that it is sensitive to different surface treatments and can
discriminate between bonding processes that give good and poor
service performance. The specimen consists of two thin adherends
or a thick-thin configuration to concentrate stresses along the
interface. A wedge is driven into the bondline and the position
of the crack leading edge is determined microscopically. Then the
specimen is subjected to various external stimuli such as changes
in temperature and relative humidity. The propagation of the
crack tip is followed with time. Sometimes when the wedge is
driven into the bondline, separation of the specimen occurs over a
portion of the bondline, first causing cohesive failure in the
adhesive then apparent adhesive failure at the adhesive/adherend
interface during testing at 71°C (160°F) and 95 percent relative
humidity, followed by cohesive failure when the specimen is opened
following the test. The adherend shows no indication of adhesive
either visually or in the SEM, as seen in Figure 10 in the area
"ABC". There are slight reflectivity differences across the
surface. Perhaps staining techniques would be effective in
outlining areas which contained thin films of adhesive when they
exist. Wi th extremely thin films, however, it is doubtful if
staining would provide any information on the locus of failure
where a primer is used and therefore several interfaces exist. An
example of such a wedge test specimen is shown in Figure 10.
Spectra (not shown) indicated failure in areas a and b near the
oxide/adhesive interface. ISS/SIMS data shown here for area c
along with depth profiles suggest that with time the failure locus
changes to a failure either in the oxide or near the oxide/metal
interface. The appearance of chromium (from the original etch) in
area c and depth profiling data provide important clues to the
exact location of the failure.

Some failures which could be classed as pure adhesi ve or

cohesive have been examined by ISS/SIMS. Adherend surfaces which
were obviously not wet by the adhesive showed no trace of the
adhesive on the adherend. Often these surfaces were 'dirty" and
showed a thin layer of contaminating elements on the adherend.
This kind of failure probably should not be considered adhesive if
proper bonding between the two surfaces never occurred. An
example of this phenomenon is seen in Figure 11. ISS and SHtS
spectra are shown for a low carbon steel surface following
debonding of the thermoplastic Nylon 12 from the surface. The
first ISS spectra show only weak peaks near potassium and iron.
THE ROLE OF SURFACE AND BULK CHARACTERIZATION 13

The corresponding SIMS data show strong sodium and potassium peaks
compared to iron. As sputtering continues, sodium and potassium
decrease and iron increases. As was indicated earlier, this
segregation of monovalent ions to the surface of adherends which
have been heated is a common occurrence. When the temperature of
adhesive bonding materials was raised to approximately the
adhesive cure temperature+(121 ~o 177°C) ~250 to 350°F), the very
mobile ions such as Li, Na, and K were concentrated on
the surface. Perhaps if the conditions for bond failure under
water attack were those for hydrolysis, the diffusion of alkali
ions to the surface would increase the osmotic potential and
enhance the destruc~~ve ingress of water at the interface.
Gledhill and Kinloch have studied the iron-epoxy system and
have found cohes;i. ve failure under dry conditions and adhesive
failure under wet conditions. They show that substrate corrosion
is not an operative mechanism in environmental failure but,
rather, a post-failure mechanism.
Appendix 1 is a background bibliography of a few of the many
publications which pertain to this subject. These examples
illustrate test methods, problems, and surface characterization
techniques related to the subject at hand.

Q)
.2
<II
>
.>t.
<II
Q)
Co

B
.>t.
<II
Q)
\
,,
Co
.. - .......... _----
(}O 4· B· 12' 6· 20'

time(min)

Figure 8. Elemental profile from AE'S data for 2024 aluminum

alloy.
14 W.L. BAUN

..
N •

r;J ..P.- - - -

Figure 9. ISS and SIMS data from "As Received" 2024 aluminum
alloy showing mill scale .

. / ·COneslv.· ' "

adnerend from
/" In ter18c lal "
I,,"wedoe ... ,

A.I

ISS
E/Eo
amu

Figure 10. Failure propagations in wedge test specimen.

THE ROLE OF SURFACE AND BULK CHARACTERIZATION 15

Na

ISS

SIMS

Ci N9 : ~I
lb io Na3b
'!)""e -- -

Figure 11. ISS and SIMS data for steel side of nylon/steel bonded
lap shear specimen.

CONCLUSIONS

Spectrochemical techniques such as ISS, SHIS, AES, and XPS

combined with microscopy can be usually used to gain a clear
picture of where an adhesive joint has failed following testing or
service. Unfortunately thi.s information still does not always
tell us the exact mechanism of failure. For instance, many
failure surfaces, especially those exposed to high concentrations
of water vapor, show elements of both adhesive and adherend. Such
a re~ult would suggest the presence of a weak boundary layer, but
Good shows that the transfer of some material from one phase to
the other does not prove the existence of a WBL at or adjoining
the interface before failure. He does not rule out the WBL, but
states only that it is not a universal concept which explains all
apparent interfacial failures. Regardless of these interpretive
difficulties, chemical information about the surface certainly
will help clarify the mechanism of failure in adhesive bonds.
16 W.L. BAUN

APPENDIX I
PARTIAL BIBLIOGRAPHY RELATING TO LOCUS OF FAILURE AND ADHESIVE
BONDING.

R. A. Gledhill and A. J. Kinloch, J. Adhesion 6, 315 (1974).

M. Gettings and A. J. Kinloch, J. Mater. Sci. 12, 2511 (1977).
W. J. Van Ooij, Surf. Sci. 68, 1 (1977). --
L. H. Lee, Editor, "Adhesion Science and Technology", Plenum
Press, New York, 1975.
W. D. Bascom and R. L. Cottington, J. Adhesion 4, 193 (1972).
W. L. Baun, N. T. McDevitt, and J. Solomon in- "Surface Analysis
Techniques for Metallurigcal Applications", pp. 86-101 ASTM STP
596, ASTM, 1976.
\01. Bascom, C. Timmons, and R. L. Jones, J. Mater. Sci. ..!.Q., 1037
(1975) .
W. L. Baun, J. Adhesion 7, 261 (1976).
W. L. Baun, in "Characterization of Metal and Polymer Surfaces"
L.-H. Lee, Editor, p. 375, Academic Press, New York, 1977.
N. J. DeLollis and O. Montoya, J. Appl. Poly. Sci., 11 983 (1967).
J. M. Burkstrand, J. Appl. Phys. 52, 4795 (1981). --
K. L. Mittal, Poly. Eng. Sci. 17,~67 (1977).
W. L. Baun, in "Surface Treatments of Plastics and Metals", D. M.
Brewis, Ed.,]). 45, Applied Science Pub., London, 1982.
H. E. Bair, S. Matsuoka, R. G. Vadimsky and T. T. Wang, J.
Adhesion 3, 89 (1971).
W. L. Baun, Appl. Surf. Sci. 4, 291 (1980).
L. B. Sargent, ASLE Trans. 21-280 (1978).
A. J. Kinloch and N. R. Smart, J. Adhesion, 12, 28 (1981).
J. M. Burkstrand, J. Vac. Sci. Technol., 161072 (1979).
M. Gettings and A. J. Kinloch, Surf. Interface AnaL,.!., 189
(1979).
F. Yamamoto, and S.Yamakawa, J. Appl. Poly. Sci., 25 2427 (1980).
W. L. Baun in "Industrial Applications of Surface Analysis" L. A.
Casper and C. J. Powell, Editors, ACS Symposium Series No. 199,
pp. 121-143 American Chemical Society, (1982).
K. L. Mittal, Pure Appl. Chern. 52, 1295 (1980).
W. L. Baun, in "Adhesion Aspects of Polymeric Coatings", K. L.
Mittal, Editor, pp. 131-146, Plenum Press, 1983.

REFERENCES

1. R. J. Good, in, "Adhesion Measurement of Thin Films, Thick

Films and Bulk Coatings", pp. 41-54, ASTM STP 640, K. L.
Mittal, Ed., ASTM, Philadelphia, (1978).
2. J. J. Bikerman, ibid. pp. 30-40; and many earlier
publications including J. J. Bikerman, "The Science of
Adhesive Joints", 2nd Edition, Academic Press, New York,
(1968).
THE ROLE OF SURFACE AND BULK CHARACTERIZATION 17

3. W. L. Baun, in .. Adhesion Measurement of Thin Films, Thick

Films and Bulk Coatings" ,K. L. Mittal, Editor, pp. 41-53,
ASTN, Philadelphia, PA, 1978.
4. K. L. Mittal, Electrocomponent Sci. Tech., 1. 21 (1976).
5. L. Sharpe, in "Recent Advances in Adhesion"~ L.-H. Lee, Ed., pp.
437-453, Gordon and Breach, New York, (1973).
6. w. D. Bascom and R. L. Cottington, J. Adhesion 4, 193 (1972).
7. R. L. Park, in "Surface Analysis Techniques for Metalurgical
Applications", R. Carbonara and J. Cuthill, Eds. pp. 3-18,
ASTM, Phila., PA, (1976).
8. D. W. Dwight, J. Colloid and Interface Sci. 59, 447 (1977).
9. T. S. Sun, J. M. Chen, J. D. Venables and ~Hopping, Appl.
Surf. Sci. 1, 202 (1978).
10. A. J. Kinloch, H. E. Bishop and N. R. Smart, J. Adhesion 14,
105 (1982). -
11. N. T. McDevitt and W. L. Baun, these proceedings, pp. 381-394
12. R. A. Gledhill and A. J. Kinloch, J. Adhesion ~, 315 (1972).
PRACTICAL ADHESION MEASUREMENT IN ADHERING SYSTEMS A PHASE

BOUNDARY SENSITIVE TEST

A.A. Roche, M.J. Romand and F. .

S~doroff *
Applied Chemistry Department (CNRS, ERA # 300)
University Claude Bernard - LYON I
69622 Villeurbanne Cedex, France

* Solid Mechanics Department

Ecole Centrale de LYON
69130 Ecully, France

The strength of adhesive bonds is assessed by

means of mechanical tests in which an increasing load
is applied until failure occurs. This paper describes a
three point flexure test using a single adherend/adhe-
sive structure. Contrary to what occurs with usual
tests such as single or double lap shear and peel tests,
this configuration is sensitive to the adherend surface
chemistry variations and adhesive cure conditions and
thus allows evaluation of the properties of the adhesive
bond joint interphase. Data are presented which demons-
trate the flexure test sensitivity to the effects of
surface prebonding treatments on interphasial mechanical
properties, such as rigidity modulus. Photoelastic iso-
chromatic fringes are recorded simultaneously with
flexure test data as a means to monitor stress distri-
bution, failure initiation and crack propagation. This
test shows that the stresses are uniformly distributed
along the specimen length and remain parallel to the
interfacial region and insure failure totally along
the adherend/adhesive interface.

19
20 A.A. ROCHE ET AL.

INTRODUCTION

Industry is continuously seeking ways of producing durable

and higher performance materials at rower costs. By way of example,
adhesively-bonded structures are extensively developed and used
in the aircraft and aerospace industries. Their advantages over
mechanical fastening structures are well known, e.g., weight sa-
ving, aerodynamic and environmental resistance improvements and
stress concentration reduction.

In single adherend/adhesive combinations - called adhering

systems - various interfaces and interphases are created in addi-
tion to the bulk materials. Consequently, under stresses, failure
can occur at any of the interfaces, interphases or bulk phases,
and it is suggested that the bond strength in such systems be
expressed as practical adhesion. For a detailed discussion of the
concept of practical adhesion and how it relates to the interface/
interphase between the adhesive (coating) and adherend, see ref.
la, b, c. To improve bond joint performance, it is, then, very
important to find out "WheJLe, how an.d why the bon.d 6aili".

Until recently, most adherend prebonding treatments (chemical

etching, conversion, anodization, ••• ) have been developed mainly
empirically, but today, many scientists are trying to establish
why some prebonding processes yield better results than others.The
present work shows firstly, how mechanical interphasial properties
can be obtained using a flexure test and an appropriate specimen
geometry ; and secondly, how analytical information from failed
surfaces can be correlated with bond performances.

EXPERIMENTAL

Adherend sheets, prepared by machine process to provide iden-

tically sized strips (50 x 10 x .4 mm), are submitted, prior to
bonding, to different surface treatments.

A single adherend-epoxy adhesive specimen is used. The speci-

men preparation fixture is shown in Figure I. The adhesive forming
mold (Figure I a) is made of RTV-501. It is held against the adhe-
rends by a clamping assembly (Figure 1 b and c). An exact amount
of adhesive is applied to each adherend with a syringe (Figure
I d). Care is taken to ensure that no bubbles evolve within the
adhesive near the interfacial region. A similar specimen prepara-
tion fixture has been previously described 2,3.

Figure 2 is a schematic drawing of a specimen under applied

load after debonding along the adherend/adhesive interface.
A three point flexure tester (INSTRON 1102 table model),
PRACTICAL ADHESION MEASUREMENT 21

Figure 1. Specimen preparation fixture: a) forming mold,

b) metallic holder, c) metallic retention plate, d) complete
assembling.
22 A.A. ROCHE ET AL.

load
metal
failure

Figure 2. Illustration of the single adherend/adhesive test

specimen under applied load.

fitted with a cross-head speed reducer and an appropriate load

cell, is used. The cross-head displacement speed is typically 0.5
rnrn per minute and the loading sensitivity is as low as 100 N full
scale. The bonded single adherend specimen is mounted in the tester
behind a light polarizer (Figure 3).

For each tested specimen, the following can be observed simul-

taneously: - the recording of the load versus displacement
- the pattern modification of the photoelastic fringes
through the adhesive bulk.

Figure 4 shows a typical load-displacement curve and the

fringe pattern modification occurring during testing. The photograph
numbers correspond to the various points along the load-displacement

Figure 3. Specimen mounted in the three point flexure test appara-

tus fitted with a light polarizer.
PRACTICAL ADHESION MEASUREMENT 23

Figure 4. Photoelastic isochromatic fringes recorded during the

testing of a single titanium adherend specimen.
24 A.A. ROCHE ET AL.

curve at which they have been recorded. The failure front is

pointed out by an arrow.

The following information cap be derived from these experiments

- the slope of the load-displacement curve
- the ultimate load before failure
- the stress distribution within the adhesive bulk
- the failure initiation and propagation.

SYSTEM MECHANICAL ANALYSIS

As a mechanical system, the specimen can be considered as an

inhomogeneous beam with variable cross-section, simply supported
at both ends (x = ± l/2) and subjected to a vertical load P for
x = 0 (Figure 5).

As a first approximation, this problem can be analyzed within

the framework of the classical flexure theory for beams 4,5. The
bending moment M along the beam is given by

M= P (x - ~) for o~ x ~ l/2
2 2
- ~ (x + ~) for - i?/2 ~ x ~ 0
2 2
and the elastic energy of the beam is
W =/ + l/ 2 M2 dx
_ lI2 2H
=/ II
0
2 M2 dx
H

where H is the bending stiffness of the beam.

p ...

Figure 5. Specimen geometry and cross-section.

PRACTICAL ADHESION MEASUREMENT 25

Em bm ~ n )I,

r=
H
~
= HS
Em 1m
12
for

for
T~ Ixl

o ~ Ixl
~

~
2
n
2
where Hm = Em 1m = (EI)m is the usual bending stiffness for the
rectangular metal sheet while HS is the bending stiffness of the
metal/interphase/adhesive composite system.

The elastic energy of the beam is then given by

laO -n 2
. )I, )I,
p2 )I, p2 )I, 2
- (x - _)2 dx + - (x --)
W
4 HS 2 /o2_n & 4 IIm 2
dx
2

According to Castigliano's theorem, the vertical displacement 8 for

x = 0 is

ClW P )1,3 [£3 o - £3) ] P-

8 = ClP +
48 Hm HS k
48 HS HS
k n
)1,3 o - £3) + n£3 Hm
which shows that for a given geometry and for a given Em the global
stiffness k (P = k8) of the system only depends on HS' This rela-
tion can be inverted to give

H = -
k )1,3
s 48 (1 - £3) 48 Hm
[, _H3 !;3J'

or
P )1,3
(EI)S = -
486
o - £3) [, P ~3
48 8 (EI)m
,3 ] (1)

It is quite clear that this approach is a very crude one since

it neglects transverse shearing, edge effects and others and only
gives a rough evaluation (in fact an excess value) for k. However
it clearly shows that the slope of the load-displacement curve is
directly related to (EI)S and, since all other parameters are kept
constant, to the quality of the phase boundary layer. In fact, for
all the different tested specimens, the whole geometry, the adhe-
rend size, the formulation, size and curing process of the adhesive
are precisely controlled and kept constant. Under these conditions
the interphase (metal/adhesive) properties vary with respect to
26 A. A. ROCHE ET Al.

the different possible reactions between adherend surface and adhe-

sive during bonding. The above relationship is verified (see Figure
6) by considering experimental values of k for various precondi-
tioned titanium adherend/adhesive systems (Table II) versus calcu-
lated (EI)S values according to Equation (1).

Starting from a similar analysis along the line of the

80

70

60

60 80 100

Figure 6. Experimental k values versus calculated (EI)S values.

PRACTICAL ADHESION MEASUREMENT 27

classical flexure theory,an approximate value can be obtained for

the value Hp of HS in the ideal case of a perfectly adherend
contact. Starting from Navier-Bernoulli assumption 4 ,5 (the cross-
section remains plane and normal to the neutral line) the longitu-
dinal strain E can be written as

E =K (y - Yo) K = R-1
where R is the curvature radius of the beam and Yo the neutral
line position (i.e. the line without length variation). The stress
o is then given by :

0 E K (y - Yo) o ~ Y ~ ha (adhesive)
a
Em K (y - Yo) ha ~ Y ~ ha + hm (metal)

ffsOdS = K { b a Ea fa
The value of Yo is obtained from the condition

ha (Y-Yo) dy + bm Em
1 ha+hm
(y - Yo) dY} =0
ha
which express the fact that there is no longitudinal force. The
bending moment is then obtained as

H = b
p a

A somewhat lengthy calculation then leads for Hp to the following

value 6
E E b h3 E b h E b h h h 2
a mmm aaammm (-.E!. + ~)
H + -~12:-= + =E~b---:;-h----="-+--'E=-b-:::::-O;-h::':'
P 12 2 2
a a a m m m

which is found to be much higher than the values of HS obtained

from experiments. A reason for this may be the very hlgh normal
stresses induced in the very thin bonding layer and which cannot
be analyzed by such a simple approach and which requires some
kind of finite element analysis. Without such an analysis it is
not possible to go beyond the qualitative discussion presented and
to get quantitative information about the bonding layer (thickness,
mechanical properties).

As for the photoelastic fringe patter~it essentially describes

the initial residual stresses and strains and is therefore related
to the specimen preparation. Its evolution during the loading
sequence qualitatively agrees with the above analysis.
28 A.A. ROCHE ET AL.

RESULTS AND DISCUSSION

For a set of identical adherend/adhesive systems using the

same sample geometry and size, a change in the slope of the load-
displacement curve must be correlated with a change in the inter-
facial properties when the adherend has been subjected to various
surface conditioning treatments.

Table I depicts a repeatability test. The rigidity modulus

(EI)S and ultimate load values are given for a set of identically
prepared specimens. Relative standard deviations for these values
are shown to be 1 % and 10 %, respectively. Effects of the sample
exposure to room atmosphere before testing are also shown. Such a
statistical result takes into account reproducibilities of the
mechanical tester, specimen sizing and preparation, and adherend
surface state.

Table II gives system rigidity modulus (EI)S and ultimate

load values for a set of titanium adherends submitted to various
chemical etchings. A change in the ultimate load and rigidity
modulus (EI)S values is observed for different titanium pretreat-
ments. As seen previously, such changes cannot be explained in
terms of the statistical error ; and therefore are characteristic
of the different mechanical properties of the interphase created
after titanium surface conditionings. Also listed in Table II are
the respective adherend surface oxide thicknesses determined from
a set of identically prepared specimens by Auger sputter profile
analysis 7 • The optimum (EI)S value is shown for the thicker Ti02-
like oxide layer. The specimen subjected to treatment 4 has the
thickest film layer but a lower (EI)S' The relevant load-displace-
ment curve shows that failure occurs in a slow, step-wise manner.
In this case, the surface layer is fo~~ to be a mixture of tita-
nium oxide and titanium fluoride TiF4 • As already reported
for Ti-6Al-4V alloy bonded with another adhesive 2 such specie
appears to be responsible for this type of failure propagation
process and interphase properties.

CONCLUSION

The results reported here show that a three point flexure

test on a single metal adherend/adhesive bonded structure is sen-
sitive to the adherend surface chemistry and topography, and
therefore, to the material properties of the so-created interphase.
According to the qualities of an "ideal test" as attributed by
K.L. Mittal 1 , the test described here fulfills more than half of
the different requirements and therefore appears to be undoubtedly
an important test.
PRACTICAL ADHESION MEASUREMENT 29

Table I. Repeatability of the Ultimate Load and Rigidity Modulus

for a Set of Fluoro-ammonium Etched Titanium Bonded to the CIBA
Epoxy (LY 564) and Effect of the Exposure to Room Atmosphere before
testing.
Exposure time
1 week 10 weeks
(room atmosphere)
Sample number 1 2 3 4 5 6 7 8
~ltimate Load (N) 18.6 21.4 21.7 21.7 17.7 17.7 19. 117.6
20 ± 10% 18 ± 10 %*
System rigidity
j:nodulus 129 1127 I 127 I 130 I 127 I 130 94 I
94
(EI)s(l03N.mm 2) 128 ± 1 % 94 ± 1 %*

* Relative standard deviation based on the first six values

Table II. Effects of the Adherend Prebonding Treatments on the

Values of Ultimate Load, Rigidity Modulus and Surface Oxide
Thickness as Obtained by AES Sputter Profile Analysis.

~ode Treatment Ultimate System Adherend Surface

Load (N) Rigidity surface Oxide
Modulus composition Thickness
(EI)S (nm)
(l03N.mm 2 )
1 Degrease 10.3 91 Ti02 17
NaOH 6
2 8.3 78 Ti02
(Alkaline)

3 HN03/HF 13.0 81 7
(Fluoro-ni tric) Ti02

4 Na3P04/NaF/HF 10.1 87 TiF4 + oxide 83

(Fluoro-phosphate)

5 NH4HF2 18.7 94 Ti02 20

(Fluoro-ammonium)

6 H2S0~/cr03 7.9 78 Ti02 14

(SuI o-chromium)
HN0 3 /HF/H Z02/NH 4HF2
7 (Fluoro-m. tro- 8.3 64 TiOZ 7
ammonium)
NaOH/HZ0 2 (65°C) 16
8 7.2 86 Ti02
(Hot alkaline)
30 A.A. ROCHE ET AL.

ACKNOWLEDGMENT

Acknowledgments are made to W.L. Baun and J.S. Solomon for

their helpful discussions.

REFERENCES

la. K.L. Mittal, in "Adhesion Measurement of Thin Films, Thick

Films and Bulk coatings", K.L. Mittal, Editor, ASTM STP 640,
pp. 5-17, American Society for Testing and Materials,
Philadelphia, 1978.
lb. K.L. Mittal, Electrocomponent Sci., Technol., 3, 21 (1976).
lc. K.L. Mittal, Pure Appl. Chern., 52, 1295 (1980)~
2. A.A. Roche, A.K. Behme and J.S.:Solomon, Int. J. Adhesion
and Adhesives, 1., No.4, 249 (1982).
3. A.A. Roche, J.S. Solomon and M.J. Romand, in "Microscopic
Aspects of Adhesion and Lubrication", J.M. Georges, Editor,
Tribology Series 7, pp. 333-342, Elsevier, 1982.
4. S.P. Timoshenko, D.H. Young, "Theory of Structures",
McGraw Hill, New York, 1965.
5. W. Fliigge, "Handbook of Engineering Mechanics", McGraw Hill,
New York, 1962
6. A.A. Roche, Thesis, Lyon, France, 1983.
7. A.A. Roche, Air Force Wright Aeronautical Laboratories, Dayton,
OH, Technical Report AFWAL-TR-80-4004, 1980.
8. A.A. Roche, J.S. Solomon and W.L. Baun, Appl. Surf. Sci., L,
83 (1981).
9. A.A. Roche, M. Charbonnier, F. Gaillard, M. Romand and R.
Bador, Appl. Surf. Sci., ~, 227 (1981).
EFFECT OF A PHENOL FORMALDEHYDE RESIN OF THE ADHESION OF A

POLYSULFIDE SEALANT

R. Ramaswamy and P. Sasidharan Achary

Polymers and Special Chemicals Division

Vikram Sarabhai Space Centre
Trivandrum 695 022 (India)

The results obtained in a study to evaluate the

effect of a phenol formaldehyde resin on the adhesion
of a polysulphide sealant is presented. The adhesion
of the sealant is improved significantly by the incor-
poration of phenol formaldehyde resin. This is expla-
ined on the basis of increased boundary layer strength
and the results obtained are compared with the 'theory
of attachment site'. Addition of phenol formaldehyde
resin to the sealant caused an increase in cross-link
density, ultimate tensile strength, stress at 100%
elongation, hardness and a decrease in ultimate elong-
ation and water absorption. The observed property
changes are explained on the basis of polymer-filler
interactions. Higher concentrations of phenol formal-
dehyde resin in the sealant caused a decrease in bond
strength, cross-link density and mechanica~ properties.
This is attributed to the lesser extent of cure of poly-
sulphide resin.

INTRODUCTION

Polysulphide sealants which find major utility as elastic

sealants in aircraft fuel tanks, aircraft runways, ship building
and building construction are formulated from liquid polysulphide
polymer, fillers, adhesion promoter, plasticizer and curing agent.
The liquid poly sulphide polymer generally used is mercaptan termi-
nated having the structure,
HS-C H -O-CH -O-C H (-S-S-C H -O-CH -O-C H). - SH
2 4 224 2 4 2 2 4 n

31
32 R. RAMASWAMY AND P. SASIDHARAN ACHARY

with a proportion of thiol terminated branched chains. By the oxi-

dation of thiol group, the liquid polymer is cured to a rubber.

In order to perform as a sealant, good adhesion (in this paper

the term 'adhesion' refers to 'practical adhesion' 1,2) of the sea-
lant to the substrate is an essential requirement. However, poly-
sulphide liquid polymer, in view of its low polar character, exhi-
bits poor adhesion. In an earlier paper 3 , the probable reasons for
poor adhesion and possible mechanisms of adhesion development in a
poly sulphide sealant have been described. It was shown that by
coreaction.of poly sulphide resin with epoxy resin before compounding
into a sealant, adhesion of the sealant could be increased signfi-
cantly. Core action with epoxy resins results in introduction of
polar hydroxyl groups in the poly sulphide backbone resulting in
increased adhesion.

In the present study,we have investigated the effect of a resol

phenol formaldehyde resin (PF resin) on the adhesion of a polysul-
phide sealant composition to aluminium substrate. The substrate
studied was B5l SWP aluminium. PF resin is chosen because of its
high polarity and commercial use in adhesives and coatings. Also,
it has been reported that the use of specific phenolic additives
improve the adhesion of polysulphide sealants 4 .

EXPERIMENTAL

Polysulphide sealants were prepared according to the formula-

tion given in Table I. The constituents in the base component
(Component A) are mercaptan terminated liquid polysulphide resin,
PF resin and fillers. The constituents in the curing agent (compo-
nent B) are dibutylphthalate which is a plasticizer, manganese di-
oxide and diphenyl guanidine which causes curing of the poly sulphide
resin.

Table I. Composition of the Sealant.

Component A
Liquid poly sulphide polymer lOOg
Titanium dioxide 40 g
Silica 10 g
PF resin (resol) varied from 0 to 5 g

Component B
Manganese dioxide 12.5 g
Dibutyl phthalate 4.0 g
Diphenyl guanidine 2.5 g
EFFECT OF A PHENOL FORMALDEHYDE RESIN 33

The po1ysu1phide resin used in this study was prepared in the

laboratory by a conventional procedureS from dich1oro diethy1 formal
and sodium po1ysu1phide (Na S , x = 2 to 2.2) with 2 mole % trich1oro
propane as a cross-linking ~g~nt. The PF resin (reso1) was also
prepared in the laboratory from phenol and excess formaldehyde using
sodium hydroxide as catalyst and was used as 70% solution in methyl
ethyl ketone. Other ingredients used in the formulation are commer-
cially available materials. A single batch of raw materials was
used throughout the study to minimise variations in the ingredients.

Component A (resin portion) with various concentrations of PF

resin was prepared by mixing the ingredients in a three roll mill.
A large batch of Component B (curing agent) was prepared separately.
Component A is a homogenous,viscous, Newtonian fluid white in colour
with a yield value of 7.15 N m- 2 (as indicated by Rheogram). Compo-
nent B is a soft paste black in colour.

Appropriate amounts of Component A and Component B were mixed

together, adhesion test specimens and slabs prepared as per proce-
dure reported ear1ier 3 . The specimens and slabs were cured for 24
hours at room temperature followed by 24 hours at 70°C.

Adhesion of the sealant was measured as the force required to

separate the sealant from the substrate (aluminium) in terms of
shear and peel bond strengths. The force required to separate the
sealant from the substrate was determined in INSTRON at a crosshead
speed of 5 em/min. The nature of failure was visually observed and
recorded.

Stress-strain properties were measured in INSTRON at a cross-

head speed of 5 em/min. from the dumbe11 cut from the slabs. Ulti-
mate tensile strength, stress at 100% elongation and ultimate elong-
ation were calculated from the stress-strain curves obtained. Hard-
ness was determined from the slabs by Durometer.

The cross-link density of the sealant was measured by swelling

method. Cured samples of the sealant were immersed in toluene and
they were allowed to swell to equilibrium for 24 hours. The swollen
samples were compression tested at a cross-head speed of 0.5 em/min.
in INSTRON, keeping the sample immersed in toluene throughout the
testing. The cross-link density was calculated using Equation (1)

(1)

where Ve is the cross-link density in mo1es/cc, R is the gas con-

stant, T is the absolute temperature and C1 is the elastic consta~t
defined by modified Mooney re1ation 6 which was determined 7 by us~ng
34 R. RAMASWAMY AND P. SASIDHARAN ACHARY

Equation (2)

(2)
2(h 3 _ h 3)
D s

where ho is the height of the unswollen sample, h is the height of

the swollen sample and h =h ~h where ~hsis the change in
height for a compressionDload 3f cr , determined from the graph.

Water absorption was measured with specimens cut from slabs of

constant thickness between 0.2mm to 0.3mm. Approximately 10 g. sam-
ples were cut, weighed accurately and immersed in 100 ml distilled
water at 30°C. The weight changes were determined after blotting
dry with filter paper.

The cure rate of the sealant was determined from the measurement
of pot life using a gel timer.

RESULTS AND DISCUSSION

Adhesion of the polysulphide sealant is improved significantly

I/l
C>16
~
15
~
11.
0
« 13
0
2 12 1 SHEAR FORCE (KG/Cm 2 )

z 11 I PEEL FORCE (KG/Cm )

0
10
I rN1ERFAC IAL FAILURE
0-
« e COHESIVE FAILURE
I( 9 ( Ie MIXED FAILURE
« 8
CL
w 7
I/l
u- S
0 5
w I.
u
oc
0 3
u-
2

o '5 1 1·5 2 2·5 3 3-5 I. 1.·5 5

P F-RESIN CONCEN TRATION (ph r J

Figure 1. The effect of PF resin (Resol) concentration on the

adhesion of the sealant.
EFFECT OF A PHENOL FORMALDEHYDE RESIN 35

by addition of PF resin as seen in Figure 1. The incorporation of

highly polar PF resin into the polysulphide sealant probably helps
to remove loosely bonded water layers (WEL)8 present on the substrate
surface by a process of solvation and diffusion and establish strong
interfacial forces with the substrate. It is reported 9 that PF resin
can establish strong interfacial forces with metal surfaces through
dipole, ionic, covalent and co-ordinate interactions. The possible
interfacial interactions between PF resin and metal surfaces are
shown in Figure 2.

It is clear from Figure I that with the increase of PF resin con-

centration, both peel and shear bond strength increase accompanied
by change in mode of failure from interfacial to cohesive. This
result is comparable with the 'theory of attachment site' proposed
by Lewis and Natarajan lO • The theory proposes that as the degree of
interfacial attachment increases, the joint strength also increases
accompanied by a change to mixed mode of failure. At a certain satu-
ration level of attachment site density, an optimum boundary strength
is reached, where the boundary strength is equal to or greater than
cohesive strength. Increasing the number of attachment sites beyond
this point does not increase the joint strength as the strength of the
joint has reached the so called cohesive plateau. Figure 1 indicates

H
DIPOLE INTERACTION I
-M-O"'H-O
WITH PHENOLlC HVDROXVLS

2 IONIC INTERACTION
WITH PHENOLIC HVDROXYLS

3 COVALENT BOND FORMATION

WITH METHVLOL GROUP

4 CO-ORDINATE BOND FORMATION

Figure 2. possible interfacial interactions between PF resin (resol)

and metal surfaces.
 w
280r C!l

1B- 260
:z: ..:
0- 16 :!... 24 0
(!)
z z
w 52 22 0
a::",
:n <l 200
uE 12 (!)
w ..... Z
...J (!) o 180
Vi ~ ...J
z~ w 160
w 8
0-

0 I 2 3 4 5 0 I 2 3 {. 5
P F RESIN CO NC E NTRAT ION ( p hr I PF RESIN CONCENTRATION ( ph r I
?J
:0
1 If- T
»
~
N
»
(J)
«
-.! E w
.u ~
o ..... a:: 5' ~
o (!) 0 -<
~ 8
0- ~ 52 »
z
« z 0
7 \1'1 50 •
\1'1 0 \1'1 :u
\1'1 ~ w (J)
W <l 6
a:: (!) z '6 »
(J)
0- Z 0
\1'1 0 5
...J a::
« '6
a:::r
W I »
:0
0 I 2 3 to 5 0 1 2 3 to 5 »
P F RESIN CONCENTRATION ( p h rI P F RE SIN CONCENTRATION (phr)
z
»
Cl
Figure 3. The effect of PF resin (resol) concentration :::r
»
:0
on the mechanical properties of the sealant. -<
EFFECT OF A PHENOL FORMALDEHYDE RESIN 37

that the cohesive plateau is reached at 3 phr of PF resin. At and

above this PF resin concentration, the boundary layer strength is
greater than strength of the sealant and cohesive failure is observed.
However, at higher levels of PF resin the peel and shear strengths
decrease due to decrease in mechanical properties as seen in
Figure 3.

The effect of PF resin on the cross-link density of the sealant

is shown in Figure 4. We see that upto 3 phr the cross-link density
increases ana then decreases. Now the question arises, whether the
increase in cross-link density is due to the additional cross-links
formed in the polymer matrix or bonding with the filler surface. A
separate study indicated that in the absence of filler, addition of
PF resin reduces the cross-link density of the sealant (Table II).
This suggests that addition of PF resin to the sealant produces
additional cross-links with the filler surface due to the bonding
between the polymer matrix and filler surface. The data obtained
on the water absorption studies (discussed later) also supports
this view. Thus the increase in ultimate tensile strength,
stress at 100% elongation and hardness can be attributed

u 12
....u 11
UJ
..J
0 10
~
\Il
9
10

)C 8
-r
CII
7
>
..... 6
Vi
z 5
UJ
0 t.
:!II::
Z
3
..J
I
III 2
III
0
cr:
u
0
2. 3 4 5
P F RESIN CONCENTRATION (phr)

Figure 4. The effect of PF resin (resol) concentration on the

cross-link density of the sealant.
38 R. RAMASWAMY AND P. SASIDHARAN ACHARY

Table II. Effect of PF resin (resol) concentration on the cross-link

density of the polysulphide sealant (without fillers).
T-------------------------T-------T-----T-------T----------T
: PF resin(resol) : : : : :
: concentration (phr) : 0: 1 : 2 : 3 :
, " , , ,
~-------------------------~-------~-----~-------~----------~
: Cross-link density : : : : :
_t ve xlO- 5 mole/cc : 14.5: 12.6: 10.3: 5.6 :
.!.-------------------------~-------~-----~-------~-,---------~

to the increased interaction between the polymer matrix and filler

surface. But the decrease in mechanical properties at higher levels
of PF resin requires explanation of different sort. It is evident
from the measurement of pot life of the sealant (Table III) that PF
resin retards the cure of polysulphide sealant. The cure reaction
is the simple oxidative coupling of terminal - 8H groups of the
polysulphide resin 2-8H ~ 2e + -8-8- +2W. Due to the acidity of
phenol, PF resin retards the cure of polysulphide resin and can re-
sult in reduced extent of cure of polysulphide sealant. The decrease
in cross-link density by the addition of PF resin (Table II) also
indicates that PF resin inhibits the cure of poly sulphide resin.
Therefore, the decrease in shear strength, peel strength, mechanical
properties and cross-link density at higher levels of PF resin can
be attributed to the lesser extent of cure of polysulphide resin.

Figure 5 illustrates the effect of PF resin on the water

absorption of the cured sealant samples. Contrary to the expectation,
the water absorption of the cured sealant samples decrease with the
increase of PF resin concentration. Water absorption of a polymer-
filler system involves diffusion of water through the polymer and the
adsorption of water at the polJffier-filler interface. PF resin in-
creases the polarity of the polymer system; therefore intensity of
interaction between the sealant and water molecules increases which
should result in the increased diffusion of water through the sealant.
If the polymer and filler surfaces are not in intimate contact, water
will displace the polymer easily from the filler surface and get

Table III. Effect of PF resin(resol) concentration on the pot

life of the polysulphide sealant.
--------------------T---T----T----T---T---T-----T-----T---'-----T
: PF resin concen- : 0: 0.5: 1: 2 : 3 : 3.5! 4 : 4.5 :
: tration (phr) :: : : :: : : :
t-------------------t---t----t----t---t---t-----t-----t-------t
,, Pot life ,'15', 35', 70 ''105'200'
" 285 ,' 375', 480 ,'
: (minutes) :: I : :: : : :
,
~-------------------~---~----~----~---~---~-----~-----~-------~
EFFECT OF A PHENOL FORMALDEHYDE RESIN 39

A AFTER 3 DAYS
WATER IMMERSION

• AFTER 7 DAYS
'·0 WAT ER IMMERSION
0.9

..... 0·8
0
0

0.7
z
-
0
a..
a::
0
0·6
0·5
\I)
CD 0·4
4:
a:: 0·3
UJ
4: 0.2
~
a.'
a 2 3
P F RESIN CONCENTRATION (phri

Figure 5. The effect of PF resin (resol) concentration on

the water absorption of the sealant.

adsorbed around the filler particles. This will result in increased

water absorption. If the polymer and filler surfaces are in intimate
contact by chemical linkages, the appearance of water at the inter-
face is delayed and the adsorption of water at the filler surface by
the displacement of polymer is impossible until the breakage of chemi-
cal linkages occursll. Therefore, the decrease of water abso!ption of
the cured sealant samples with PF resin indicates that PF resin pro-
motes the establishment of intimate contact when exposed to water
probably by chemical interactions. Although there is not much direct
evidence, it is assumed that PF resin establishes chemical interact-
ions with high energy oxide surfaces 9 . It is also postulated that
the very good durability characteristics of phenolic based adhesives
is due to the presence of these interfacial chemical bonds 12 . There-
fore, the reduced water absorption of the cured sealant with the in-
crease of PF resin concentration can be due to the bonding between
the polymer matrix and filler surface probably by chemical inter-
actions which can show an increase in cross-link density l3.

In conclusion, PF resin improves adhesion of the polysulphide

40 R. RAMASWAMY AND P. SASIDHARAN ACHARY

sealant. It also improve£ mechanical properties and water resistance

of the polysulphide sealant. PF resin retards the cure of polysul-
phide resin and therefore extends the pot life of the sealant.
Higher concentrations of PF resin in the sealant affects the extent
of cure and therefore reduces the mechanical strength properties
and hence bond strengths.

ACKNOWLEDGEMENTS

The authors thank Dr. K.V.C. Rao, Head, Chemicals Group for
encouragement, Mr. S. Bhagwan, Propellant Engineering Division for
determining cross-link densities and Mrs. B. Syamala Aroma for assis-
tance in the laboratory.

REFERENCES

1. K.L. Mittal, Polymer Engg. Sci.,17, 467 (1977).

2. K.L. Mittal in "Adhesion Measurement of Thin Films, Thick Films
and Bulk Coatings" K.L. Mittal, Editor, ASTM STP 640, p.5, ASTM,
Philadelphia, 1978.
3. R. Ramaswamy and P. Sasidharan Achary, J.Adhesion.,ll, 305(1981).
4. J.R. Panek in "Handbook of Adhesives" I. Skeist, Editor, p.372,
Van Nostrand Reinhold Company, New York, 1977.
5. M.B. Berenbaum in "P'olyethers Part III, Polyalkylene Sulphides
and Other Polythioethers" N.G. Gaylord, Editor, p.82, Inter-
science Publishers, New York, 1962.
6. L.R.G. Treloar in "The Physics of Rubber Elasticity"
L.R.G. Trel.oar, Editor, p. 165, Clarendon Press, Oxford, 1975.
7. B.B. Moore and O.K. Thomas, RAE Technical Report No.66388,
Ministry of Aviation, Farnborough,Hants (U.K.), 1966.
8. R.L. Patrick in "Treatise on Adhesion and Adhesives"
R.L. Patrick, Editor, Vol.4, p.72, Marcel Dekker, New York, 1973.
9. J.C. Bolger and A.S. Michaels in "Interface Conversion for
Polymer Coatings" P. Weiss and G.D. Cheever, Editors, p.3,
Elsevier, New York, 1968.
10. A.F. Lewis and R.T. Natarajan in "Adhesion Science and Techno-
logy" L.H. Lee, Editor, Vol.9B, p.565, Plenum Press, New York,
1975.
ll. W.C. Wake in "Fillers for Plastics" W.C. Wake, Editor, p.l.
Ilffe Books, London, 1971.
12. A.J. Kinloch, J. Adhesion, 10, 193 (1979).
13. B.B. Boonstra in "Rubber Technology and Manufacture", C.M. Blow,
Editor, p. 252, Butterworths, London, 1971.
POLY SULFIDE-POLYURETHANE INTERFACIAL ASPECTS

A. M. Usmani

The Research Institute

University of Petroleum & Minerals
Dhahran, Saudi Arabia

We studied polyurethane coating aging and determined

polyurethane coating/polysulfide sealant interfacial
integrity using scanning electron microscopy. A rapid
microspecimen hand-pulled peel test was developed that
will find application in designing future aircraft seal-
ants. Dynamic mechanical analysis was used to follow
polyurethane aging.

INTRODUCTION

A polyurethane coating is used on fuel tank interior surfaces

prior to application of polysulfide sealant. The sealant/coating
combination works well initially and there is apparently adequate
adhesion of the polysulfide rubber to the polyurethane coating.
Upon extended aircraft usage, however, leaks do develop. These
leaks cannot be repaired by removing the old polysulfide rubber
and then applying a fresh sealant because of insufficient adhesion.

The factors responsible for poor adhesion of the polysulfide

sealant to "aged" polyurethane coating were determined by Usmani
et al. 1 Using transmission and surface reflectance infrared spec-
troscopy, we monitored concentration changes in NCO groups of
artificially aged polyurethane coatings. Heat treatment of poly-
urethane coating at 121°C for 48 hours produced a surface to which
the polysulfide sealant would not adhere.

This paper describes sealant-coating interface examination by

SEM and a hand-pull test on small specimens for rapid adhesion

41
42 A.M. USMANI

screening. It is hoped that these methods will find application in

development of future generations of aircraft polysulfide sealant.
Polyurethane reaction with polysulfide has also been treated.

BRIEF OVERVIEW OF POLYURETHANE AND POLYSULFIDE CHEMISTRY

Polyurethane

Polyurethanes are formed by condensation of an isocyanate and

a polyol. The propagation involves the nucleophilic attack of the
hydroxyl group on the isocyanate group to yield a substituted amide
ester of carbonic acid, i.e., polyurethane. Much has been written
about polyurethane chemistry. Excellent treatments of polyurethane
chemistry and technology have been provided by Saunders and Frisch,2
Dombrow,3 and more recently by Usmani and Salyer.~

The chemical reactions involved in urethane polymerization are

complex but reasonably well understood. Polyfunctional isocyanate
reacts with polyfunctional polyol containing labile hydrogen to
make the urethane structure. The formation of urethane groups is
generally accepted to be the chain propagating reaction. If the
polyol contains amine groups, a substituted urea becomes a princi-
pal product. Reactions that result in branching and crosslinking
are the isocyanate-urethane reaction leading to allophanate link-
ages, and the isocyanate-urea reaction products which produce biuret
linkages. These reactions are well known and are summarized in
Table 1.

Polysulfide

The highly reactive terminal mercaptans of liquid polysulfide

polymers can be converted into high molecular weight products very
easily at ambient temperatures by oxidation or by reaction with
other active polymers. Polysulfides have excellent solvent resis-
tance, predictable cure, good low-temperature performance, good
weathering, and excellent adhesion to many substrates. These com-
bination of properties have given polysulfide polymers a modest but
significant position as specialty polymers since their commerciali-
zation in 1929.

Polysulfide sealants have found wide acceptance in the air-

craft industry and although facing strong challenge by silicones
and polyurethanes, continue to be the dominant segment of the
elastomeric sealant market.

In a recent paper, Usmani has reviewed the chemistry, proper-

ties, compounding, processing and manufacturing, characterization
and testing, applications, and future prospects of polysulfide
POLYSULFIDE-POLYURETHANE INTERFACIAL ASPECTS 43

Table I. Polyurethane Coating Reactions.

oII
R-N=C=O + R-OH ----+ R-NH C-O-R

Urethane
H0
I II
R-N=C=O + H20 ----+ R-N-C-OH --+R-NH2 + C02
Unstable
HOH
I II I
R-NH2 + R-N=C=O ----+ R-N-C-N-R

Substituted urea
000
I) II II
R-N=C=O + R-NH-C-OR ----+ R-NH-C-N-C-OR
I
R
Allophanate

0II 0
II
0II
R-N=C=O + R-NH-C-NH-R ----+ R-NH-C-N-C-NH-R
I
R
Biuret

sealants. 5 The author has also presented a generally accepted

mechanism of cure which is as follows:

2-RSH + (0) -+ -RSSR + H20.

CHEMICAL COMPOSITION OF COATING AND SEALANT

The polyurethane coating consisted of strontium chromate,

titanium dioxide, phthalocyanine green, and talc dispersed in a
polyester solution. The other component consisted of an aromatic
diisocyanate. The NCO/OH ratio in the coating was 1.03/1.00. The
coating was ~prayed and allowed to cure for 14 days under ambient
conditions to give a 3 mil thick film.

The chemical composition of the polysulfide sealant used in

this work was as follows:
44 A.M. USMANI

Sealant Base Component %

Calcium carbonate 26.16

Titanium dioxide 3.10
Liquid polysulfide polymer 58.50
Volatile diluent 2.25

Accelerator Component

Manganese dioxide 5.53

Processing oil 3.59
Volatile diluent 0.51

SEALANT-COATING INTERFACE SEM STUDY

Events occurring at the polysulfide-polyurethane interface

are important in defining the cause of poor adhesion and ways to
improve adhesion of sealant to the polyurethane surface. For this
reason SEM was used to study the polysulfide-polyurethane inter-
face on an aluminum aircraft alloy.

The method involved preparation of polyurethane coated mini-

panels (0.5 in. by 0.5 in.), followed by application of about 100
mil thick polysulfide sealant. The sealant was allowed to cure
(two days under ambient conditions plus one day at 60°C). Then
the mini panels were conditioned at room temperature for 4 hours
followed by exposure to one of three different test environments.

• Two days at 60°C in air,

• Two days at 60°C in water, and
• Two days at 60°C in jet reference fuel (JRF).

After exposure, samples were examined for integrity at the inter-

face or the extent of debonding (desorption).

In Figure 1 SEMs of sealant-coating interface, at 3,000X mag-

nification, are shown. Most debonding was found with water. It
can be also seen from Figure 1 that water has penetrated the seal-
ant and that the morphology of the sealant has changed.

From this work we are led to believe that SEM study of the
sealant-coating interface is useful for examination of the perform-
ance of both established and new sealants.

ARTIFICIAL AGING OF POLYURETHANE AND DEVELOPMENT

OF HAND-PULLED PEEL TEST

An important aspect of our work was to artificially age poly-

POLYSULFIDE-POLYURETHANE INTERFACIAL ASPECTS 45

Figure 1. SEM of sealant-coating interface, at 3,OOOX magnification

under conditions (a) new (b) 2 days at 60°C in air
(c) 2 days at 60°C in JRF, and (d) 2 days at 60°C in water.
46 A.M. USMANI

urethane coating to produce surfaces to which the polysulfide seal-

ant would not stick. We were able to determine necessary condi-
tions to artificially age the polyurethane coating (see Table II).
After establishing the broad conditions, we proceeded to narrow
down the conditions of aging at which limited or no adhesion with
polysulfide will result. To narrow down the exposure time at 121°C
at which no adhesion with polysulfide will result we developed a
rapid screening method. Microspecimens (1 in. by 1 in.) of poly-
urethane coated panels were aged at 121°C in air for 1 through 10,
20, 30, 40, and 50 hours; the polysulfide sealant was then applied
and cured. A strong but very fine fabric leader was embedded in
the polysulfide (1" wide, ~" embedded and 1" outside to provide
grip between thumb and index finger). A dual beam razor with 3/4"
gap was used to indent the polysulfide sealant reaching the alu-
minum alloy substrate. The prepared test panel was held on the
edges by the left hand. The fabric leader was pulled by the right
hand. The same pull was used in all specimens although this is
subject to operator error. Results of hand pulled peel tests are
provided in Table III.

Table III indicates that exposure time greater than five hours
at 121°C will result in a nonsticking surface. We feel that this
technique has merits as a screening test method but its results
should be checked with the standard peel test. Advantages of the
test method are that it requires only about 109 of sealant and
that the test can be quickly run.

DYNAMIC MECHANICAL ANALYSIS (DMA)

The polyurethane coating curing was studied by observing

changes in damping properties and the glass transition temperature
(T g). DuPont 981 was used to record DMA spectra of polyurethane
coatings on glass fabric (about 75 ~m thick). The cure of poly-
urethane coating was found to be slow with steady increase in Tg
with time over a three-week period. The coating was found to
crosslink with an ultimate Tg of 80°C. The cure was accelerated
by temperature and water immersion. In Figure 2, DMA of polyure-
thane coating cured for two-weeks at room temperature is shown.
The film coated on glass fabric is then soaked in water for one-
week and the resultant DMA is shown in Figure 3. Water soaking
produces crosslinking and therefore increased Tg • This hardens
the polyurethane surface and this is one of the principal reasons
for insufficient adhesion of a polysulfide sealant to an aged
polyurethane coating.
 ""0
0
r
10° -<
STORAGE MODULUS (f)
I!I 9
C
~EJ LOSS TANGENT 8 r
+ 7 :!!
EJ 0
0 LOSS MODULUS 6 m
I!I 5
I!I -0
4 0
I!I 4 r
I:l -<
C
/J)
..J
r~ 3 IEIIEI
I:le I:l
I:l 3 :Il
m
-l
ct +' lEI I
++ .. +~++O
~ 2 + ' I!I I!I + + • 2 »
ct +
Z
Q. + m
+
/J)
Z
:::> -l
+ -I m
..J I c! 10 I- :Il
9 . z -n
9 ILl »
8 (!)
6o
8 ()
::z: 7 z
+ 0 '
+A0 7 »
6 o 0 ~ r
6 /J)
5 +0 00 »
5 /J) (f)
++0 0 000 0
4 ""0
;;;JP %00 0 4 ..J m
4 ()
-l
3 (f)

:t 2

I 01 1 1 , 1 1 , 1 1 , I II 0"2
-120 -80 -40 0 40 80 120 160 200 240 280 320
TEMPERATURE ( Co)

Figure 2. DMA of polyurethane coating cured for 2 weeks at room

temperature.
.j::.
-...J
 .po.
ex>
I~ 10°
9 (;] STORAGE MODULUS 9
8 8
7 + LOSS TANGENT 7
6 o LOSS MODULUS 6
5 5
4 4
[H] 0 8E1 (;] (;](;][] 8 0 []88
8
3 3
I/) I:l
...J
« 2 ~ 0(;] ++++++++++++++ + 2 I-
(.) Z
I/) (;]+ + W
« o C>
Q. + (;] z
+ []
«
I-
U) 8 -I
+ 0
~IO + [) 10 U)
...J 9 U)
9 Q
~
0 8 , . 0 0
. . . 8 80 8 3
0 7 7
~ +
6 6
+++
5 5
+
4 ++ 4
++
3 ++++++ .. ++++++ 3
~
~ 00
2 o 2
• 0 °0%..
Q00
o ~00
.<9
o71 I
~,
~0Q9 · I 1 1 0 "
1 2
-120 - 80 -40 0 40 80 120 160 200 240 280 320
!>
TEMPE RATURE (C) s:
Figure 3. DMA of polyurethane coating cured for 2 weeks at room cCJ)
temperature followed by 1 week water soak. s:
~
z
POLYSULFIDE-POLYURETHANE INTERFACIAL ASPECTS 49

Table II. Peel Tests on Polysulfide Sealants Applied over Artifi-

cially Aged Polyurethane Coating Surface.
Sample Treatment Peel Stren~th, N/cm

5 days @ 60°C in air 104.2

12 days @ 60°C in air 86.0
5 days @ 60°C in water 106.2
2 days @ 121°C in air 0
5 days @ 121°C in air 0
5 days @ 60°C in air +
7 days @ 60°C in water
} 0
2 days @ 121°C in air +} 0
2 days @ 60 o C/95% R.H.

Table III. Microspecimen Hand-Pulled Peel Test Results.

Exposure Time at Hand-Pull Peel Test

121°C in air, hrs Result

0 CF
1 "'80% CF
2 "'70% CF
3 "'50% CF
4 "'50% CF
5 "'50% CF
6 AF
7 AF
8 AF
9 AF
10 AF
20 AF
30 AF
40 AF
50 AF

CF == Cohesive Failure
AF == Adhesive Failure

POLYURETHANE REACTION WITH POLYSULFIDE

The polyurethane reaction is slow in the coating because of

the absence of tin or other catalyst compound. Newly coated poly-
urethane surfaces therefore contain residual NCO group. These
groups can function as an internal coupling agent for the poly-
50 A.M. USMANI

sulfide sealant. The coupling reaction is shown below. The forma-

tion of thiourethane is responsible for excellent adhesion.

o
R-N=C=O + R~-SH ---+ R-NH-3-S-R~
Contained in Polysulfide Thiourethane
the cured Sealant
Polyurethane

CONCLUSIONS

We have developed an accelerated aging method for polyurethane

coatings. The hand-pull test that we developed on small specimens
is a rapid adhesion screening nethod. This method, in conjunction
with peel tests, can be utilized for developing new sealants.

ACKNOWLEDGEMENTS

The author appreciates help and support of Dr. Ali G. Maadhah

and Mr. P. F. Cooke (both of the Research Institute) in prepara-
tion of this chapter. Technical help provided by D. E. Miller
(University of Dayton) is acknowledged. Special thanks to Nihal
Ahmad for typing this work.

REFERENCES

1. A. M. Usmani, R. P. Chartoff, W. M. Warner, J. M. Butler, 1. O.

Salyer, and D. E. Miller, Rubber Chem. Technol., 54, 1081 (1981).
2. J. H. Saunders and K. C. Frisch, "Polyurethanes: Chemistry and
Technology," Vols. 1 and 2, Wiley-Interscience, New York, 1962
and 1964.
3. B. A. Dombrow, "Polyurethanes," Reinhold, New York, 1957.
4. A. M. Usmani and I. O. Salyer, Polym.Plast. Technol. Eng.,
Q, 61 (1979).
5. A. M. Usmani, Polym.Plast. Technol. Eng., ~, 185 (1982).
EPOXY RESIN WETTING OF E GLASS SINGLE FILAMENTS AS IT RELATES TO
SHEAR STRENGTH

Elisabeth J. Berger and Yona Eckstein

Owens-Corning Fiberglas Corporation
Technical Center
Granville, Ohio 43023

The work of adhesion, W ,of epoxy resin, water,

and methylene iodide on sin9"e E-glass filaments was
measured using a modified Wilhelmy apparatus. The resin
used was a bisphenol A epoxy resin, and the fibers were
coated with differing concentrations of silanes. Shear
strengths were measured using a single filament adhesion
test. It was found that increasing W of resin on the
fibers correlates with increasing sheJt strength of dry
and water-boiled composites. This result and the
results of methylene iodide and water wetting are
analyzed in terms of the geometric mean approximation
for the work of adhesion, and good agreement is found.

INTRODUCTION
It has long been ~ppreciated that wetting has a significant
effect on adhes ion 1 n compos i tes , and thus on compos i te
properties. 1 ,2 However, the nature and extent of this effect have
not been clearly defined. This work is a study of the correlation
between glass/resin shear strength and wetting characteristics of
the components of a compos ite system. No attempt is made to
interpret our results in terms of the structure of the glass/resin
interface.
Wetting experiments were done on single E-glass filaments,
coated with differing concentrations of two silanes: y-aminopropyl
(triethoxy) silane and y-methacryloxypropyl (trimethoxy) silane.
The works of adhesion of epoxy resin, of methylene iodide and of
water to these filaments were measured with a wetting balance. In
51
52 E.J. BERGER AND Y. ECKSTEIN

the following, the results from these measurements and their

correlation with the results from a single filament/resin
interfacial shear test are presented and discussed.
EXPERIMENTAL
Single filaments of E-glass (10 micron diameter) were pulled
from a one-hole bushing and coated from an applicator immediately
under the bushing. The coatings were water or dilute aqueous
silane solutions ·of concentrations from 0.2% to 2.5% by weight.
The silanes used were Union Carbide A-llOO (amino silane) and
A-174 (methacryl silane) dissolved at a pH of 3.5 (pH adjusted
with acetic acid). A release agent, octyl(triethoxy)silane from
Petrarch Systems Inc., was also used as a fiber coating in one
experiment. The fibers were not rinsed.
The wetting balance used is similar to that described by
Miller 3 . It consists of a Cahn 2000 electrobalance, an elevator
and enclosure from Rame-Hart, and a Fisher strip chart recorder.
The system is isolated from building vibrations by a Peabody air
table. A fiber is glued to a small wire hook which is hung on the
weighing ar.m of the electrobalance. The weight is tared off. The
elevator platform is slowly raised until the beaker of DER331
epoxy resin (Dow Chemical Co.) or other wetting liquid resting on
it touches the fiber. The elevator then continues rising for four
millimeters at a rate of 0.15 microns per second. The contact of
the resin with the fiber generates a force described as
(1)

where m is the measured mass, g the acceleration due to gravity, d

the fi ber di ameter, y the 1i qui d surface tens ion, and e the
contact angle. The worlvof adhesion of the resin on the fiber can
then be calculated, since
(2)

About ten fibers from each sample were measured and the results
averaged.
Liquid surface tensions were determined using a Wilhelmy
plate apparatus. The contact angle of resin against polypropylene
was measured with d% Rame-Hart NRL-I00 contact angle goniometer in
order to measure Yl v'

The glass fiber/resin interfacial shear strength was

determi ned us i ng the method deve loped by Fraser et a1 .5, and
modified by Drzal et al. 6 In short, the technique consists of
EPOXY RESIN WETTING AND SHEAR STRENGTH 53

embedding a continuous glass monofilament in a resin. The

dogbone-shaped specimen is then tensile loaded until the
fragmentation of the embedded glass filament ceases. In order to
analyze the results, the strain to failure of the resin must be
four to five times larger than the ultimate elongation of the
fiber. Since the fracture elongation of glass fiber is about 5%,
the elongation at break of the matrix must be at least 25%. This
requirement was satisfied by curing DER331, a diglycidyl ether of
bis-phenol A, with the polybasic acid EM,POL 1052 (a mixture of 55%
dibasic and 41% trimer fatty acids), using DMP-10
(dimethylaminoethylphenol) as an accelerator. The elongation at
break of this resin varies between 35 and 55%.
The glass fibers were mounted on a stainless steel collecting
fork by cementing their ends adjacent to the positioning notches.
The single filaments were embedded in the resin using a silicone
RTV-3112 resin mold. The mold consists of eight dogbone-shaped
specimen cavities (1/8 inch wide x 1/16 inch deep x 1 inch long).
Slots for the fibers are molded 1/32 inch deep through the ends of
each specimen cavity. Since the distance between the slots in the
mold is similar to the distance between neighboring notches on the
steel fork, mounting and alignment of the fiber in the mold are
readily accomplished.
The resin is injected into the mold with a disposable syringe
to a level just above the height of the mold surface. The mold is
then transferred to an oven and the resin cured. After curing,
the mold is allowed to cool to room temperature. The specimens
are easily removed by curling the silicone mold parallel to the
fiber. In order to eliminate from consideration specimens with
defects (e. g., broken fibers or improperly al igned fibers) each
specimen is examined under a microscope prior to testing. After
thi s i nit i ali nspect ion, those that passed are deformed on an
Instron at jaw speed of 0.2 inch/min to just below their ultimate
yield elongation.
Uniaxial tensile deformation of the specimens causes
successive fragmentation of the fiber within the matrix.
Theoretically, all lengths of fiber fragment should fall within
the range of Q /2 ~Q~Q, where Q is the critical length.
Assuming an evencdistribbtion of fr~gment lengths, the critical
length can be calculated from
Q
c
= (4/3) .£ (3)

where Q is the average fragment 1ength produced ina specimen

after tensile deformation. The average interfacial shear
strength, t, is calculated from
54 E.J. BERGER AND Y. ECKSTEIN

t = (/2) (oFfd/Qc) (4)

where d denotes the fiber diameter, and oFf is the average

strength of the glass fragments.
After substituting Equation 4 into Equation 3, the formula becomes
(5 )

Q, the average length of glass fragments, can be estimated to

a first approximation by dividing the gauge length of the specimen
(one inch in our speci mens) by the number of fragments produced
upon tensile deformation. oFf is calculated from the relationship
L 11m
off = OF Q (6)

where ° denotes the mean strength of gl ass fi ber of a known

length Land m is a statistical parameter describing the data
scatter obtainable from a Weibull plot of the strength
distribution of the fibers, measured prior to embedding the fibers
in a resin.
The interfacial shear strength was measured in samples both
before and after immersion in boiling water for six hours. These
measurements are denoted dry strength and wet strength,
respectively.
RESULTS AND DISCUSSION
The results of the glass/resin interfacial shear strength
measurements on specimens coated with different concentrations of
the silanes are given in Table 1. As demonstrated in Figures (IA)
and (IB), the strength of the interface is a function of silane
concentration, with the A-llOO-coated fibers reaching a maximum
dry strength at 1.0%, and the A-174-coated fibers giving the best
dry strengths at the lowest concentration tested, 0.2%. For wet
strength values, opposite trends prevailed for the two silanes.
The wet strength of the A-llOO-coated glass decreased with
increasing silane concentration (Figure 3) while the A-174-coated
samples increased in wet strength as the silane concentration
increased.
The fiber coated with only water and with the release agent
had shear strengths above most of the AllOO- and A-174-coated
fibers. Both systems may be viewed in terms of epoxy/glass
adhesion, since the release agent coating is believed to be very
nonuniform. It is known that epoxide groups are very reactive
with water and adhere strongly to the glass surface 7 . McGarry 8,
for example, reported that the epoxi de/gl ass joi nt is about as
EPOXY RESIN WETTING AND SHEAR STRENGTH 55

strong as the cohesive strength of both. He observed that

fracture itself never occurred at the interface. The large
interfacial shear strength values reported for the water and
release agent coated glass can thus be explained by the presence
of strong glass/epoxy bondings. Apparently, even a small
concentration of such bonding is sufficient to produce high
interfacial shear strength.
Also listed in Table 1 are the measured works of adhesion of
epoxy resin on glass filaments from the wetting experiments. As
the work of adhes i on increases, an increase in the i nterfaci a 1
shear strength of the dry specimens is observed. Thi s effect is
more easily observed when the dry strengths are plotted against
the work of adhesion for the resin, W (Figure 2). The better
mo 1ecul ar contact between the fi ber ancf the uncured res in, whi ch
is suggested by the higher wetting, correlates with better
interfacial adhesion in the cured composite. However, examination
of Table 1 shows that no such correlation exists between the
interfacial shear strength of the water-boiled samples and the
work of adhesion of resin on the glass fibers.

13.0

12.0

11 .0
iii
~

&.
g,
t: 10.0
!
iii...
..
01
&.
1/1
9 .0

8 .0

7 .0

6 .0+--~----"--"--"T"""'---r----r­
O~ O~ 1~ 1~ 2~ 2~ 3~

Welghl Percenl A174

Figure lAo Single Filament Interfacial Shear Strength vs.

Concentration of A-174 on Glass Filaments.
56 E.J. BERGER AND Y. ECKSTEIN

11 .0

10.0

·iii
~

.:::
g,
c:: 8.
~

...
iii
<II
.:::
<f)

4 .04--~--.---r----r--'--T"
0 .0 0.5 1.0 15 2~ 25 3~

Weight Percent Al100

Figure lB. Single Filament Interfacial Shear Strength

vs. Concentration of A-HOO on Glass Filaments.
EPOXY RESIN WETTING AND SHEAR STRENGTH 57

I
13 .0
I
I

12 .0 I
I
I
I
11.0 I
I
I
I
I
10 . I
·iii I
~
r.
C,
c: 9.0
~
1/1
~

'"41
r.
1/1 8 .0

7 .0

6 .0
I
I

5.0
68 .0 70.0 72 .0 74 .0 76.0 78.0 80 .0
Work 01 Adhesion In Resi n (dynes/ em)

Figure 2. Single Filament Interfacial Shear Strength vs.

Work of Adhesion of Epoxy Resin on Glass Filaments.
58 E.J. BERGER AND Y. ECKSTEIN

5 .5

5 .0

';;;
~ 4 .5
.r:.
D.
c:
e 4.0
iii
:...
.r:.
V>
a; 3.5
~

3.0

2.5

2 .0·~-"""----'----r---.---.------r-
0.0 0 .5 1 .0 1 .5 2 .0 .5
·2 3 .0
Weight Percent All00

Figure 3. Single Filament Interfacial Shear Strength After Water

Boil vs. Concentration of A-1100 on Glass Filaments.
Table I. Interfacial Adhesion and Wetting by Epoxy Resin,
Water, and Methylene Iodide on Silanized Glass Fibers.
W in Win Win
Coatin t wet(a) E ox Resin(b) Wa~er(b) CHaI (b)
A-1100 c
0.2% 6.8 4.8 69.2 ± 3 95.1 ± 6 68.9 ± 2
0.5% 7.1 3.6 69.5 ± 2 101. 3 ± 9 64.3 ± 3
1.0 8.7 3.9 71. 8 ± 2 108.6 ± 10 67.4 ± 5
2.5% 5.4 2.8 71. 3 ± 2 119.3 ± 10 67.9 ± 1
A-174(c)
0.2% 12.2 1.6 77.0 ± 2 111.1± 7 67.1 ± 2
1. 0% 8.2 2.6 69.8 ± 2 114.1 ± 9 64.2 ± 1
2.5% 7.1 3.4 69.7 ± 2 111. 5 ± 12 64.2 ± 1
Water 9.1 4.1 73.9 ± 2 103.3 ± 3 65.4 ± 2
Release
agent
1. 0% 8.0 69.7 ± 3
(a) Units are kpsi, errors are approximately 15%
(b) Units are ergs/cm 2 , errors are 95% confidence intervals.
(c) Silane concentrations are weight percent in water.
EPOXY RESIN WETTING AND SHEAR STRENGTH 59

In order to elucidate the glass/resin/water interactions,

wetting experiments with water and methylene iodide were conducted
on the glass fibers. These results are also shown in Table l.
The water wettabil ity of the A-174-coated fibers appeared
constant, but the A-lIDO-coated fibers showed a marked increase in
hydrophobicity (lower water W) with decreasing silane
concentrations. This corresponds toan increase in wet strength
at lower A-lIDO concentrations (Figure 3).

The degradation of the gl ass/res i n interface duri ng water

boil may be seen as a competition between water and resin for the
glass surface. Therefore, the larger the ratio of the work of
adhes i on of res i n on gl ass to the work of adhes i on of water on
glass, the greater the wet strength should be. The trend of a
genera 1 increase of wet strength wi th a greater ratio of res in
wettabi 1ity to water wettabil ity is shown for our systems in
Fi gure (4).

It is frequently assumed that the surface free energy of a

substance can be di vi ded into a polar and di spersi on component
(e.g. reference 8)

(7)

The yd is due to London dispersion forces, and the yP is due

to dipole-dipole interactions and hydrogen bonding. One of the
most prevalent theories of adhesion is that good adhesion is the
result of a close match between the polar and dispersion
components of the surface free energy of the adherend and the
adhesive 9 . A difficulty with this theory is the lack of exact
methods to measure the components of the surface free energy of
solids. Several useful approximations have been suggested,
however, including the geometric mean approximation!O, which
describes the work of adhesion as

(8)

Subscri pts s and I refer to the so 1i d and 1i qui d,

respectively. Dispersion components for liquids can be calculated
from contact angle measurements against a nonpolar surface, such
as polyethylene!!. The polar component is then the total surface
tension minus the dispers1p,1i component. If both sides of
Equation 8 are divided by 2(Yl)~ we have
60 E.J. BERGER AND Y. ECKSTEIN

4 .5

4 .0

3.5
'iii
.>L.
s;
a, 3.0
c:
~
en
0; 2.5
~

2.0

1.5

1.0+---.----r--.-----,,--r--"T""
0.59 0.62 0.65 0 .68 0 .71 0.74 0 .77

Wa resin/ Wa water

Figure 4. Single Filament Interfacial Shear Strength After Water

Boil vs. The Ratio of Work of Adhesion of Epoxy Resin on Glass
Filaments to the Work of Adhesion of Water on the Filaments.

(Y~)~(Y~)~
= (9)
(Y~)~
Thus a soli d with unknown surface characteri st i cs can be wetted
with t~okor more liquids with known ~u~fac~ ~nergy components, and
W12(y,)2 can be plotted against (Y1) 2/(Yl) 2. From Equation (9),
tRe intercept is the square root of the cri spers i on component and
the slope is the square root of the polar component.
Once the solid surface characteristics have been
approximated, the adhesion can be discussed diagramatically8. The
square root of the dispersion component is plotted against the
square root of the polar component. A circle is then drawn
through the coordinates of the adherend (phase 1) and the adhesive
(phase 3) with radius Ro and center (K,H):
EPOXY RESIN WETTING AND SHEAR STRENGTH 61

R2 = 1/4((al - a3)2 + (~l -~3 )2)

o
H = 1/2 (al + a3) (10)

K = 1/2 (~l +~3)

where
a = Ci)~
1
~ = Cl)>2

Sensitivity of thi s bond to another environment (phase 2; e. g. ,

immers i on in water) is then descri bed by the Griffith fracture
equation

16

14

12

10

~ 8
o·
i"-
K
~ 6
to
~

0
0.0 0.5 1.0 1.5 2.0

( Y~l '''/( Y'; l'Il

Figure 5. Plot for Surface Energy Calculation of Water-Coated

Gl ass. Using the geometric mean approximation, the intercept
is the square root of the dispersion component, and the slope
is the square root of the polar component.
62 E.J. BERGER AND Y. ECKSTEIN

(ll)

in which
(12)

and y is the critical fracture energy. If the immersion fluid

has c~ordinates inside the circle, there will be spontaneous
debonding, since R < R giving a negative YG' For fluids outside
the circle, the degreeOof sensitivity of the bond to the immersion
fluid is described by the magnitude of YG' The farther the
coordinates of immersion fluid are from the circle describing a
particular bond, the less the bond will be affected by the fluid.
In this work, three wetting liquids were used to determine
the surface characteristics of the fibers: water (yi = 50.8
ergs/cm 2 , Y~ = 22.0 ergs/cm 2 ), DER331 resin (yi = 10.8 ergs/cm 2 ,
Y~ = 36.7 ergs/cm 2 ), and methylene iodide (yi = 2.3 ergs/cm 2 ,
Yl = 48.5 ergs/cm 2 ). A typical plot giving surface energy of a
fiber is shown in Figure (5). The calculated polar and dispersion
components of the fibers coated with A-llOO and A-174, and the
Griffith fracture energy, are listed in Table II.

Table II. Square Roots of Polar and Dispersion Components of the

Surface Free Eneregy of Silane-coated Fibers. Griffith Fracture
1:
Energy for Epoxy Composites in Water. Units are (ergs/cm 2 )2.

Coating (Y~)~ (yp)~

s YG
A-llOO(a)
0.2% 3.68 ± 0.26 4.22 ± 0.17 9.89
0.5% 3.29 ± 0.29 4.88 ± 0.28 7.24
1.0% 3.60 ± 0.24 4.82 ± 0.27 7.74
2.5% 3.10 ± 0.37 5.87 ± 0.38 3.70

A-174(a)
0.2% 3.60 ± 0.26 5.23 ± 0.25 6.24
1.0% 2.91 ± 0.40 5.80 ± 0.37 3.87
2.5% 2.95 ± 0.43 5.68 ± 0.40 4.34

Water 3.57 ± 0.12 4.76±0.13 7.94

(a) Silane concentrations are weight percent in water.

EPOXY RESIN WETTING AND SHEAR STRENGTH 63

Circles are drawn through the coordinates of the cured epoxy

resin and coated fiber (Figure 6). The immersion fluid of
interest is water, and it can be seen that the closer the circle
is to the water coordinates, the lower the wet strength. If this
di stance, whi ch corresponds to the Gri ffi th fracture energy, is
plotted against the wet strength, a correlation can be seen
(Figure 7). As the Griffith fracture energy increases, more
energy is needed to break the adhesive bond, and the wet strength
increases.

0
o:i

q
oa 4.8 ksi
I'- water-coated, 4.1 ksi
1..Q44, 3.9 ksi
O.5~ 3.8 ksl
2.5~ 2.8 ksi
q

-
(I)

~
~

i
2
q
II)

~water
)(

.,q

q
N~+O----~3~.O----~
4D------5~D------
8.-0----~7.0----~
8D

1/2
(ganvna p)

Figure 6. Square Root of the Dispersion Component of the Surface

Free Energy vs. the Square Root of the Polar Component for Fiber
Coated with A-lIDO or Water, p-nd for Liquid Water and Cured Epoxy
Resin. Units are (ergs/cm2)~. The circles describe the bond
between the fiber and the resin, with a smaller radius indicating
a closer match between surface energy components of resin and
fiber. Numbers associated with the circles are percent A-lIDO and
wet shear strength.
64 E.J. BERGER AND Y. ECKSTEIN
6.0

5.0

4 .0
...
'(ij

3.0

2.0

1.0

O.O+--..,......--,--....,....-"'"'T----r----,r-
0.0 2 .0 4 .0 6 .0 8 .0 10.0 12.0

R' -Ro'

Figure 7. Single Filament Interfacial Shear Strength After

Water-Boil vs. the Griffith Critical Fracture Energy.

It should be noted that in both Figure (4) and Figure (7) the
0.2% A-174 point does not fit the trend of the rest of the data.
The reason for this is not known. We are confident, however, that
this is a real phenomenon.
While it must be emphasized that yd and yP are
approximations, this analysis works well for most of the ~stems
studi ed. It demonstrates that whil e both components play a role
in adhesion of coated E-glass fibers to epoxy resin, the polar
component is the more important in determining resistance to
debonding by water.
CONCLUSIONS
Interfacial shear strength in a glass fiber/epoxy resin
composite is seen to be a function of the concentration of silane
solution for y-aminopropyl(triethoxy)silane and
y-methacryloxypropyl(trimethoxy)silane used as fiber coatings.
Also, as wetting of the glass fiber surface by epoxy resin
increases, i nterfaci a 1 adhes i on in the compos ite increases,
probably due to increased contact between the resin and the fiber.
The ratio of wettability of the fibers by resin to wettability by
water is shown to be an indicator of wet strength. Wet strength
EPOXY RESIN WETTING AND SHEAR STRENGTH 65

was also analyzed as a function of closeness of polar and

dispersion components of the cured epoxy and the fibers, compared
to their closeness to the water coordinates. On a plot of the
square roots of the polar and the di spers i on components, the
distance from the liquid water coordinates to a circle describing
the closeness of each fiber to the resin is seen to correlate to
the wet strength. A smaller distance indicates a lower resistance
of the glass/epoxy bond to water.

ACKNOWLEDGEMENTS
D. William Stroud and Sarah K. Wesson are gratefully
acknowledged for help in data collection. Thanks are also due to
Sheldon P. Wesson for helpful discussions.

REFERENCES
1. M. Yamamoto, S. Yamada, Y. Sakatani, M. Yaguchi and
Y. Yamaguchi, Int. Conf. Carbo Fib., Compos. Appl. Pap.,
Plas. 21/1-21/6, 179, 1971.
2. W. D. Bascom, NRL Report 6140, Washington D.C., 1964.
3. B. Miller, in "Surface Characterization of Fibers and
Textiles,1I M. J. Schick, Editor, Vol. 7, Part 2, Marcel
Dekker, New York, 1977.
4. F. M. Fowkes, Ind & Eng Chern, 56 (12), 40 (1964).
5. W. A. Fraser, F. H. Achker, ana-A. T. DiBenedetto, Proc.
Conf. Reinforced Plastics/Composites, SPI, Section 22A, 1975.
6. L. T. Drzal, M. J. Rich, J. D. Camping, and W. J. Park, Proc.
Conf. Reinforced Plastics/Composites, SPI, Section 20C, 1980.
7. K. L. Loewenstein, in "Composite Materials," L. Halliday,
Editor, Elsevier, p. 150, 1966.
8. F. J. McGarry, Proc. Conf. Reinforced Plastics/Composites,
SPI, Section lIB, 1958.
9. D. H. Kaelble, J. Adhesion, 2,66 (1970).
10. D. H. Kaelble, "Physical Chemistry of Adhesion," pp. 153-170,
Wiley-Interscience, New York, 1971.
11. T. Smith, J. Adhesion, 11, 243 (1980).
 Part II
Factors Influencing Behavior of Adhesive Joints
INFLUENCE OF SURFACE ROUGHNESS ON MECHANICAL PROPERTIES

OF JOINTS

Y. Gilibert 0 and G. Verchery 00

o Ecole Nationale Superieure 00 Ecole Nationale Superieure

de Techniques Avancees des Mines
Centre de l'Yvette 158, cours Fauriel
91120 Palaiseau, France 42023 St Etienne, France

Mechanical properties of joints are dependent on the

geometry of surface defects caused during manufacture of
the adherends. This paper presents experimental results on
the influence of the roughness, due to grinding and sand
blasting, on tensile properties of joints. We studied spe-
cimens made of low carbon steel rods, bonded by an epoxy
resin containing mineral fillers. The surfaces were pre-
pared with various roughness conditions : 2 grindings
(coarse and fine), 8 grindings with sand blasting (combi-
nation of 2 grindings and 4 diameters of sand particles),
and one shot blasting. The roughness parameters were mea-
sured, particularly the total depth and the average spa-
cing of roughness. These measurements showed that the
prepared surfaces had well defined and reproducible rough-
ness. Electric strain gauges at specific points on the
metallic parts were used ; these provided measurements
of local strains (at ±l ~m/m) during the loading. The
variation of these strains represents indirectly the
joint behavior : the elastic limit, the appearance of
micro-cracks and their propagation up to failure. Optical
and scanning microscopic observations were used to check
the results obtained by these mechanical measurements.
The principal results of the present work are the fol-
lowing: with or without subsequent blasting, fine grin-
ding produces better mechanical properties than coarse
grinding ; sand blasting improves these properties
when compared to shot blasting or pure grinding ; and
finally, sand blasting gives the best behavior when the
total depth of roughness is equal to the mean diameter
of the dispersed particles in the resin.

69
70 Y. GILIBERT AND G. VERCHERY

INTRODUCTION

It is commonly observed that surface defects influence me-

chanical properties of joints : elastic limit, strength, and type
of fracture l - 4 • In this work, we have studied, in particular, the
effects of surface defects caused during machining of the adherends.

We realized from the beginning the necessity of precise con-

trols during all the steps so as to obtain reproducible and depend-
able results. We have paid special attention to the following
points :
- procedure for the pretreatment of the surfaces and the
measurements of the resulting geometry of the adherend surfaces
- analysis of the adhesive characteristics, all the proper-
ties of which are not provided by the manufacturer ;
- design and manufacture of the test-specimens ;
- loading conditions ;
- selection of the characteristic mechanical parameters, and
the procedure for measuring these parameters.

We used visual, optical and scanning microscopic observations

to determine the aspect of fracture and to study the distribution
of the fillers in the resin.

Most of the results of the present work can be found with more
details in Reference 5, relevant chapters of which will be referred
to in the following text.

PREPARATION OF THE SURFACE AND MEASUREMENT OF THE GEOMETRY

Properties of the Adherends

A low carbon steel ( 0.18 % carbon; XC 18 French standard

equivalent to SAE-AISI 1017) was used for the adherends of the spe-
cimens studied. The properties of the steel were controlled using
mechanical measurements and microscopic observations. The Young's
modulus (ET = 207 700 MPa) and the Poisson's ratio were determined
using tension test with strain gauges at a very low strain rate
(E about 10- 6 s-l) on steel specimens with 100 mm 2 square cross
section and 250 mm length. The Brinnell hardness (HB = 210) test
showed the uniform quality of the material, whereas microscopic
observations indicated a ferritic microstructure.

Preparation of the Rough Adherends, Finishing and Blasting

A shaping machine with a traversing tool was used in the pre-

paration of the rough adherends.
INFLUENCE OF SURFACE ROUGHNESS 71

This was followed by milling and after which a grinding was

effected with a horizontal surface grinding machine under different
conditions : coarse or fine grain grinding wheels. For all -these
operations, machining conditions (speed, feed, depth of cut, lubri-
cation, sharpening of the tools) were strictly controlled and are
described in Reference 5, Chapter 4.
The finishing procedure for some of the specimens was comple-
ted with sand blasting under four conditions using different par-
ticle diameters (lIS, 169, 282 and 423 ~m) or with shot blasting.
The particles with a lIS ~m diameter were made of high purity white
alumina ( AI 20 3>99.5%), whereas the particles with diameters of 423,
282 and 169 ~m were of brown corundum, i.e. less pure alumina (com-
position: AI203>94% ; Ti0 2<2.5% ; Si02<.5%). The blasting pressure
used was 0.4 to 0.5 MPa. The sand jet was inclined at 60° from the
specimen surface and displaced at a speed of 24 mm s-! The shot
blasting was conducted with iron shots I mm in diameter.

In this paper, the various surface states are designated as

follows :
RF fine ground state ; RFS - fine ground sand blasted state ;
RG - coarse ground state ; RGS - coarse ground sand blasted state
RGG - coarse ground shot blasted state.
The numbers following RFS and RGS designate the particle mesh size.

Measurement of the Surface Roughness Parameters

The surface profiles were determined using probes with an air

bearing of 25 ~m and 750 ~m travel distances. The specimens used
had the same section as the adherends and were machined and treated
at the same time under the same conditions. The surface profiles
were measured, before and after blasting, along reference paths,
the ends of which are marked with two notches remaining after
blasting.

Data were recorded and processed by a computer to determine

the various parameters of the surface defects describing the total
profile, roughness and waviness. The calculations were done using
a method developed by B. Scheffer of Renault, and J. Bielle and
C. Berger of ENSAM 6 ,7.

We observed that the surface roughness parameters alter very

quickly during the first few seconds of blasting and do not change
after a while. We selected a time lapse of 30 seconds which ensured
a stable state of these parameters.

Measurements showed that the surface roughness parameters are

not dependent on the direction of measurement for blasted states,
whereas condition as ground showed some anisotropy.
72 Y. Gill BERT AND G. VERCHERY

Table I presents values of the surface roughness parameters

for the various surface states studied. These are the total depth
of roughness, Rt ; the average depth of roughness, R ; the maximum
depth of roughness, Rmax ; the roughness levelling depth, Bp ;
the arithmetical mean deviation from mean line of roughness, R ;
and average spacing of roughness, ~ • In Figure I, these para~eters
are plotted against the mean diameter d of sand particles.

Table I. Roughness Parameters of the Various Surface States. AVB

is a trade name (Aluminium Oxide Vapor Blast) for an alumina with
a purity over 99.5%.

Designation dlL m ROUGHNESS (11 m )

-- - Rt R Rmox AR Rp Ra
RF -- 2.73 0.82 2.73 40 0.44 0.29
RG -- 9.09 4.39 9.09 90 236 1.39
RGG -- 2,(09 18.68 24.68 5650 9.00 5.20
RGS60C 423 21.61 12.83 21.61 285 6.77 3.47
RGS 90C 282 7.20 4.96 7.20 185 2.77 1.40
RGS 150C 169 6.21 3.92 6.21 135 2.12 1.20
RGS220AVB 115 9.01 4.73 9.01 155 2,51 1.41
RFS60C 423 18.09 11.51 18.09 195 5.95 3.14
RFS 90C 282 14.42 8.06 14.42 100 4.34 2.54
RFS150C 169 5.60 3.47 5.60 110 1.88 1.01
RFS 220AVB 115 3.64 2.27 3.64 105 1.25 0.69

From these data, it can be seen that the surface roughness

parameters for sand blasted states depend on both the initial grind-
ing and the mean diameter of the particles used for blasting. For
all surface preparations, the maximum depth of roughness Rmax is
very close to the total depth Rt , which indicates that surfaces have
no aberrant defects of roughness and are consequently fairly homoge-
neous ; for the fine ground state, the roughness parameters increase
steadily with the particle diameter whereas for the coarse ground
state these increase notably only for a diameter superior to 200 ~m.

Other surface defect parameters were computed but proved of

little importance as far as the mechanical properties of the joints
are concerned. However, they confirmed that the prepared surfaces
had well defined and reproducible parameters.
INFLUENCE OF SURFACE ROUGHNESS 73

Roughness J•
J
11 m
J
J
I

A~o.
R F5
- --------- R G 5

• J
J
I
I

I
I ,,
I
J I
I I

10 - __ J
I
I
--"',, I
I

,, I R
/

\.
'\ - --- I '
" ",,,,-+

+ R,/ ~~
5 ________ ~ _------::-----~
" .~-~///' -1
--- - -- ---- --- - ~~------:~
----- ;~
I==c----~
-
100 115
R
-- -----
169 200
--- R
-"'----'-'-......

26 2
d
~23
m

Figure I. Roughness parameters versus sand particle diameter. The

solid line refers to fine ground blasted states (RFS) ; the broken
line refers to coarse ground blasted states (RGS).

ANALYSIS OF THE ADHESIVE

We used the commercial adhesive, Eponal 317, of Sonal, France,

which is a two component system (an epoxy resin containing mineral
fillers and a hardening agent). It was processed at room tempera-
ture ; at such temperature, about 20 minutes are available for use
and polymerization rate reaches 90% within 2 hours.

We determined its mechanical, physical, structural and chemi-

cal properties. The elastic constants were measured at room tempe-
rature (20°C) using two methods. Tensile test using strain gauges
and a very slow loading rate (e
about 10- 6 s-l) showed an elastic
brittle behavior. Using traction test the Young's modulus was found
to be 5800 MPa, the Poisson's ratio was 0.327 and the strength was
30 MPa ; whereas ultrasonic measurements gave 5940 MPa for the
Young's modulus and 0.323 for the Poisson's ratio. The measured
density was 1.433 kg dm- 3 •
74 Y. GILIBERT AND G. VERCHERY

Figure 2. Fracture surface through the adhesive (scale bar 0.5 mm).

Figure 3. Fracture surface through the adhesive (scale bar 50 llm).

INFLUENCE OF SURFACE ROUGHNESS 75

The resin was calcined at 1000°C, which produced a mineral

residue weighing 40 % of its initial weight. A qualitative analysis
of this residue was performed using two types of methods : chemi-
cal reactions and use of aX-ray spectrometer associated with the
scanning electron microscope. It showed the presence of Mg, Fe
and Si in significant quantities, barium sulphate, Ti, and traces
of AI, Ca and various fluorides. The hardening agent was also calci-
ned at 1000°C. The chemical reactions on the residue showed the
presence of Ca, small quantities of Fe, carbonates and traces of AI.

Observations of the surfaces of the fractures in the adhesive

using the scanning microscope revealed that the fillers are small
particles with a prolate shape, a mean diameter of 5 vm, and a
length up to 10 or IS vm. Scanning electron micrographs (Figures
2, 3 and 9) showed that the fracture in the adhesive went through
the mineral particles, with a smooth-looking surface, revealing a
brittle behavior of the fillers. Cracks perpendiculars to the adhe-
rends were observed in the adhesive, and a brittle behavior of the
adhesive is revealed. More details can be found in Reference 5,
Chapter 3.

DESIGN AND MANUFACTURE OF THE TEST-SPECIMENS

LOADING CONDITIONS

We used specimens with two double-lap ]01nts. One of these

double-lap joints was used for strain measurements. The other
joint, which was tightened transversely by a binding clip, ensured
axial and uniform load in the adherends, while the clip eliminated
any local bending. The extremities of the test-specimen were fixed
on a universal testing machine by knee joints. In fact, this design
of the specimen ensured symmetrical fractures in the tested part,
and the clip decreased considerably the statistical deviation in
the measurements, as shown in Reference 5, Chapter 2. Figure 4 shows
a schematic view of the specimen.

The parts (10 mm x 4.5 mm cross-section for parts 2' ; 10 mm

cross-section for parts I ') were produced at a ± 0.01 mm tolerance.
The length of the overlap in the tested part of the specimen was
88 mm. A thickness of 0.5 mm was selected for the adhesive joint
and controlled by gauged spacers. Preliminary tests, described in
Reference 5, Chapter II, showed that this thickness gives the best
mechanical behavior for these dimensions of the metallic parts and
this length of the joint.

The adhesive was applied at room temperature by hand with a

spatula. A very thin layer was first spread on the adherends.
Bubbles formed at the interface during this application vanished
within a minute, after which a I mm adhesive layer was spread.
76 Y. GILIBERT AND G. VERCHERY

The various parts were positio~ed together into a frame. The excess
adhesive was removed after the polymerization.

The test-specimens were loaded in tension at a slow rate

(1000 N per minute) at room temperature (20°C).

t
positions or the strain gouges
2 3 4 5 6
3'
F
- . ~ . -.

3'

z
a

Figure 4. Schematic view of the test-specimen (dimensions in mm).

A view of the cross-section is superimposed between gauges 1 and 2.
l ' and 2' : adherends ; 3' : binding clip.

PRINCIPLE OF THE MECHANICAL MEASUREMENTS

Since mechanical quantities inside the joint cannot be obtained

directly, we used electric strain-gauges placed at different points
on external surfaces of the metallic adherends. Figure 4 shows the
position of the gauges ; their distances from the middle of the
joint are as follows
- gauge 1 -37 mm
- gauge 2 -18.5 mm
- gauge 3 o
- gauge 4 +18.5 mm
- gauge 5 +37 mm
The positioning of the gauges is described in detail in References
5 (Chapter 5) and 8.
 Z
TI
r
C
m
Z
(")
m
oTI
(J)
C
:xl
TI
}>
(")
;/ . m
.~ ®-K~ 0-7"'_." :xl
-", o
..-.L:=:J /+.' .' C
.•... . G)
;G 2
~
I
Z
,of. m
.• (J)
,t ' (J)

If_
...~ ... D:3
"• .• " ,t ·
D ,... :t.. . ....... .
... ~ .... +...
10 \ t"+ ' , -• • •~ - ..... ' ••

1 :t.... " .....

. ~.., . . '
~ .'• . + "."
5 " Ii .,.- 4-
+ ,. 0,+' ...
.
.t '. '.
i.. ~~f 't -6
!f~-' (;10
o 200 400 600 800
Figure 5. Strain versus tensile load for a ground sand blasted state
(outputs of gauges 1-4). Non-linearity occurs over point D2.

-...I
-...I
78 Y. GILIBERT AND G. VERCHERY

A sensitive and precise ( ±0.25%) bridge (VE 20 Micromeasure-

ment) was used.

Figure 5 shows typical strain gauge recordings for a ground

sand blasted state. Other recordings are given in Reference 5.

The gauges numbered 2, 3 and 4 show a linear relationship up

to points D2' D3, D4 respectively. The fact that gauge 2 is the
fir'st to deviate from linearity shows that initiation of micro-
cracks takes place in the joint in area 2 under this gauge (at
abscissa -18.5nnn). When the load is increased, over the point D2,
micro-cracks continue to develop in area 2 and begin to appear in
area 3 (point D3), then 4 (point D4)' At G2, micro-cracks in area
2 reach a sufficient size to propagate and later (point G3), propa-
gation begins in area 3. For a given load, this propagation is
stable, but the speed of stable propagation increases when load
increases. During this phase, area 2 can support less and less load.
The absolute value of the slope of the load versus the strain of
gauge 2 decreases. When it reaches zero, the propagation becomes
unstable and fracture occurs very quickly.

Gauge 5 is in fact of little use as far as the analysis of

the fracture process is concerned. Gauge 1 delivers a weaker output
than the other gauges ; its values are not easy to analyze, espe-
cially at low loads. It can be seen that a compressive effect
occurs about point D2 and developed up to point G2' Generally, non-
zero values could be analyzed as the sign that the extreme end of
the joint is not fractured until the final step.

Visual and microscopic observations of the fractured specimens

have confirmed that in the central part of the joint a progressive
shear fracture occurs inside the adhesive ; at the ends of the joint
one can see either a shear fracture inside the adhesive or an adhe-
sive failure between the adhesive and the metallic adherend along
the upper or lower face, induced during the final phase of the
fracture.

These facts differentiate the ground sand blasted state from

the ground state, which showed predominantly metal-resin adhesive
fractures.

The various stages for the ground sand blasted states, (i)
elastic behavior, (ii) initiation of micro-cracks, (iii) coalescence
of the micro-cracks, (iv) stable propagation of the cracks, and
finally, (v) un~table propagation up to final fracture, proved to be
absolutely reproducible (Reference 5, Chapter 8).

These experimental results can be partly explained by the va-

rious theoretical analyses 9- 10 , which have analyzed stress fields
INFLUENCE OF SURFACE ROUGHNESS 79

in adhesive joints. However, these theoretical analyses, which are

based on continuum mechanics and ,use idealization of interfaces,
cannot explain the significant differences observed using the above
measurement procedure, and are presented below.

RESULTS

Here are presented the effects of roughness on the three typi-

cal mechanical parameters : elastic limit (crack initiation thres-
hold), crack propagation threshold, and ultimate strength.

We found that roughness can be adequately represented by using

the total depth of roughness Rt.

We observed a very small scattering in the measured values,

which gives proof of the importance of our careful and well-defined
process in preparation and testing. Deviation in the results is
generally higher for the propagation threshold and the ultimate
strength than for the elastic limit. Further, deviation in the ulti-
mate strength is higher for non-blasted surfaces than for blasted
surfaces (Reference 5, Chapter 8). Typical values, corresponding to
the results presented in Figures 6 and 7, are, (i) 0.25 kN for the
crack initiation threshold (about 12 kN), (ii) 0.5 kN for the crack
propagation threshold (about 24 kN), and (iii) 0.5 kN for the ulti-
mate strength (about 30 kN).

Mechanical Results

For the sand blasted states, we found that the mechanical para-
meters were influenced by the initial grinding and the roughness,
which is related to the sand particle size.

The fine grinding is equivalent or better than the coarse grin-

ding for the three mechanical parameters ; it gave better values
for the elastic limit with the large particle diameters, and better
values for the propagation threshold with all diameters.

For both types of grinding, we observed that the strain mea-

surements given by the gauges at the external surface of the adhe-
rends were significantly smaller (for any position) and more uni-
form along the specimen, when the surfaces were prepared using
particles with a diameter of 169 pm. Furthermore, this preparation
shows the best values for the three mechanical parameters.

Figures 6 and 7 summarize the results. Almost all of the

plotted points represent the mean value of four tests. In Refe-
rence 5, Chapter 9, one can find the curves representing F versus
80 Y. GILIBERT AND G. VERCHERY

the other roughness parameters, R9 , Rp ' R ; they have the same

behavior as the curves in Figure b.

Figure 6 shows that for fine grinding, with or without blas-

ting, the three mechanical parameters reach a maximum value when
the total depth of roughness equals 5.6 }.lm. It seems possible to
associate this property with the fact that the total depth of
roughness is approximately the size of the fillers in the resin.

No precise trends can be found in the case of coarse grinding.

F
kN
~-
30 ~/ ~ 0
t:Y' ' '0------------ £ - - ---0
25

20

15
/'
~/ 0,
10
I f ' '0- - - - - - - - +
- - - - - - - - - -0--
~RF
-t-RFS
0RGS
oRGG
o 2.73 5.60 9.01 14.42 18.09 21.61 2{0
3.64 6.21 7.2

Figure 6. Mechanical parameters versus the total depth of rough-

ness : from the bottom, (i) crack initiation threshold, (ii) crack
propagation threshold, and (iii) ultimate strength. For fine grin-
ding without blasting (RF), the values are in accordance with the
extrapolation (with res'pect to the roughness) of those obtained
for fine grinding with blasting (RFS).

Visual and Microscopic Observations of Fractures

Visual observations of the fractures show that for all the

sand blasted states mostly shear fractures occur in the adhesive
the shot blasted states show no such shear fracture.
INFLUENCE OF SURFACE ROUGHNESS 81

F 6. RF
kN + RFS
o RGS
30

25

20

15

~
• •• +
····· ·0 ~
10 "" "' .. ,-0

o 61.1 115 169 282 423

Figure 7. Mechanical parameters versus the mean diameter of the

sand particles : from the bottom, (i) crack initiation threshold,
(ii) crack propagation threshold, and (iii) ultimate strength.
For fine grinding without blasting (RF), a fictitious diameter is
considered, so as to retain the property shown in Figure 6, i.e.
linear extrapolation from the nearest values for fine grinding with
blasting (RFS).

We measured the length, L, of the shear fractures as they

occurred in various states. Figure 8 is a plot of the mechanical
parameters versus this length ; the corresponding curves are
approximately linear.

Microscopic observations with use of aX-ray spectrometer

were performed. For sand blasted states, they confirmed the predo-
minance of shear fractures inside the joint. Furthermore, we ob-
served that the mineral fillers tended to fill the cavities on the
surface and, when prolate, to be inclined at 45° to the surface.
Figure 9 shows a typical view. For the shot blasted states, the
fracture surface revealed the metallic adherends with flakes of
adhesive about 0.1 nm in diameter. (Figure 10).
82 Y. GILIBERT AND G. VERCHERY

F
kN
o ___--~r-----0t----0
I__-----------t-e
30 ~ 0

0RGS
tRFS

20

10
9

Lmm
o 20 21; 27 50 60

Figure 8. Mechanical parameters versus the length of shear frac-

ture : from the bottom, (i) crack initiation threshold (two lines),
(ii) crack propagation threshold (two lines), and (iii) ultimate
strength (one line).

Figure 9. Fracture aspect for a sand blasted state ; the arrow is

the direction of the surface of the adherend (scale bar 20 ~m).
INFLUENCE OF SURFACE ROUGHNESS 83

Figure 10. Fracture aspect for a shot blasted state the scale
bar represents 0.2 mm.

CONCLUSION

Our experimental results show that the surface roughness has

a significant influence on the mechanical properties (crack initia-
tion threshold, crack propagation threshold, and ultimate strength)
of joints. We have found the best conditions for joint design as
follows
fine grinding is better than coarse grinding ;
sand blasting improves mechanical properties ;
sand particle diameter must be selected so as to give
a total depth of roughness close to the size of the
adhesive fillers.
84 Y. GILIBERT AND G. VERCHERY

REFERENCES

1. J.J. Bikerman, "The Science of Adhesive Joints," Academic Press,

New York, 1968.
2. R. Houwink, "Adhesion and Adhesives, Vol. 2 : Applications,"
Elsevier, Amsterdam, 1967.
3. J. Corey McMillan, in "Bonded Joints and Preparation for
Bonding," AGARD Lecture Series, No.102, Part 7, 1, 1979.
4. D. Hennemann and W. Brockmann, J. Adhesion, 12, 297 (1981).
5. Y. Gilibert, "Contribution a l'etude de l'adhesivite de
materiaux colles par l'intermediaire de resines epoxydiques,"
Doctoral Dissertation, Universite de Reims, France, 1978.
6. J. Bielle, Revue Arts et Metiers (Paris), No.3, 44 (1973).
7. J. Bielle, "Etat geometrique des surfaces techniques," lecture
notes, EN SAM , Chalons sur Marne, France, 1981.
8. Y. Gilibert, H. Miller and G. Gautron, Annales de l'Universite
et de l'ARERS, Reims, France, 14, 17 (1976).
9. M. Goland and E. Reissner, J. Appl. Mech., 11, A17 (1944).
10. D. Volkersen, Construction Metallique (PariS), No.4, 3 (1965).
11. R.D. Adams and N.A. Peppiatt, J. Strain Analysis, 8, No.2,
134 (1973). -
12. Y. Gilibert and A. Rigolot, J. Mecanique Appliquee, 1, No.3,
341 (1979).
THE EFFECTS OF SURFACE CONDITIONING OF AISI-304 STAINLESS STEEL ON

THE INTERFACIAL PROPERTIES OF ALLOY/EPOXY COMPOSITE STRUCTURE;

ADHEREND SURFACE CHARACTERIZATION USING X-RAY EMISSION SPECTRO-

SCOPY (LEEIXS)

F. Gaillard, A. A. Roche and M. J. Romand

Applied Chemistry Department (CNRS, ERA # 300)

University Claude Bernard - LYON I
69622 Villeurbanne Cedex, France

AISI-304 stainless steel adherend-epoxy bonded

specimens were subjected to a three point flexure test.
This test used single interface type specimens. The
measurements were conducted in order to: (i) investigate
the effects of adherend surface reproducibility and metal
surface conditioning on joint strength, (ii) determine
the sensitivity and repeatability of the test, (iii)
gain a better idea of the range of applicability of the
test.

The results presented emphasize that mechanical

properties such as modulus of ri.gidity and ultimate load
values of the metal/epoxy interface are highly dependent
upon adherend surface treatments (chemical etching,
electropolishing, cathodic reduction, thermal treatment).
In order to understand the chemical factors affecting
adhesive strength, a near-surface sensitive technique
(Low Energy Electron-Induced X-ray Spectroscopy (LEEIXS»
has been used and effects of surface conditioning of the
stainless steel on the surface layer composition and
thickness were studied.

85
86 F. GAILLARD ET AL.

INTRODUCTION

Different standard mechanical tests are commonly used to eva-

luate adhesive bonded joint strength or quality. These tests may
be categorized in three major groups i.e. tension, shear or peel,
even though various geometries have been adapted from these confi-
gurations.
In order to investigate the influence of the adherend prepa-
ration processes on the strength and the mechanical properties of
the adhesive bonded joint interfacial region, it has been shown
recentlyl,2 that a three point flexure test applied to a single
interface type specimen can be employed advantageously. The objec-
tive of the present work is to gain a better idea of the range of
applicability of the test and therefore to define ways and means
to achieve fully reliable bonded joints. The relevant study deals
with AISI-304 stainless steel adherend/epoxy bonded specimens. In
addition LEEIXS (Low Energy Electron-Induced X-ray Spectroscopy)3,
a near-surface sensitive technique, was used in order to understand
the chemical factors which affect the adhesive strength.

I. INSTRUMENTATION

1.1 Three point flexure test

A table model 1102 INSTRON tester fitted with a cross head

speed reducer and a 500 N load cell was used for testing. The cross
head speed was typically 0.5 mm.mn- 1 with a loading sensitivity
down to 100 N full scale.
The specimen preparation fixture has been described previously
1,2,4. A Plexiglass holder was used as a form for the RTV 501 sili-
cone forming mold. Rectangular Plexiglass strips (25 x 5 x 10 mm)
were prepared simultaneously by a machining process to an exactly
identical size. In this manner, the area of all mold cavities was
identical. By applying an exact amount of adhesive (0.5 cm 3 ) to
each adherend with a syringe, bonded adhesive size and volume were
then precisely controlled.
Industrial AISI-304 stainless steel foils 0.47 mm thick were
milled simultaneously to an identical size (10 x 50 mm).

1.2 LEEIXS instrument

A detailed description of the LEEIXS instrument has been given

elsewhere 3 ,5. We shall only mention here the basic principles of
such an apparatus, which is shown schematically in figure 1. This
equipment consists of a wavelength-dispersive X-ray spectrometer
equipped with a cold cathode tube (CCT). This gas discharge tube is
an open window device which operates directly in the spectrometer
primary vacuum. Normal operating reduced pressure is 1-10 Pa. The
CCT is used to bombard the sample surface with a quasi monoenerge-
tic electron beam which may be selected in the range of 0.5-5 keV.
EFFECTS OF SURFACE CONDITIONING OF STAINLESS STEEL 87

Primary Vacuum Enclosure

Figure 1. Schematic drawing of the Low-Energy Electron Induced

X-ray Spectrometer (LEEIXS).

The beam spot size at the sample surface is limited by the anode
aperture. Typically, the spot diameter is 1 cm. During experimen-
tation, for a given acceleration voltage, the selected electron
current is stabilized by an automatic pressure regulator in the
range of 0 to 2 mAo Both voltage and current are digitally control-
led. The soft X-ray spectrometer employed in the present measure-
ments uses flat analyzing crystals or pseudo-crystals. The detector
is a flow proportional counter with an ultra-thin (0.5 wm alumini-
zed polypropylene) window and uses, at atmospheric pressure, an
argon-methane (P-IO) gas flow. The bombarding electron beam inci-
dence angle is 70° and the take-off angle of X-rays is 35°.

The effective depth of analysis of LEEIXS is determined by the

material thickness at which characteristic X-rays are produced
within the sample. Obviously, this depends upon the target mate-
rial, the electron energy and the electron beam incidence angle.

From our previous results and their comparison with Rutherford

Backscattering Spectroscopy data 6 ,7, it appears that the average
excitation depth (Re) at which X-rays are produced can be formula-
ted as follow
88 F. GAILLARD ET AL.

Table 1. Excitation Depth for OKa , CrLa and FeLa Radiations.

E(keV) 1.5 4.5

Radiation OK CrL FeLa OK CrL FeLa

a a a a

Re (nm)
A1S1-304 - 13 11.5 - 136 134.5
matrix

Re (nm)
Fe2°:) 21 21 19 203 202 200
matr1x

where n = 1.2/(1 - 0.29 log Z), Eo is the electron energy in keV,

~x !he excitation energy in keV of the concerned emission process.
A, Z and p are respectively the atomic weight, the atomic number
and the density of the material under investigation. This expres-
sion for the depth of excitation has been deduced from the Feldman
relationship8 giving the practical maximum range of 1-10 keV elec-
trons in solids. This semi-empirical relationship takes into ac-
count the electron backscattering effects and the energy loss of
electrons during their path in the target material.

The calculated excitation depths for OK , CrLa and FeL radia-

tions are given in table I for 1.5 and 4.5 k~ electron ene~gies.

1.3 Cathodic reduction device

The oxide layers formed at the sample surfaces can be reduced

using a standard amperostatic circuit 9 shown in figure 2.

Reduction takes place in a medium of concentrated H2S0 4 and

the reduction current density can be fixed in the range from 0.6
to 13 ~.cm-2. Under reduction, the sample potential varies from
oxide to metal/medium potential. The amount of charge necessary
for this potential variation can be related to the cou10mbic
thickness of the reducible oxide 1ayer 9 ,10,11.
EFFECTS OF SURFACE CONDITIONING OF STAINLESS STEEL 89

A
t-It-I

H
B

Figure 2. Schematic amperostatic circuit diagram : A, 75 V battery;

B, ammeter ; C, current integrator ; D, sample E, reference
electrode; F, auxiliary electrode (Pt) ; G, E = f(t) recorder;
H,2MDresistance.

II. STATISTICAL STUDY

Different adherend surface conditioning treatments were used

as depicted in Table II.

Firstly, it is necessary to clarify that the different chemi-

cal conditionings do not affect the modulus of rigidity of the al-
loy (EI)M and therefore the substrate mechanical properties. For
this purpose, unbonded metallic strips were submitted to different
cleanings or etchings (degreasing, electropolishing, concentrated
and diluted sulfuric acid oxidations) and then subjected to the
three point flexure test. The modulus of rigidity of the metallic
strip (EI)M is deduced from the slope p/o of the load/displacement
curve in the region of elastic deformation by :

where l is the distance between the two supports. The values of

(EI)M, shown in table III, exhibit a relative standard deviation
value of ± 1.6 %. Such a statistical result takes into account
reproducibilities of the mechanical tester, specimen sizing and
adherend surface state.
90 F. GAILLARD ET AL.

Table II. Different Treatments for AISI-304.

Code Treatment

A Sample degreased in acetone for ten minutes.

B Sample degreased as in A, submerged in 1 % HC!/" 2 % HF,
10 % RN03 at BO°C for one minute, rinsed with deionized
water for two minutes.
C Sample degreased as in A, submerged in 5 % H2S04 at 60°C
for 15 minutes, rinsed as in B.
D Sample degreased as in A, electropolished in a CH3C02H,
HC!/,04 3.2 : 1 solution, rinsed as in B.
E Sample cathodically reduced in concentrated H2S04 till
dissolution of any oxide layer.
F Sample degreased as in A, oxidized at 300°C at air for
one hour, air cooled.
G Sample degreased as in A, submerged in boiling concen-
trated H2S04 for 15 minutes, rinsed as in B.

Secondly, it is necessary to emphasize the test reproducibi-

lity for the composite materials prepared. After bonding, sets of
identically prepared specimens (identical surface preparation,
epoxy formulation, curing cycle ••• ) were tested. The results are
tabulated in table IV.

For panels oxidized by means of treatment C, relative stan- 4

dard deviation a % of the modulus of rigidity of the system (EI)S
was less than 1 %, while this deviation is less than 7 % for the
values of the ultimate load (F max ). Such a statistical result
includes reproducibilities of the tester, specimen sizing, adhe-
rend surface state and bonding processing.

Table III. The Effects of Surface Conditioning on the Modulus of

Rigidity of the Metal.

Treatment A B C D

(EI)M (10 3 N • lllIri 2 ) 16.5 17.1 17.0 17.0

(EI)M ± a % = 16.9 ± 1.6 %

EFFECTS OF SURFACE CONDITIONING OF STAINLESS STEEL 91

Table IV. The Effects of Surface Conditioning on the Mechanical

Properties of the System.

Treatment A C D G

Sample (EI)S Fmax (EI)S Fmax (EI)S Fmax (EI)S Fmax

number (N.mm 2 ) (N) 2 2
(N.mm ) (N) (N.mm ) (N) (N .mm 2 ) (N)

I 122.10 3 31 172.10 3 33 104.10 3 19 128.10 3 30

2 86.10 3 12 174.10 3 38 97.10 3 18 128.10 3 31
3 121.10 3 9 174.10 3 33 112.10 3 18 124.10 3 33
4 85.10 3 9 174.10 3 34 104.10 3 20 150.10 3 30
5 172.10 3 32 145.10 3 35
6 174.10 3 37 133.10 3 29

Hean 103.10 3 15 173.10 3 35 104.10 3 19 135.10 3 32

values
± (J % ± 20 % ± 67 % ± 0.9 % ± 7 % ± 6 % ± 4 % ± 8 % ±8%

(EI)S ± (J % = 129 ± 26 %
F ± (J %= 25 ± 38 %
max

Table IV also shows the (EI)S and Fmax values for different
series of specimens subjected to various conditioning treatments.
(EI)S and Fmax are respectively the average of the various (EI)S·
and Fmax values obtaine~for each surface conditioning._Relative
standard deviation of (EI)S was found to be 26 % while Fmax rela-
tive standard deviation was 38 %. Such variations cannot be explai-
ned by statistical error (i.e. reproducibilities of apparatus,
specimen sizing, adherend surface state) and are therefore charac-
teristic of the properties of the created interphase.

III. INFLUENCE OF ADHEREND CONDITIONING

A set of eight samples was identically electropolished
(treatment D) ; four of them were then submitted to cathodic reduc-
tion (treatment D + E). The whole set was immediately bonded and
submitted to the three point flexure test. For panels bonded after
a cathodic reduction, there is a considerable decrease in the va-
lues of Fmax whereas the modulus of rigidity of the system remained
constant (see table V). The cathodic reduction curves are shown in
figure 3. These curves were obtained by reducing the surface oxide
layer at constant current density (j = 5 ~A.cm-2) in a concentra-
ted sulfuric acid (18 N) bath 12 . The amount of charge Q (mC.cm- 2 )
used for this reduction has been plotted against the sample poten-
tial relative to the standard calomel electrode (S.C.E.). The Q
92 F. GAILLARD ET AL.

Table V. The Effects of Cathodic Reduction on the Mechanical

Properties of the System.

Treatment D D+ E

Sample (EI) S F (EI) S F

max max
Number (l03N.mm 2 ) (N) (l03N.mm 2 ) (N)

1 104 19 127 7
2 97 18 111 7
3 112 18 106 8
4 104 20 103 11

Bean 104 19 111 8

values
± cr % ± 6 % ± 4 % ± 10 % ± 25 %

value necessary to reach the metal potential is then re~resentative

of the apparent coulombic thickness of the oxide layer 1 •

Figure 3 shows the amount of charge Q necessary for the reduc-

tion of (a) the AISI-304 natural oxide and (b) the AISI-304 surface
layer formed during cathodic reduction and / or immediately after.

E (mV/SCE)

+ 200

Q (mC.ciih

- 200

Figure 3. Cathodic reduction curves.

EFFECTS OF SURFACE CONDITIONING OF STAINLESS STEEL 93

In this last case the metallic surface is covered by a very

thin layer. The subtle metal potential difference exhibited by
curves (a) and (b) may be due to differences in surface state
and/or contaminant deposition on the electrodes. Inversely, treat-
ment by sulfuric acide (0) provided a thick oxide layer (table VI)
and gave the highest Fmax values.

It is alsQ worthy of note that (EI)S and Fmax values and their
corresponding a % are strongly dependent upon the preconditioning
steps. As can be seen in table VII, treatment C provided Fmax and
(EI)S values higher than those obtained after treatment B. If
treatment C was performed after treatment B (treatment B + C), cor-
responding values were lower than those obtained for treatment C
alone.

Table VI. The Effects of Surface Conditioning on Oxide Thickness.

OK INTENSITY OXIDE LAYER

a(a.u.)* THICKNESS (nm)**

Fe single crystal after 13

5690
long time air exposure
Etched AISI-304 (B) 2210 5
Electropolished AISI-304 3010 7
(D)
Etched AISI-304
followed by sulfuric 9090 21
acid treatment (B + C)
Sulfuric acid treated 10140 23
AISI-304 (C)
Thermally oxidized 10440 24
AISI-304 (350°C, 1 hour)
Thermally oxidized 21610 49
AISI-304 (500°C, 1 hour)

* LEEIXS measurements performed with 4.5 keV electron beam and

a 0.5 mA.cm- 2 electron current density.

** Thickness calculated using Feldman's relationship, assuming

that the surface oxide layer formed after each treatment is an
homogeneous Fe 30 4 film (density = 5.18).
94 F. GAILLARD ET AL.

Table VII. The Effects of Treatments B, C and B + C on the

Mechanical Properties of the System.

Treatment B C B + C

Sample (EI) S F (EI) S F (EI)S F

max max max
number (l03N.mm 2 ) (N) (l03N.mm 2 ) (N) (l03N.mm 2 ) (N)
I 151 17 172 33 117 32
2 130 19 174 38 119 23
3 152 16 174 33 126 26
4 160 16 174 34 126 30
5 172 32 104 26
6 174 37 116 29

Hean 148 17 173 35 118 28

values
± (J % ± 9 % ± 9 % ± I % ± 7 % ± 7 % ± 9 %

The standard deviations in (EI) Sand Fmax for (B + C) or (B)

treated metallic strips were similar. This emphasizes that the ma-
terial surface keeps remanent traces of each treatment step.

Figure 4 shows the LEEIXS spectra of (B), (C) and (B + C)

treated specimens. These spectra were obtained by bombarding the
sample surface with 1.5 keV electrons, the electron current densi-
ty being 0.5 mA.cm- 2 • Under these conditions, the depth probed is
less than 25 nm. X-ray dispersion is provided by a T£AP single
crystal analyser. These spectra show that surface enrichment of
nickel is produced by the 5 % H2S0 4 oxidation process (C). This
enrichment is revealed by the lower ratio of intensities FeLa/NiLa
exhibited by spectra C and B + C compared to this observed in the
case of spectrum B.

It is also noteworthy that the fluorine impurity at the B

etched sample surface, remained within the surface oxide layer
even after the sulfuric oxidation (B + C).

Both the mechanical and the spectroscopic studies emphasize

the effect of fluorine contaminant incorporated within the oxide
layer. This impurity is probably responsible for the poor adhesion
(lower Fmax and (EI)S values), as has been previously reported for
titanium and titanium all oys5,14,15.
EFFECTS OF SURFACE CONDITIONING OF STAINLESS STEEL 95

Felo:

OK",

8+C

800 700 E ( eV ) 600 550

!
i
!
i
!
, !

, q 16 18 20
A ( A) 22 24

Figure 4. LEEIXS spectra of AISI-304 after treatments (B), (C),

(B + C).
96 F. GAILLARD ET AL.

IV. INFLUENCE OF EXPOSURE TIME TO AIR BEFORE BONDING

In order to investigate surface ageing effects on bonded spe-

cimen performances, the treated panels were submitted to room
atmosphere exposure before bonding. After the desired ageing, aged
and identically treated unaged samples were bonded at the same time
and tested. The results are shown in table VIII.

In order to understand further these results it is necessary

to clarifY what are the physical meanings of (EI)S and Fmax para-
meters. Account been taken of the sample preparation processing,
an increase (or decrease) in the value of (EI)S~st~m is associated
with an increase (or decrease) in the value of \EI)lnterphase. For
the materials investigated in this study (EI)Oxid~ 1S h1gher than
the corresponding (EI)Metal. Consequently, format10n of an oxide
of high rigidity induces an increase in (EI)Interphase value and
therefore an increase in (EI)System 4 • Fmax values depend firstly
upon the viscoelastic properties of the so-created interphase
(i.e. upon the presence or not of a weak boundary layer) and secon-
dly upon diffusion of hydroxide groupsl6.

In the case of C-treated specimens, (EI)S and Fmax values

decrease and their respective cr % increase drastically. These

Table VIII. Adherend Ageing Effects on the Mechanical Properties

of the System.

Treatment C G

Ageing no one week no 24 hours

Sample (EI)S Fmax (EI) S Fmax (EI) S Fmax (EI) S F

max
number (N.mm 2 ) (N) (N.mm 2 ) (N) (N.mm 2 ) (N) I (N;mm 2 ) (N)
1 173.10 3 33 185.10 3 24 128.10 3 30 100.10 3 45
2 174.10 3 38 144.10 3 39 128.10 3 31 131.10 3 44
3 174.10 3 33 151.10 3 25 124.10 3 33 150.10 3 26
4 174.10 3 34 137.10 3 14 150.10 3 30 200.10 3 44
5 172.10 3 32 124.10 3 19 145.10 3 35 165.10 3 40
6 174.10 3 37 133.10 3 29 145.10 3 36
7 157.10 3 34

Mean 173.10 3 35 148.10 3 24 135.10 3 32 150.10 3 39

values
± cr % ± 1 % ± 7 % ± 16 % ± 40 % ± 8 % ± 8 % ± 20 % ± 18 %
EFFECTS OF SURFACE CONDITIONING OF STAINLESS STEEL 97

results are associated with changes in the properties of the oxide

layer which is well known to decay when exposed to air and to lose
its protective properties lO •

In the case of G-treated specimens (EI)S and Fmax values

increased with the surface ageing. This is in agreement with elec-
trochemical data which report that stainless steel treated with
boiling sulfuric acid undergo an increase in passivity after a 24
~ exposure to air l7 •

In these conditions the oxide so-produced improves the adhe-

sion properties of the system.

V. INFLUENCE OF AGEING BONDED SPECIMENS

In order to investigate the ageing effects on bonded structu-

res; single interface type specimens were submitted to a three weeks
exposure in ambient atmosphere. Although such an ageing may be con-
sidered as a short time, it has been reported that it is long
enough to modify rheological properties of bonded joints and bulk
adhesive I8 ,19. After ageing, identically prepared unaged samples
and aged samples were tested at the same time. Table IX shows the
ageing effects on the modulus of rigidity of the system and on the
maximum load values. The results concern adherend sheets submitted
either to chemical or thermal oxidation processes.

In the case of C-treated specimens, two opposite phenomena can

be observed. The decrease in (EI)S reflects a decay of the oxide
layer with exposure time in air. This correlates with previous
results (table VIII). This oxide layer degradation would be expec-
ted to lead to a decrease in Fmax' In fact, experimental Fmax va-
lues are nearly constant with, or without, ageing. This can be due
to the diffusion of water and/or ox6gen into the adhesive volume
and/or into the interfacial region l ,18. The diffusion process of
water may relieve internal stresses and/or transform some surface
oxide to hydroxide and so may create a weak boundary layer. In both
cases, adhesion is impaired. Diffusion process of oxygen through
the polymer may modify its mechanical properties and so enhances
adhesion. The combination of these two phenomena may allow to ex-
plain why the values of Fmax remain constant.

It is well known that treatment F produces a homogeneous oxide

layer, stable to air exposure. The oxide layer produced thermally
has a low permeability to water. In this case, a constant value of
(EI)S is predictable and was observed. Since the oxide layer is
unaffected by exposure to air, diffusion phenomena occur mostly at
oxide/adhesive interface or through the adhesive to the interface.
Then a Fmax increase can be expected if oxygen diffusion is pre-
vailing. This was experimentally observed.
98 F. GAILLARD ET AL.

Table IX. Ageing Effects on the Mechanical Properties of the System

Treatment C F

Ageing no three weeks no three weeks

Sample (EI)S Fmax (EI) S Fmax (EI) S Fmax (EI) S Fmax
number (N.mm ) (N) (N .mm ) (N) (N .mm ) (N) (N.mm 2 ) (N)
2 2 2

1 173.10 3 34 146.10 3 26 108.10 3 20 160.10 3 43

2 176.10 3 32 139.10 3 39 115.10 3 24 127.10 3 41
3 175.10 3 38 148. 10 3 40 135.10 3 19 153. 10 3 56
4 170.10 3 31 165.10 3 37 154.10 3 29 165.10 3 31
5 169.10 3 37 162.10 3 36 164.10 3 32 124.10 3 30
6 174.10 3 35 152.10 3 34 131.10 3 27 148.10 3 30
7 173.10 3 41 148.10 3 41 155. 10 3 21 143. 10 3 40
8 172.10 3 31 143.10 3 53 145.10 3 21 108. 10 3 25

Mean 177.10 3 35 150.10 3 38 138.10 3 24 141.10 3 37

values
± 0" % ± 1 % ± 10 % ± 6 % ± 20 % ± 14 % ± 21 % ± 14 % ± 27 %

VI. INFLUENCE OF THERMAL TREATMENTS

a) Influence of the duration of thermal treatments

The degreased adherend sheets were oxidized in air for various

durations at 300 and 600°C. CrLa , FeLa and OKa radiation intensities
were mesured using LEEIXS. These data are plotted with respect to
the duration of the thermal treatment as shown in figure 5.

The OKa intensity can be related to the thickness of the oxide

layer. Using a 4.5 keV electron beam, the depth probed is higher
than. 150 urn. This value is considerably greater than studied oxide
thicknesses.

For an identical duration of thermal treatment, the surface

laver formed at 600°C was found to be thicker than that formed at
300°C.

The CrLa radiation intensity is representative of the outer

21 urn of the oxide layer (electron energy: 1.5 keV). For 300°C
treated samples a regular decrease in the intensity of the CrLa
intensity was observed when the duration of treatment was increased
Conversely at 600°C, samples treated for more than 15 minutes exhi-
bited an enrichment of surface chromium. This is supported by the
uniform increase in the ratio of CrLa/FeLa intensities.
EFFECTS OF SURFACE CONDITIONING OF STAINLESS STEEL 99

_------·0
~C r

,~ _________________ oFe

>-
I-
Ui
Z
W
I-
oCr
~
.Fe
'0
-·600 ·C - 300 ·C TIME (min. )
30 60 90 120

Figure 5. CrLa , FeL a , OKa radiation intensities versus duration of

thermal treatment.

Panels oxidized in air at 600°C for various lengths of time,

were bonded and tested. This temperature was chosen to determine the
effects of chromium diffusion within the oxide layer. The results
of the mechanical test are given in table X.

Figure 6 compiles CrLa, OKa radiation intensities, Fmax and

(EI)S values versus the duration of treatment. The highest Fmax
values have been found to correspond to treatment durations provi-
ding chromium depleted surfaces. Cohesive failure of the oxide was
observed for specimens treated for more than 15 minutes at 600°C.
This leads to the hypothesis of failure below the chromium-enriched
surface layer. In addition a value increase in (EI)S can be obser-
ved up to 30 minutes treatment. For longer durations a decrease of
this narameter occured. This fact can be related to the apparent

Table X. Thermal Exposure Time Effects on the Mechanical Properties

of the System.

Treatment 15 30 60 120
0 I 5
duration (min. )
(EI) S 129 140 152 153 155 134 114
(10 3 N.mm 2 )
F (N) 15 21 28 29 21 23 23
max
100 F. GAILLARD ET AL.

~. .Cr
/ . _ _ _ _...-._ _ _ ·0

j."/
iJ

30 6 IZO

Figure 6. OKa, CrL radiation intensities, (EI)S and F values

versus duration ofathermal treatment. max

.- ----- -----../
Cr ./

e
"

(·C)
200 400 600

Figure 7. OKa , FeLa and CrLa radiation intensities versus treatment

temperature.
EFFECTS OF SURFACE CONDITIONING OF STAINLESS STEEL 101

inhomogeneity of the oxide layer which can be easily observed for

treatments of long duration.

b) Influence of thermal treatment temperature.

Degreased adherend sheets were air-oxidized at various tempe-

ratures for half an hour. The choice of this duration was deduced
from previous experiments results (figure 5). In figure 7, the
LEEIXS data are plotted against the air oxidation temperature ; the
relevant experiments were performed in the same conditions as des-
cribed previously.

The intensity of the OKa radiation and therefore the oxide

thickness increased with temperature of treatment up to 750°C. An
enrichment of surface chromium was observed for 8 ~ 450°C. This
fact is indicated both by the continuous increase of the intensity
of the CrLa radiation an by the decrease in the FeLa radiation
intensity. These data are in good agreement with the transformation
of the oxide layer structure at high temperatures (over 400°C)20.

Sheets treated at different temperatures were bonded and tes-

ted, the results are tabulated in table XI.

In the temperature range studied, Fmax increased continuously

while (EI)S increased and then remained constant. A delamination
along the oxide/adhesive interface was observed for specimens oxi-
dized up to 450°C, whereas cohesive failure, within the oxide layer,
was observed for higher temperatures. The failure locus changes
occur at temperatures at which chromium diffusion is observed. This
last phenomenon seems to be responsible for the localization of
failure within the oxide layer.

Table XI. Treatment Temperature Effects on the Mechanical Proper-

ties of the System.

Treatment 20 150 300 450 600 750

temperature (DC)

(EI) S 104 122 160 157 157 161

(10 3 N.mm 2 )

F (N) 15 18 22 23 24 30
max
102 F. GAILLARD ET AL.

CONCLUSION

The test used in this work undoubtedly appears to be a promi-

sing technique for measuring the practical adhesion in adhering
systems. Surface analytical techniques such as LEEIXS, combined
with an appropriate mechanical test provide chemical information
on the adherend surfaces prior to bonding or after delamination.
and therefore can allow a better understanding of the chemical
parameters which influence practical adhesion.

REFERENCES

1. A.A. Roche, J.S. Solomon and M.J. Romand, in "Microscopic

Aspects of Adhesion and Lubrication", J.M. Georges, Editor,
Tribology Seri~s 7, pp. 333-342, Elsevier, 1982.
2. A.A. Roche, A.K. Behme and J.S. Solomon, Inter. J. Adhesion
Adhesives,2, No.4, 249 (l982}.
3. R. Bador, M. Romand, M. Charbonnier and A. Roche, Adv. in X-ray
Analysis, 24, 351 (1981).
4. A.A. Roche~M.J. Romand and F. Sidoroff, these proceedings, pp.19-30.
5. A.A. Roche, M. Charbonnier, F. Gaillard, M. Romand and R. Bador,
Appl. Surf. Sci., 9, 227 (1981).
6. A. Roche, A. Cachard, R. Bador, F. Buiguez, M. Charbonnier and
M. Romand, J. Microsc. Spectrosc. Electron., ~, 627 (1977).
7. R. Bador, Thesis, Lyon, 1980.
8. C. F. Feldman, Phys. Rev., 117, 455 (1960).
9. R.P. Frankenthal, J. Electrochem. Soc., 116, 580 (1969).
10. P. Berge, Etude de la Stabilite de la Passivite des Aciers
Inoxydables en Solution Sulfurique, Publications Scientifiques
et Techniques du Ministere de l'Air, p. 37, Paris, 1961.
11. M. Nagayama and M. Cohen, J. Electrochem. Soc., 110, 670 (1963).
12. P. Berge, C. R. Acad. Sci., Paris, 245, 1239 (1957).
13. M. da Cunha Belo, B. Rondot, F •. Pong:-J. Ie Hericy and
J.P. Langeron, J. Electrochem. Soc., 124, 1317 (1977).
14. C. Hamilton, Appl. Polymer Symp., 19,-rD5 (1972).
15. P.M. Stifel, New Ind. Applications-ror Advanced Materials
Techn., 19th National SAMPE Symp. Exhibition, 75 (1974).
16. D.M. Brewis, J. Comyn, B.C. Cope and A.. C. Moloney, Polymer,
21, 1477 (1980).
17. ~ Harwood and Schulman, Corrosion, 203 (1949).
18. J. Schultz and K.C. Sehgal, Proc. l3eme Congres FAT IPEC, EREC Ed.,
Cannes, p. 575, 1976.
19. A. Roche, Thesis, Lyon, 1983.
20. S. Storp and R. Holm, Surf. Sci., 68, 10 (1977).
SURFACE CHARACTERIZATION OF ANODIC OXIDES ON ALUMINIUM ALLOYS BY
MEANS OF SURFACE POTENTIAL DIFFERENCE, SURFACE IMPEDANCE AND
SURFACE MORPHOLOGY

A. Kwakernaak, R. Exalto and H.A. van Hoof

Fokker B.V., Technological Centre

P.O. Box 7600, 1117 ZJ Schiphol-Oost
The Netherlands

To obtain acceptable adhesion and durability of

adhesive bonded structures, anodizing of aluminium
alloys in chromic acid or phosphoric acid is commonly
used in the aircraft industry. In order to find a
non-destructive testing technique to inspect pretreated
parts and to gain further understanding, we used some
simple techniques to characterize surfaces.
Using Transmission Electron Microscopy (TEM) as a tool
to characterize surface morphology showed that porous
oxides gave the best adhesion quality. The Surface
Potential Difference (SPD) measurement was found to be
very sensitive to adsorbed dipole layers on surfaces.
Surface Impedance (Z) measurements were introduced for
determination of the sealing quality of anodic oxides,
but can also be used to characterize non-sealed oxide
layers. On the basis of experimental data obtained, the
effects of several variables in the anodizing process
on SPD and Z measurements and TEM morphology are
discussed. Also the effects of ageing of anodic oxides
by dipping in ambient temperature water are shown. None
of the techniques discussed alone can fully
characterize the anodic oxide; the methods used are
supplementary to each other. SPD, Z, and surface
morphology can be described as functions of all
influencing parameters. With the use of such functions
characterization nomograms of anodizing processes can
be determined. Finally a new further refined impedance
analysis is proposed as a technique to characterize
more fully the anodic layer.

103
104 A. KWAKERNAAK ET Al.

INTRODUCTION

Adhesive bonded aluminium structures have been used in the

aircraft industry to an increasing extent for the last thirty
years. During these years more and more knowledge and understanding
of adhesive bonding has been gained by operational experience and
experimental research work. Most of the problems encountered were
at the interface between the metal surface and the adhesive and it
was recognized soon that, in order to obtain optimal adhesion and
durability, the surface treatment of the aluminium alloy was of
utmost importance. Nowadays, it is widely accepted that adhesion as
such can be attained by etching processes, such as the sodium di-
chromate-sulphuric acid etch or the chromic acid-sulphuric acid
etch 1 • However, to obtain a more durable bonded structure a
thicker oxide layer is needed. The chromic acid anodizing process
without hot water sealing, as used in Europe since 1953, and the
phosphoric acid anodizing, as used in the U.S.A. since 1974, are
the two most commonly used processes in the aircraft industry.

Since the first introduction of adhesive bonding as a

production process, much emphasis has been placed on the quality
control aspects of bonding. In addition to inspecting the cohesion
quality by nondestructive techniques, close control of the
pretreatment variables has always been carried out. To ensure
proper etching and anodizing of components, separate samples are
pretreated together with the production components. After bonding,
these samples are tested destructively and if their peel strength
is sufficient the pretreatment is accepted. In order to support the
empirical approach of the early days, and to find a non-destructive
technique to inspect pretreated parts for surface quality , new
ways to characterize surface quality have been explored. Today,
several techniques for surface characterization are available. The
methods range from simple physical measurements, such as Surface
Potential Difference or Voltapotentia1, to sophisticated methods
for surface analysis in ultra high vacuum such as X-ray
Photo-electron Spectroscopy (XPS or ESCA). In this paper attention
will be focussed on the three techniques which were introduced at
Fokker in the sixties, viz:
- Surface morphology by electron microscopy
- Surface potential difference
- Surface impedance.

Morphology of Anodic Oxides

Transmission Electron Microscopy (TEM) has earlier been used

to study the structure of anodic oxides 2 ,3. Early TEM work
carried out in cooperation between N.t.R., Fokker and T.N.O.4
gave an indication of the importance of the oxide morphology on
adhesion. From studies with variations in e.g. electrolytes,
SURFACE CHARACTERIZATION OF ANODIC OXIDES 105

Figure 1. Scanning Electron Micrograph of Cr03 anodized surface.

solution temperatures and treatment times, correlations were found

between oxide morphology and the adhesion quality as measured by
the peel test l ,5. New high-resolution scanning electron
microscopes provide even more detailed information on surface
morphology. Using stereopairs Venables et al. 6 were able to
develop oxide structure models. From both the earlier TEM work 7
and from micrographs using the same technique (Figure I), we found
an open pore structure for the European chromic acid anodizing
process with a pore size of about 400 A, and an average pore wall
thickness of about 150 A (Figure 2), which is quite different from

400 A

...
(
o
o

U.·n:ff..D·~.: ~
·:U ~:U·;
~ " .,
BARRIER LAYER .,.
~

Figure 2. Schematic representation of Cr03 anodic oxide.

106 A. KWAKERNAAK ET AL.

the model proposed by Venables et ale We varied the anodizing

parameters, but could not reproduce their results.

Surface Potential Difference of Anodic Oxides

The Surface Potential Difference (SPD) or Voltapotential is

the difference in workfunction between a reference surface and the
sample of interest. The extreme sensitivity of this technique to
changes in the oxide and to the presence of contamination had been
reported about 20 years ago 8 • The SPD can be measured with the
vibrating capacitor technique known as the Kelvin method 9 • Based
on this principle the Fokker Contamination Tester was developed lO •
With specimens etched in chromic acid-sulphuric acid solutions,
correlations were found between SPD and bondabilityl,lO.

With anodic oxides the situation seems to be more complicated.

Effects due to adsorption and desorption of water were monitored
with SPDlO and can be explained in terms of oriented dipoles or a
residual space charge ll ,l2 in the barrier layer.

This effect is also illustrated in Figure 3, where

experimental data on different anodic oxides (barrier layer and
porous oxides) are shown. The effect of adsorption of methanol on
the SPD change (A SPD) was measured and is plotted against the SPD
measured after anodizing. The same effect was also measured for two

1500

[j. SPO = SPO M -SPO 0

1000

>
E

~
tJ)
500 Cr03 -~S04 ETCH
<l

HF-HN03ETCH ----~
O +---------~------~----~~--~~~

-1000 -500 o 1000

SPO o (mV)

Figure 3. The effect of adsorption of methanol on the SPD of

anodic oxides (. barrier layer oxides, + porous oxides).
SURFACE CHARACTERIZATION OF ANODIC OXIDES 107

etching solutions and is shown in the same diagram. It seems that

when the residual space charge is large, it controls the
adsorption. The influence of the surface morphology is seen only
when the residual space charge is small (SPD~ 0). This shows the
sensitivity of the SPD measurement for adsorbed dipole layers on
substrate surfaces, while it can also be used to detect organic
contaminants 13 and changes in anodic oxides due to ageing.

Surface Impedance Measurements of Anodic Oxides

The surface impedance (Z) measurement of anodic coatings was

introduced as a simple method to measure the quality of sealing14
and nowadays is commonly used in industry15,16. The measurement
is carried out with a test cell with a well-defined measuring area
using 3.5% sodium chloride or 5% potassium sulphate as an electro-
lyte. The impedance or admittance of the oxide layer is generally
measured at a frequency of 1000 Hz. Some instruments are also able
to measure the dissipation or loss factor. The method was used in
our laboratory to measure all kinds of oxide layers. The effect of
different alkaline cleaning solutions prior to etching could be
measured, and thus we could discriminate the aggressiveness of the
alkaline cleaners 17 •

THE INFLUENCE OF PARAMETER VARIATIONS ON

SPD, Z AND MORPHOLOGY OF ANODIC OXIDES

Variations in aluminium alloy, pretreatment, anodizing process

and post-anodizing treatment are considered to be of great impor-
tance for the oxide film properties. In order to determine the
effect of some of these parameters on oxide layer properties,
several experiments were carried out. SPD was measured using the
Fokker Contamination Tester, which has a galvanic gold layer as a
reference surface. The impedance was measured with a Radiometer
vector impedance bridge at 1 kHz on an area of 12.9 mm 2 with a 5%
NaCI solution as electrolyte. The structure of the anodic oxide was
determined with transmission electron microscopy (TEM) which
provides information about the porosity of the anodic oxide and by
measuring the oxide layer mass (w), by stripping in a selective
oxide stripping solution (CrO) 20 g/dm3 , H3P04 50 g/dm3 ,
100°C, 5 min.).

The Effects of Pretreatment, Anodizing Temperature and

Anodizing Solution

1100 aluminium alloy (99.0% AI) was subjected to the following

treatments be'fore chromic acid anodizing:
- solvent cleaning with methyl ethyl ketone,
108 A. KWAKERNAAK ET AL.

alkaline cleaning in Turco 4215 S for 15 minutes,

- alkaline cleaning followed by etching in a chromic acid-
sulphuric acid solution.

Chromic acid anodizing was carried out in a fresh solution:

- Total chromic acid present as Cr 6+: 44 g/dm 3 Cr03

and in an aged solution:

- Chromic acid present as Cr 6+: 40 g/dm3 Cr03;

total present as Cr 3+ and Cr 6+: equivalent to 81 g/dm3 Cr03.

Anodizing was carried out at temperatures of 30, 40 and 50°C, with

voltage program for 40 minutes according to the Fokker standard
process 7 • The results of SPD, Z and surface morphology are shown
in Table I.

The SPD results show the interaction between pretreatment and

anodizing temperature. The effect of the aged anodizing solution on
SPD disappears at higher temperatures.

Table I. The SPD, Z and Surface Morphology After Cr03 Anodizing

of 1100 Aluminium Alloy.

PRETREATMENT

w
a:
...
:t
<C
a:
w
~
:Ii
...w
~
z
N
Q
o
zc

4.38 4.57
FRESH AGED FRESH AGED

ANODIZING SOLUTION
SURFACE CHARACTERIZATION OF ANODIC OXIDES 109

0 . 5 ,A.!!11

CLEANING MEK Cr03 -H:lS04 ETCHING

Cr03 ANODIZING 50°C Cr03 ANODIZING 50 ° C

Figure 4. The effect of pretreatment and anodizing temperature on

surface morphology (part of Table I).

Z decreases with increasing temperature and is not significantly

influenced by the pretreatment. The surface morphology shows the
effect of temperature and pretreatment and is shown in more detail
in Figure 4.

The Effect of Ageing by Water Soaking

Specimens of 2024-T3 A1c1ad (Clad layer: 99.5% A1) chromic

acid anodized were stored for different times up to 30 days in
deionized water at room temperature. After removing from the tank,
the specimens were dried at 60°C for 1 hour and SPD, Z, and the
mass increase were measured. The results are shown in Figure 5. The
SPD shows a rapid change to positive values after storage for a
short time and a change in behaviour after 150 hours of exposure.
The mass increases linearly with time. Z stays constant until 30
days when the specimen was visually corroded, which resulted in a
lower impedance. The surface morphology after 70 hours did not show
any change, but after 210 hours, the structure is changed (Figure 6).

As a reference, the morphology, SPD and Z are shown for a

specimen after dipping for 15 minutes in 100°C water, which shows
clearly a sealing effect. Figure 7 shows the results of Infrared
Spectroscopy, utiing specular reflection, of chromic acid anodized
surfaces.
110 A. KWAKERNAAK ET AL

'\
"'-
.....
>
S
~
E
J
-
Q
E

Go
N <l f/)

4 400 800

3 300 600

2 200 400

100 200

o o O#---~r---~----'-----~
100 200 300 400
-200 I
STORING TIME (hr)
-400

Figure 5. The influence of storing Cr03 anodic oxides in 25°C

water.

Cr03 REF 210 hr 25°C 15 min 100·C

SPD -400 mY SPD 370 mY SPD 440 mY
Z 3.15 Kl\ Z 3 . 3 Kl\ Z 14.2 Kl\

Figure 6. Electron micrographs of Cr03 anodized surfaces after

storing in water.
SURFACE CHARACTERIZATION OF ANODIC OXIDES 1 11

~~
I Cr0 3

~~68hr
<II
<II
L <II
i i SO°C 2 hr
<II
...
~~
Z
II:
~

'"~> '">
...-' ~
-'
100°C 15 min

'"
II: '"
II:

4000 2000 1000 4000 2000 1000

WAVENUMBER eM- 1 WAVENUMBER CM - 1

Figure 7. Reflection IR spectra of Cr03 anodic oxides after

storing in water.

The change in SPD after 150 hours corresponds with the appearance
of a peak at 3400 cm- l • The sealed surface morphology which
appears after 15 minutes in 100°C water or after 30 minutes in 60°C
water, results in a different peak in the IR spectrum at
3100-3600 em-I.

The Effects of Anodizing Time and Voltage

Chromic acid anodizing at a temperature of 40°C in a solution

of 40 g/dm3 Cr03 was carried out with 2024-T3 Alclad specimens.
The anodizing time and voltage were varied, viz. 5 and 15 minutes
and 20 and 40 Volts. The results are shown in Table II. Phosphoric
acid anodizing was carried out with specimens of the same alloy at
a temperature of 25°C in a solution of 100 g/dm3 H3P04. Also
anodizing time and voltage were varied, viz. 5 and 25 minutes and
10, 25 and 40 Volts. The results are shown in Table III.
112 A. KWAKERNAAK ET AL.

Next to the SPO, Z and w values in Tables II and III, climbing

drum peel test results using test specimens with a width of 25 nun
bonded with a vinyl phenolic adhesive (Redux 775 Lip) are given.
The electron micrographs (TEM) shown were taken using two different
specimen preparation techniques; on the left-hand side by the
carbon replica technique and on the right-hand side by the oxide
transmission technique. The results of both the chromic acid and
the phosphoric acid anodizing show generally a similar behaviour.
The pore size increases with increase in voltage and the thickness
of the porous layer (or layer mass) increases with time and
voltage. The gPO decreases and Z increases with increase in
voltage. At higher voltage Z decreases with increase in time. Only
the effect of time on gPO shows a difference between chromic acid
and phosphoric acid anodizing. With phosphoric acid the gPO did not
show any effect, but with chromic acid the gPO increases with
increase in time.

Table II. er03 Anodizing of 2024-T3 A1c1ad.

~
!
w
::IE
j:

"N
Z -+-=--..;....~

Q
o
Z
<C

15
-535
3.50 5.iS
2180 3080
402 412
SURFACE CHARACTERIZATION OF ANODIC OXIDES 113

Table III H3P04 Anodizing of 2024-T3 Alclad.

40Y

-380
L&I 5 . 13
:I 830
l- 87
e!
z
N
Q
o
z
«
SPD (mY) 270 -390
25 Z (KI\) 1.62 4 .25
w(mg / m 2) 650 2050
P (N) 370 159
O. 5AUl1

REPRESENTATION OF OXIDE FILM PROPERTIES BY CHARACTERIZATION

NOMOGRAMS

From these experiments, it can be seen that several parameters

largely influence the quality and structure of the anodic oxide.
Next to the effects of pretreatment, anodizing voltage, time,
temperature and oxide ageing by water adsorption, there are other
influences such as the alloy composition, impurities in solutions,
rinsing temperature and time, which will lead to changes in SPD, Z,
layer mass and surface morphology. None of the discussed techniques
alone can fully characterize the anodic oxide. The measuring
methods are supplementary to each other.

To find the relationship between a measured quantity y and the

investigated processing parameters, the method of regression
analysis was used l8 ,l9. In general the dependent variable y ~s a
linear function of m independent variables Xj which can be
written as:
114 A. KWAKERNAAK ET AL.

Yi = bO + bl xil + b2 xi2 ••••• bm xim + ei,

where the index i identifies a particular set of processing

parameters, and ei is the experimental error. The regression
coefficients bO' bl' ••••• bm can be determined by least
squares fitting. Also the covariance matrix can be developed, which
shows the dependency of one parameter on the others and gives the
standard error for each of the regression coefficients.

SPD, Z, oxide film mass, and surface morphology were

determined in a chromic acid (50 g/dm3 Cr03) anodizing
experiment on 2024-T3 Alclad material. The following parameters
were varied: process time, 5, 25 and 45 minutes; voltage, 10, 25
and 40 Volts; and solution temperature, 30, 40 and 50 G C.

With the measured data the following regression equations were

deduced:

SPD -590 - 13 t' - 521 V' + 36 T' + 42 t'V' + 79 ViTI (mV)

Z 3.28 - 0.31 t' + 1.69 V' - 0.10 T' - 0.31 t'V' (ka)
w = 4174 + 3138 t' + 1876 V' + 2000 T' + 1547 t'V' + 1366 to'T' +
1139 ViTI + 929 t'V'T' (mg/m 2 )

The parameters t', V', T' represent, respectively, the process

time, anodizing voltage and solution temperature and are expressed
as dimensionless parameters, where -1 ~ t', V', T' ~ 1.
The values of the multiple correlation coefficient (r) as well as
the standard error of estimate (s) are:

rSPD 0,973 sSPD 115 mV

rZ 0,973 Sz 0,37 kn

0,994 Sw = 490 mg/m 2

To allow a graphical characterization, nomograms can be developed

with the regression equations, showing for a fixed value of one
parameter the effect of the other parameters on SPD, Z and w.
Figure 8 shows the characterization nomogram for the Cr03 acid
anodizing process at 40 V, showing the effects of time and tempera-
ture. For this nomogram, the change in surface morphology is also
given by the electron micrographs in figure 9.

Such a characterization nomogram, developed for a certain

production process, would allow the detection of processing
deviations.
SURFACE CHARACTERIZATION OF ANODIC OXIDES 115

(,)
!.. 45
I&.t
CIC
::I
~
c
ffi 40
0.
~
I&.t
~

35

30 ~~~~~--~~-3~~~
5 10 15 20 25 30 35 40 45
TIME (min)
- - - -SPD (mY)
_.- Z (Kn)
- - - w (mg / ni)

Figure 8. Characterization nomogram for Cr03 (50 g/dm 3 )

anodizing at 40 v.

0 . ) ,AIm

Figure 9. Electron micrographs of the characterization nomogram

for Cr03 anodizing at 40 v.
116 A. KWAKERNAAK ET AL.

FREQUENCY ANALYSIS OF THE SURFACE IMPEDANCE

The surface impedance measurements mentioned before were all

performed at a fixed frequency of 1 kHz. For a simple homogeneous
oxide layer the capacitance and loss factor can be obtained from
the magnitude and phase, respectively. These represent
approximately the barrier layer of the porous oxides. Most oxide
layers occurring in practice are far from being homogeneous. The
porous layer is much thicker than the barrier layer (Figure 2), but
has a high conductivity due to the penetrating electrolyte. The
layers need not be homogeneous in a direction perpendicular to the
surface nor in directions along 20 the surface. Nevertheless, the
electrical properties of such complex systems for small AC signals
can always be represented by an equivalent circuit consisting of
simple components, like resistors and capacitors. The validity of
such models can only be checked by measuring the frequency
dependence of the surface impedance 21 • The range of frequencies
should be as wide as possible. A computer program was developed to
fit a complex equivalent circuit to the measured frequency
response. Each component is switched in parallel or in series with
the circuit formed by the preceding components. The values of some
or all the components are considered to be adjustable parameters in
a non-linear least squares fit. As the derivative of the complex
impedance with respect to the parameters is known for each
component, a local linear approximation can be made. The resulting
linear least squares problem is then solved with singular value
decomposition 2Z •

The number of iterations required depends strongly on the com-

plexity of the model and on the ratio between the largest and
smallest singular value. After convergence the program' calculates
the covariance matrix of the parameters. This is used mainly to
evaluate the significance of changes in the parameters for
different surfaces. In addition to normal resistors and capacitors
a very useful component is the constant phase angle element 23
defined by its admittance: Y = Q (i~)P. Here Q and Pare J;>oth
considered to be adjustable parameters. For P close to 1, this is a
capacitor with a frequency-independent loss factor. Sometimes
inductive elements are also needed 24 • Both infinite and finite
Warburg impedances can also be treated.

Preliminary results on both Cr03 and H3P04 anodized

surfaces indicate that indeed a good fit with correlations of the
order of 95% can be attained over the frequency range 1-105 Hz.
The fit for the Cr03 anodizing process, shown in Figure 10, is
dominated by the capacitance of the barrier layer. After sealing in
boiling water (Figure 11), the effect of the barrier layer remains
but, fot higher frequencies, is exceeded by the increased impedance
of the sealed porous layer.
SURFACE CHARACTERIZATION OF ANODIC OXIDES 117

90

PHASE
60
-3
"I "w
(I)

c: w
V
30~
w
" - 4 av
>-
v
(!)
"v
0>-
o W
(I)
...J
-5 «
I:
"-
-30

-6

~: - 60

- 7~------r------'-------r------~-----+- 90
o 2 3 4 5
LOG(F ) (H-z)

Figure 10. Surface admittance (measured values and calculated

fit) of Cr03 anodic oxide.

-2~ ______ ~ ____--'______- L______ ~ +

______ 90

60
-3
"w
(I)

30~
{!)
w
" - 4 a
v

v
(!)
"v
0>-
o W
(I)
...J
-5 «
I:
"-

aln·
- 30

-6

lta1 - 60

-7~-----'r------r------~-----.-------r -90
o 2 3 4 5
LOG(F) (H-z)

Figure 11. Surface admittance (measured values and calculated fit)

of sealed Cr03 anodic oxide.
118 A. KWAKERNAAK ET AL.

CONCLUSIONS

The measurement of SPD, Z, w, and surface morphology can

provide information about the formation history of oxide layers.
The surface impedance is largely influenced by the anodizing
voltage and slightly by the process time. The surface potential
difference depends mainly on the anodizing voltage and to a lesser
extent on the time and temperature. It also provides information
about rinsing and ageing after anodizing. The oxide layer mass
depends strongly on process time, temperature and anodizing
voltage. The surface morphology is largely influenced by
temperature and less by voltage and pretreatment.

In order to obtain a more accurate view of the effect of

different parameters, it is possible to find equations describing
the relationship between the measured quantities and the
influencing parameters by regression analysis. With these equations
characterization nomograms can be developed, representing the
effect of process variations for a certain anodizing process.

Some important structural elements of the anodic oxide can be

described by an electrical analogue. This model can be developed by
fitting the equivalent circuit to the measured surface impedance of
the oxide over a wide range of frequencies.

REFERENCES

1. P.F.A. Bijlmer and R.J. Schliekelmann, SAMPE Quarterly 1, 13

(1973).
2. F. Keller, M.S. Hunter and D.L. Robinson, J. Electrochem. Soc.
100, 411 (1953).
3. ~Ginsberg and K. Wefers, Metall. 17, 202 (1963).
4. A. Hartman, N.L.R. Report TN M.209l~ational Aerospace
Laboratory the Netherlands, Amsterdam (1961).
5. P.F.A. Bijlmer, J. Adhesion 5, 319 (1973).
6. J.D. Venables, D.K. McNamara~ J.M. Chen, T.S. Sun and
R.L. Hopping, Appl. Surface Sci. 3, 88 (1979).
7. P.F.A. Bijlmer, Metal Finishing 70, 30 (April 1972).
8. A. Matting and K. Ulmer, Kautschuk und Gummi. Kunststoffe ~,
280, 387 (1963).
9. W. Thomson (Lord Kelvin), Phil. Mag. 5, 82 (1898).
10. P.F.A. Bijlmer in "Surface ContaminatIon: Genesis, Detection
and Control", K.L. Mittal, Editor, Vol.2, pp. 723-748,
Plenum Press, New York, 1979.
11. C.E. Michelson, J. Electrochem. Soc. 115, 213 (1968).
12. A.T. Fromhold, Jr. and J. Kruger, J. Electrochem. Soc. 120, 722
(1973).
13. T. Smith and G. Lindberg, Surface Technol. 2, 1 (1979).
SURFACE CHARACTERIZATION OF ANODIC OXIDES 119

14. E.T. Englehart and D.J. George, Materials Protection l, 24

(Nov. 1964).
15. H. Birtel and W. Leute, Aluminium 43, 93 (1967).
16. A. Ott, Aluminium 52, 491 (1976). --
17. R.J. Schliekelmann:-Non Destructive Testing 5, 79 (1972).
18. L.M. Rose, "The Application of Mathematical Modelling to
Process Development and Design", Applied Science Publishers
Ltd., London, 1974.
19. O.L. Davies, "The Design and Analysis of Industrial
Experiments", Oliver and Boyd, London, 1967.
20. R. Leek and N.A. Hampson, Surface Technol. 7, 151 (1978).
21. J. Ross MacDonald and J.A. Garber, J. Electrochem. Soc. 124,
1022 (1977). -
22. G.H. Golub and C. Reinsch in "Handbook for Automatic
Computation, Vol. II: Linear Algebra", J .H. Wilkinson and
C. Reinsch, Editors, pp. 134-151, Springer Verlag, 1971.
23. K.S. Cole and R.H. Cole, J. Chem. Phys. 9, 341 (1941).
24. H.J. de Wit, C. Wijenberg and C. Creveco;ur, J. Electrochem.
Soc. 126, 779 (1979).
INVESTIGATION INTO THE EFFECT OF SURFACE TREATMENT ON THE

WETTABILITY AND THE BONDABILITY OF LOW SURFACE ENERGY MATERIALS

J.P. Jeandrau

Etablissement Technique Central de l'Armement

Centre Mecanique-Chimie-Materiaux
16 bis, avenue Prieur de la Cote d'Or
94114 Arcueil Cedex, France

An experimental effort has been undertaken to

examine the effect of surface treatment on various low
surface energy thermoplastic materials to promote
wettability and bondability of these substrates.
Changes in wettability were followed by contact angle
measurements. These measurements were correlated with
the bondability of treated surfaces using a usual two-
part epoxy adhesive. Air-plasma cleaning is a more
effective treatment than the usual chemical-acid
etching in increasing the surface energy of polymers to
make them more wettable and bondable, except for
polypropylene and for polytetrafluoroethylene. This
work has enabled several satisfactory solutions for
bonding these substrates by selecting suitable surface
treatments and adhesives for each.

121
122 J.P. JEANDREAU

INTRODUCTION

The continuing increase in the use of thermoplastic materials

has led to investigate the different means of joining them to-
gether. The use of adhesives offers many advantages when compared
to other more conventional methods such as welding, riveting and
bolting. A necessary, though sometimes insufficient, requirement
for developing strong adhesive joints is the establishment of
intimate molecular contact at the interface. Many theories or the
mechanisms of adhesion are found in the literature 1 ,2,3,4. For
adhesive joints, the adsorption and wetting theory is the most
generally accepted one. Wetting may be quantitatively defined by
reference to a liquid drop resting in equilibrium on a solid
surface.

The main difficulty for bonding thermoplastic materials is due

to their low· surface free energies. Zisman and co-workers 5 , 6
established that, for low energy solids and a series of liquids, a
rectilinear relationship frequently existed between the cosine of
the contact angle, cos e, and the surface tension of the wetting
liquid, YLV • He defined the critical surface tension of wet-
ting, Yc ' as the value to which YLV tends as cos e approaches
unity, i.e. as e approaches zero degree. Then he used the
critical surface tension to characterize and compare the
wettability behavior of low-energy surfaces.

Sharpe and Schonhorn 7 proposed that the single most important

factor influencing adhesive joint strength is the ability of the
adhesive to spread spontaneously on the substrate. They developed
a criterion for the case when the adhesive will spontaneously
spread on the substrate and, using Zisman's critical surface
tension concepts, proposed that the Y of the adhesive must be
c
less than or equal to that of the substrate. When we look at the
values of y given in Table I, it becomes easy to understand
c
why thermoplastic materials such as polyethylene and poly-
tetrafluoroethylene are difficult to bond with usual adhesives.

However, other ways to characterize surfaces of low-energy

materials exist and are reported in the literature 8 ,9 In these
two articles, Mittal has reviewed and discussed the relationship
between Y or other surface energetic parameters and joint
c
strength.
EFFECT OF SURFACE TREATMENT ON WETTABILITY 123

The purpose of this work was to investigate different surface

treatments for increasing the wettability of seven
difficult-to-bond substrates. Various surface treatments of
polymers are reported in the literature 10 • These surface
treatments include the usual chemical acid-etching and also
air-plasma cleaning. This last surface treatment had been reported
by Schonhorn and Hansen 11 to be highly effective for the surface
preparation of low surface energy polymers to enhance adhesive
bonding. Essentially, the technique consists of exposing the
polymer surface to a gas plasma at a reduced pressure generated by
an electrodeless glow discharge.

Table I. Values of Critical Surface Tensions of Polymers at Room

Temperature.

Solid Surface Critical Surface Tension *

Y c (m N m -1)

Polytetrafluoroethylene 18.5
Silicone rubber 22.0
Polyurethane 29.0
Polyethylene 31.0
Polystyrene 32.8
Polychloroprene 38.0
Polymethylmethacrylate 39.0
Nylon 11 42.0
Nylon 6.6 42.5
Phenol-resorcinol adhesive 52.0
Urea-formaldehyde resin 61.0

* Values from reference 3

The change in surface energy resulting from surface treatment

was characterized by the change in the contact angle of water with
the surface of the material being evaluated.

The increase in the bondability of the substrates was

characterized by the shear strength resulting from treated
specimene bonded with usual adhesives.

The main object of this work was to select the most suitable
surface treatments and adhesives required for developing strong
adhesive joints.
124 J.P. JEANDREAU

EXPERIMENTAL

Substrates

Seven semi-crystalline thermoplastic substrates were selected

because of their low energy surfaces:

Polytetrafluoroethylene PTFE (TEFLON, Dupont de Nemours)

Low density polyethylene LDPE (ARCODUR, Arcotec)
High density polyethylene HDPE (6M50/50, Hoechst)
Polypropylene PP (PPH 2210, Hoechst)
Polyoxymethylene POM (DELRIN 15, Dupont de Nemours)
Polyamide 6.6 PA 6.6 (ULTRAMIDEA, BASF)
Polyamide 11 PA 11 (BMNO RILSAN, Ato-Chimie)

Surface Treatments

The following surface treatments were selected to improve the

wettability of each substrate:

No treatment. Samples were degreased by placing them in

absolute ethanol in an ultrasonic cleaner for 15 minutes, then
dried at 80°C for 30 minutes.

Chromic sulfuric etching. After the first operation of

degreasing, samples were immersed in etching solution for one hour
at room temperature. The samples were thoroughly rinsed in
deionized water and dried at 80°C for 30 minutes.

The etching solution was composed of:

2000 ml sulfuric acid (1.84 sp. gr.)

184 g potassium dichromate
294 ml deionized water.

Naphthalene-sodium etching. This surface treatment is

specific and recommended for polytetrafluoroethylene. After
degreasing in pure ethanol, samples were immersed for 48 hr at room
temperature in the following etching solution:

500 ml tetrahydrofuran
128 g naphthalene
24 g sodium

Then samples were thoroughly rinsed in tetrahydrofuran and in

deionized water and dried at 80°C for 30 minutes.
EFFECT OF SURFACE TREATMENT ON WETTABILITY 125

Air-plasma cleaning. The equipment used for generating the

gas plasmas consisted of a vacuum system capable of achieving
10 - 4 to 10 - 5 torr pressure and a tubular reactor (200 mm
high and 400 mm in diameter). A high voltage was applied across
the cathode (a 200 mm long and 30 mm in diameter aluminium
cylinder) and the top and bottom of the reactor. After degreasing
in absolute ethanol. samples were placed in the apparatus. at a 200
mm constant distance from the cathode. and continuously exposed to
the gas plasma. During the IS-minute air-plasma treatment the
power intensity and gas pressure were nominally 50 Wand 5.10- 2 torr
After the air-plasma treatment was completed. the samples were
immediately bonded (15 minutes maximum after treatment). Some work
was done to optimize the plasma treatment process. and the effect
of different parameters on wettability and bondability were
evaluated: plasma atmosphere. exposure time. power intensity. gas
pressure and the elapsed time between the end of the treatment and
the bonding operation. The operating conditions given above were
selected with respect to these preliminary results on polyethylene
substrate.

Measurement of Contact Angles

Distilled water was used to determine the contact angles.

Constant volume droplets of 0.5 ~l were delivered from the

stainless steel tip of a 5 ~l syringe onto the substrate. The
contact angle of the droplet was taken one minute after deposition
using a KERNCO goniometer. This interval was sufficient for the
droplet to attain an equilibrium contact angle B. yet short enough
for negligible evaporation loss. Each contact angle quoted is the
mean of at least four measurements on droplets placed on the
sample.

Evaluation of Adhesive-Bonded Joints

Bond strengths were determined using 10 mm thick thermoplastic

substrate shear specimens bonded as shown in Figure 1. The bonded
area was 625 mm 2 •

Each shear bond strength quoted is the mean of at least ten

specimens tested.
126 J.P. JEANDREAU

~F-

rn
1 "" adhes(O.1mm
Ive
thick)
T
Figure 1. Single lap shear test specimen.

Adhesives

Four structural adhesives were selected for bonding different

substrates.

EC 2216 B/A is a two-part epoxy adhesive curing at room

temperature (3M Company)

AW 116/HV 953U is a two-part epoxy adhesive curing at room

temperature (CIBA GEIGY)

UK 8212/UK 5400 is a two-part polyurethane adhesive curing at

room temperature (HENKEL)

Multibond 329 is a toughened acrylic adhesive curing at room

temperature with the initiator No. 738 (LOCTITE)

RESULTS AND DISCUSSION

Table II summarizes the effect of various surface treatments on

contact angles and shear strengths of adhesive bonded joints for
the seven low-energy thermoplastic substrates investigated.

The measured contact angles of untreated substrates are in good

agreement with those found in the literature (94 0 for low density
EFFECT OF SURFACE TREATMENT ON WETTABILITY 127

12 12
polyethylene , 108 0 for polytetrafluoroethylene and 90 0
13
for polypropylene Three important conclusions may be drawn
from the test results:

1) For untreated surfaces, the shear strengths increase as the

contact angles decrease (i.e. as the substrate surface
energies increase). It clearly shows why untreated polyamides
give stronger adhesive bonded joints than polyolefins which
have lower surface energies (when the substrates are bonded
with the same adhesive under the same conditions).

2) The second fact observed is that the usual chromic sulfuric

etching surface treatment leads to a weak improvement of the
wettability and bondability of low energy surface
thermoplastic materials. With this surface treatment, shear
strengths are increased by only a facto 7 of two or three in the
Jest cases.

3) The third conclusion is that the air-pla;ma cleaning is very

effective in increasing the surface energy of polymers
(contact angles decrease and tend to zero degree ) to make
them more wettable except for polytetrafluoroethylene. In
five cases, wettability and bondability are enhanced by
air-plasma treatment. For these five substrates, contact
angles decrease from more than 70 degrees to less than 17
degrees and shear strengths increase from very low values to
more than 7 MPa. So air-plasma treatment seems to be a very
effective treatment to improve wettability and bondability of
the five following thermoplastic materials: polyethylene (low
density and high density), polyoxymethylene and polyamides
(6.6 and 11). However, exceptions are observed for
polytetrafluoroethylene and polypropylene. For the first
(PTFE), air-plasma treatment improves neither wettability
(contact angles increasing) nor bondability (shear strength
staying constant and very low, i.e. 1.1 MFa). The only
efficient surface treatment to improve wettability and
bondability of polytetrafluoroethylene is the naphthalene
sodium etching. For polypropylene, the air-plasma treatment
increases the wettability of the substrate but surprisingly
the bondability is not improved.

The understanding of the mechanism of plasma surface treatment

in enhancing the adhesive properties of polymer surfaces is still a
.
controvers1al .
subject 13 ' 14 , 15 , 16 , 17 , 18 • The object1ve
. of our
investigation was to demonstrate a trend in improvement in the bond
strength between thermoplastic substrates. As plasma treatments
are reported to be efficient surface treatments for polymer
surfaces, we have simply used this technique without attempting to
128 J.P. JEANDREAU

Table II. Effect of Surface Treatment on Wettability (Contact

Angles) and Bond Shear Strength of Low Energy Thermoplastic
Substrates.

Thermoplastic Surface Contact Shear Mode of

Substrate Treatment (a) Angle Strength (b) failure
(MPa) (c)

Poly tetra- NT 1.1 A

fluoroethylene APC 1.1 A
PTFE NSE 6.3 C*
Low density NT 93° 1.6 A
polyethylene CSE 74.5° 3.4 A
LDPE APC 0° 8.3 C
High density NT 86° 1.6 A
polyethylene CSE 71° 3.5 A
HDPE APC 0° 8.3 C
Polypropylene NT 92° 1.3 A
PP CSE 90° 3.6 A
APC 10° 1.8 A
Polyoxy- NT 82° 2.7 A
methylene CSE 89° 4.0 A
POM APC 17° 8.9 C
Polyamide 6.6 NT 76° 5.2 A
PA 6.6 APC 0° 10.8 C
Polyamide 11 NT 4.1 A
PA 11 APC 7.6 C

a) NT: No treatment; CSE: chromic-sulfuric etching

APC: air-plasma cleaning; NSE: naphthalene sodium etching

b) Shear strengths were determined on ten test specimens bonded

with the EC 2216 BfA two-part epoxy adhesive.

c) A: interfacial failure
C: cohesive failure in the adhesive
C*: the failure occurs in the substrate in the surface
region.

The. locus of failure was determined by unaided eye. However, we

have to be careful because many times, a failure may look interfacial
by visual inspection, but the sophisticated analysis (eg, by ESCA)
shows that a thin layer of adhesive is left behind on the
substrate. The question of locus of failure and its precise
determination by surface analytical techniques has been widely
discussed in the literature 19 , 20, 21
EFFECT OF SURFACE TREATMENT ON WETTABILITY 129

contribute to a better understanding of the physical or chemical

modification of the substrate surfaces. The fact that this surface
treatment did not improve the bondability of polypropylene and
polytetrafluoroethylene under our experimental conditions does not
mean that it is inefficient on these substrates under all
conditions. As a matter of fact, Tira 13 reported that
polypropylene samples treated in an oxygen plasma at 100 W for
30 minutes could be stored after treatment over a period of two
weeks and then successfully bonded with an RTV silicone.

The shear strength results in Table III confirm the

inefficiency of plasma treatment to improve the bondability of
polytetrafluoroethylene by adhesives from different chemical
families and of different rigidities. These results confirm also

Table III. Shear Strength of Polytetrafluoroethylene Bonded to

Polytetrafluoroethylene.

Adhesive Surface Shear Strength (b) Mode of

Treatment (a) (MPa) Failure (c)

EC 2216 AlB NT 1.1 A

Epoxy APC 1.1 A
NSE 6.3 c*

UK 8812/UK 5400 NT 1.1 A

Polyurethane APC 1 A
NSE 8.4 c*

Multibond 329 NT 0.7 A

Acrylic APC 1.5 A
NSE 5.4 c*

AW 116/HV 953U APC 1.5 A

Epoxy

a) NT: No treatment; CSE: chromic-sulfuric etching

APC: air-plasma cleaning; NSE: naphthalene sodium etching

b) Shear strengths were determined on ten test specimens

c) A: interfacial failure
C: cohesive failure in the adhesive
C*: the failure occurs in the substrate in the surface
region.
130 J.P. JEANDREAU

Table IV. Shear Strength of Low Density Polyethylene Bonded to

Low Density Polyethylene.

Adhesive Surface Shear Strength (b) Mode of

Treatment (a) (MPa) Failure (c)

EC 2216 A/B NT 1.6 A

Epoxy CSE 3.4 A
APC 8.3 C

UK 8212/UK 5400 NT 1.8 A

Polyurethane APC 7.5 C

Multibond 329 NT 1.0 A

Acrylic APC 3.2 A

AW 116/HV 953U CSE 6.0 A 50% C 50%

Epoxy APC 9.0 C

Table V. Shear Strength of High Density Polyethylene Bonded to

High Density Polyethylene

Adhesive Surface Shear Strength (b) Mode of

Treatment (a) (MPa) Failure (c)

EC 2216 A/B NT 1.6 A

Epoxy CSE 3.5 A
APC 8.3 C

UK 8212/UK 5400 NT 1.8 A

Polyurethane APC 9.3 C

Multibond 329 NT 1.2 A

Acrylic APC 3.1 A

AW 116/HV 953U NT 1.7 A

Epoxy APC 8.4 C

a) NT: No treatment; CSE: chromic-sulfuric etching

APC: air-plasma cleaning; NSE: naphthalene sodium etching

b) Shear strengths were determined on ten test specimens

c) A: interfacial failure
C: cohesive failure in the adhesive
EFFECT OF SURFACE TREATMENT ON WETTABILITY 131

that the naphthalene sodium etching is the only efficient surface

treatment for bonding polytetrafuoroethylene. This last surface
treatment leads to shear strengths greater than 5 MPa with the
different adhesives used. All the failures occur in the polymer
surface region because of the decrease in the cohesive strength of
the surface layer due to the chemical etching.

Tables IV and V summarize the effect of changing various

parameters (adhesives and surface treatments) on polyethylene shear
strengths (low density and high density respectively). These
results show and confirm that air-plasma cleaning is the only
efficient surface treatment for bonding polyethylene. Usual
adhesives (epoxy and polyurethane) lead to cohesive failures (shear
strength greater than 7.5 MPa) on air-plasma treated samples.
However, the acrylic adhesive does not adhere well either on low
density polyethylene or on high density polyethylene even after
air-plasma treatment.

Table VI summarizes the shear strength results of bonded

polypropylene samples. This table shows that all failures are
interfacial in nature irrespective of the adhesive used or surface
treatment employed. These results confirm that the surface
treatments investigated on polypropylene do not improve the
bondability of polypropylene substrate.

Table VII summarizes the shear strength results of bonded

polyoxymethylene samples. These results confirm the efficiency of
air-plasma cleaning to improve wettability and bondability of
polyoxymethylene. Cohesive failures occur on air-plasma treated
samples bonded with epoxy and acrylic adhesives. Interfacial
failures at a lower strength level (~6 MPa) occur when a
polyurethane adhesive is used on the treated surfaces.

Tables VIII and IX summarize the results of adhesive bonded

polyamide shear strengths. These results show and confirm that
strong adhesive bonded joints (shear strengths higher than 7 MPa)
occur when epoxy and polyurethane adhesives are used on plasma
treated polyamide surfaces. However, the acrylic adhesive leads to
interfacial failures in every case. The shear strength is not
affected by the surface treatment when samples are bonded with
Multibond 329.
132 J.P. JEANDREAU

Table VI. Shear Strength of Polypropylene Bonded to Polypropylene.

Adhesive Surface Shear Strength (b) Mode of

Treatment (a) (MPa) Failure (c)

EC 2216 A/B NT 1.3 A

Epoxy CSE 2.8 A
APC 1.8 A

UK 8212/UK 5400 NT 1.0 A

Polyurethane APC 2.7 A

Multibond 329 NT 0.8 A

Acrylic APC 1.8 A

AW 116/HV 953U APC 3.8 A

Epoxy

Table VII. Shear Strength of Polyoxymethylene Bonded to

Polyoxymethylene.

Adhesive Surface Shear Strength (b) Mode of

Treatment (a) (MPa) Failure (c)

EC 2216 A/B NT 2.7 A

Epoxy CSE 4.0 A
APC 8.9 C

UK 8212/UK 5400 NT 2.9 A

Polyurethane APC 4.8 A

Multibond 329 NT 1.9 A

Acrylic APC 6.1 C

AW 116/HV 953U APC 10.1 C

Epoxy

a) NT: No treatment; CSE: chromic-sulfuric etching

APC: air-plasma cleaning; NSE: naphthalene sodium etching

b) Shear strengths were determined on ten test specimens

c) A: interfacial failure
C: cohesive failure in the adhesive
EFFECT OF SURFACE TREATMENT ON WETTABILITY 133

Table VIII. Shear Strength of Polyamide 6.6 Bonded to Polyamide

6.6.

Adhesive Surface Shear Strength (b) Mode of

Treatment (a) (MPa) Failure (c)

EC 2216 A/B NT 5.2 A

Epoxy APC 10.8 C

UK 8212/UK 5400 NT 3.0 A

Polyurethane APC 8.8 C

Multibond 329 NT 6.0 A

Acrylic APC 6.3 A

AW 116/HV 953U NT 5.0 A

Epoxy APC 10.2 C

Table IX. Shear Strength of Polyamide 11 Bonded to Polyamide 11.

Adhesive Surface Shear Strength (b) Mode of

Treatment (a) (MPa) Failure (c)

EC 2216 AlB NT 4.1 A

Epoxy APC 7.6 C

UK 8212/UK 5400 NT 4.6 A

Polyurethane APC 10.4 C

Multibond 329 NT 2.9 A

Acrylic APC 3.4 A

AW 116/HV 953U NT 3.7 A

Epoxy APC 9.4 C

a) NT: No treatment; CSE: chromic-sulfuric etching

APC: air-plasma cleaning; NSE: naphthalene sodium etching

b) Shear strengths were determined on ten test specimens

c) A: interfacial failure
C: cohesive failure in the adhesive
134 J.P. JEANDREAU

Table X gives the most suitable surface treatments and

adhesives required for developing strong adhesive joints on low
energy thermoplastic materials bonded together. These results show
that the difficult-to-bond substrates can be strongly bonded with
usual adhesives when suitable prior surface treatment is carried
out on the surface.

Table X. Suitable Surface Treatments and Adhesive Required for

Developing Strong Adhesive Joints on Low Energy Thermoplastic
Materials (Shear Strength> 7 MPa).

Thermoplastic Suitable Suitable adhesive Shear

Material Surface Strength (MPa)
Treatment

PTFE NSE UK 8212/UK 5400 8.4

(polyurethane)

LDPE APC EC 2216 A/B (epoxy) 8.3

APC UK 8212/UK 5400 7.5
(polyurethane)
APC AW 116/HV 953U (epoxy) 9.0

HDPE APC EC 2216 A/B (epoxy) 8.3

APC UK 8212/UK 5400 9.3
(polyurethane)
APC AW 116/HV 953U (epoxy) 8.4

POM APC EC 2216 A/B (epoxy) 8.9

APC AW 116/HV 953U 10.1

PA 6.6 APC EC 2216 A/B (epoxy) 10.8

APG UK 8212/UK 5400 8.8
(polyurethane)
APC AW 116/HV 953U 10.4

PA 11 APC EC 2216 A/B (epoxy) 7.6

APC UK 8212/UK 5400 10.4
(polyurethane)
APC AW 116/HV 953U 9.4

POM*/PA6.6** APC* AW 116/HV 953U

APC** (epoxy) 8.2

LDPE*/PA 11** APC* UK 8212/UK 5400

APC** (polyurethane) 8.4

PTFE*/PA 6.6** NSE* UK 8212/UK 5400

APC** (polyurethane) 8.0
EFFECT OF SURFACE TREATMENT ON WETTABILITY 135

CONCLUSION

Air-plasma ~reatment of surfaces is a versatile technique for

improving the wettability and bondability of low surface energy
thermoplastic materials. Five thermoplastic materials (LDPE, HDPE,
POM, PA 6.6 and PA 11) could be treated for fifteen minutes at 50 W
power successfully to render them wettable and bondable. Once the
surface is treated in an activated gas plasma, it should be bonded
within thirty minutes. Epoxy and polyurethane adhesives are
suitable for developing strong adhesive joints (shear strength
higher than 7 MPa) on these five air-plasma treated polymers).

Naphthalene-sodium etching is the only efficient surface

treatment for improving the wettability and bondability of
polytetrafluoroethylene. The polyurethane adhesive UK 8212/UK 5400
is a suitable adhesive for developing high shear strength (higher
than 8 MPa) on thoroughly treated polytetrafluoroethylene to itself
or to polyamide 6.6. No adhesive and (or) surface treatment among
those studied in this work was suitable for developing strong
adhesive bonded joints on polypropylene.

For a better understanding of the inefficiency of air-plasma

cleaning for improving the bondability of polypropylene and
polytetrafluoroethylene, additional work should be done to optimize
the plasma cleaning process. A study of the chemical and physical
effects of plasma treatment of polymer surfaces by modern surface
analysis methods (ECSA or Auger Spectroscopies) will be undertaken
in our laboratories.

REFERENCES

1. R. Houwink and G. Salomon, Editors, "Adhesion and Adhesives",

Vol. 1 and 2, Elsevier, Amsterdam, 1965.
2. D. H. Kaelble, "Physical Chemistry of Adhesion", John Wiley
and Sons, (Interscience), New York, 1971.
3. 1. Skeist, Editor, "Handbook of Adhesives", Von Nostrand,
New York, 1977.
4. A. J. Kinloch, J. Materials Sci., 15, 2141 (1980).
5. H. W. Fox and W. A. Zisman, J. Colloid Sci., 7, 428 (1952).
6. W. A. Zisman, in "Contact Angle, Wettability and Adhesion",
ACS Adv. Chem. Ser., No. 43, p. 1 - American Chemical Society,
Washington, D. C. 1964.
7. L. H. Sharpe and H. Schonhorn, ibid, p. 189.
8. K. L. Mittal, in "Adhesion Science and Technology", L. H. Lee,
Editor, Vol. A, pp. 129-168, Plenum Press, New York, 1975.
9. K. L. Mittal, Polymer Eng. Sci., 17, 467 (1977).
10. J. A. Koutsky, in "Surface Contamination: Genesis, Detection
and Control", K. L. Mittal, Editor, Vol. I, pp. 351-357,
Plenum Press, New York, 1979.
136 J.P. JEANDREAU

11. H. Schonhorn and R. H. Hansen, J. Polymer. Sci., 84, 203

(1966).
12. D. K. Owens and R. C. Wendt, J. Appl. Polymer. Sci., 11, 1741
(1969).
13. J. S. Tira, "Gas Plasma Treatment to Improve the Bondability
fo an RTV Silicone to Foamed Polypropylene", Department of
Energy, Contract Number DE-AC04-76-DP00613, BDPX-613-2224
(Rev), July 1979.
14. N. J. De Lollis, "Activated Gas Treatment of Silicone Sur-
faces", Sandia Laboratories, Albuquerque, New Mexico, report
SAND 77 - 1256C, 1977.
15. D. Briggs, D. G. Rance, C. R. Kendall and A. R. Blythe,
Polymer, 21, 895 (1980).
16. M. R. Wertheimer and H. P. Schreiber, J. Appl. Polymer.
Sci., 26, 2987 (1981).
17. T. Riley, T. C. Mahuson and K. Seibert, "Investigation into
the effect of plasma pretreatment on the adhesion of Parylene
to various substrates", National Aeronautics and Space
Administration, Lewis Research Center, Contract 8110-72-0435,
pp. 95-114, 1981.
18. G. L. Flowers and J. L. Montague, "Evaluation of Cleaning
Procedures and Bonding Materials for TPXMaterials",
Department of Energy, Washington, D. C. MHSMP-8140, Contract
AC04-76DP00487, pp. 1-9, 1981.
19. W. L. Baun, in "Adhesion Measurement of Thin Films, Thick
Films and Bulk Coatings", K. L. Mittal, Editor, pp. 41-53,
American Society for Testing and Materials, Philadelphia,
PA, 1980.
20. K. L. Mittal, Pure Appl. Chem., 52, 1295 (1980).
21. W. L. Baun, in "Adhesion Aspectsof Polymeric Coatings", K.
L. Mittal. Editor. pp. 131-146. Plenum Press •. New York. 1983.
THE EFFECT OF MOISTURE ON THE DIMENSIONAL STABILITY OF ADHESIVELY

BONDED JOINTS

E.J. Hughes, J. Boutilier, and J.L. Rutherford

Kearfott
A Division of the Singer Co.
Little Falls, NJ 07424

Torsion shear creep tests were made on two

bisphenol A-type epoxy adhesives as a function of
temperature, humidity, and applied stress. The
effects of varying the filler content of the two systems
were also determined. The relationship between creep rate
and stress was logarithmic. Although the presence of
moisture had a much greater effect in the unfilled
epoxy systems than in the filled systems, the
relationship between humidity effect and filler
content was not a simple one. The results are discussed
in terms of the residual stresses generated by
temperature changes and by moisture absorption.
The non-monotonic nature of the creep rate-stress
relationship previously shown to exist in a number
of filled epoxy systems is explained in terms of the
combined effects of thermal and moisture-induced stresses.

137
138 E.J. HUGHES ET AL.

INTRODUCTION

The use of adhesive bonding in many critical inertial guid-

ance applications where instrument stability is of paramount
importance has given rise to concerns over the effects of envi-
ronmental factors on the stability of adhesive bonds. Of particu-
lar interest is the generation of stresses in bonded adhesives
as a result of moisture absorptionl - 4 and differences in the
coefficients of thermal expansion of the adhesive and adherends.
One measure of the dimensional stability is the change in creep
behavior which occurs as a result of either moisture absorption
or thermally-induced stresses.

Previous work 5 has shown the effect of thermally-induced

stresses on creep behavior. The work described in this paper
demonstrates the effect of moisture absorption on creep behavior
and an effort is made to describe the creep behavior of bonded
joints in terms of the combined effect of thermally and moisture-
induced residual stresses.

EXPERIMENTAL METHODS AND RESULTS

All creep tests were made using napkin-ring type

torsion-shear specimens. The adherends were made of Ti-6AI-4V
alloy.6,7 Bondline thicknesses of about 0.003 inch were used
for all specimens. The adhesives studied are listed in Table
I, along with their cure cycles. Creep rates refer to the
steady-state portion of the creep curve (strain vs time).

Unfilled Epoxy Systems

The effects of moisture absorption on the creep rates of

two bisphenol-A type epoxy adhesives were measured as a function
of the humidity level of the test environment. The bonded speci-
mens were conditioned for 100 hours at the test humidity level
and temperature prior to making creep tests as a function of
applied stress. Typical results for tests made at 120 0 F are
shown in Figures 1 and 2. At any given applied stress the creep
rate increases with increasing humidity of the test environment.

Filled Epoxy Systems

To improve dimensional stability, most gyro-grade adhesives

usually have fillers added. The Epotek epoxy system normally
contains aluminum oxide as a filler, while the LCA4 system con-
tains calcium carbonate. To determine the effects of these
fillers on the susceptibility of the two systems to moisture
absorption, creep tests were made on bonded epoxy adhesives
which contained different amounts of filler. The results are
summarized in Table II and plotted in Figures 3 and 4. In the
 m
"T1
"T1
m
n
-t
0
"T1
s:0
c;;
-t
C
:Il
Table I - Bispheno1-A-type Epoxy Adhesives. m
0
Cure Filler z
0
Adhesive Hardener Cycle Material Manufacturer
~
m
Epotek H-74 amine 116°F for 50 min 70 wt % A1 20 3 Epoxy z
CJ)
212°F for 20 min 44 llm size Technology, Inc. az
:I>
Epotek H-74 amine 116°F for 50 min Epoxy r
CJ)
35 wt % A1 20 3
212°F for 20 min 44 llm size Technology, Inc.
~
!'!!
Epotek H-74 amine 116°F for 50 min unfilled Epoxy r
=i
212°F for 20 min Technology, Inc. -<
LCA4 amine 212°F for 120 min 70 wt % CaC0 3 Bacon Industries
14 llm siZe

LCA4 amine 212°F for 120 min unfilled Bacon Industries

eN
CD
 ~
o

Table II - Effect of Humidity and Filler Content on the Creep of Epoxy Adhesives.

d x 10- 4
Relative
Filler Content Humidity
~ (I>erc~Ilt by_w~ight) (percent) (sec- 1 • psr1)

Epotek H-74 0 55 8.62

0 70 14.08

Epotek H-74 35 55 4.57

35 70 5.78

Epotek H-74 70 55 3.28

70 70 4.22

LCA4 0 40 2.51
0 60 3.25

LCA4 70 40 2.37 m
L
70 60 7.07
::I:
C
G')
::I:
m
en
~
>
r
EFFECT OF MOISTURE ON DIMENSIONAL STABILITY 141

40%
RH
I--
....
'iii
5000 ~

-a.
~
III

V 60%

-
III
W RH
a: 4000

!V
I-
III
c..
W
w
a:

V
u 3000

I
2000

1000

o
o 100 200 300 400 500 600
CREEP RATE (xl0- S/sec)

Figure 1 - Effect of Humidity on Torsion-Shear Creep of Unfilled

LCA4 at 1200F.

2000

-----
20%
RH

I
r/ 55%
~
70%
RH

11/
RH
.... 1000 -'
'iii
/'
a.
III
III

V
W
a:
I-
III
c..
w
w 0
a:
u o 200 400 600 soo 1000 1200 1400
CREEP RATE (x10,S/sec)
Figure 2 - Effect of Humidity on Torsion-Shear Creep of Unfilled
Epotek H-74 at 120oF.
 142 E.J. HUGHES ET AL.

7000

55% RH
6000
<II
a.
V)
V)
5000
w
cz:
l-
V)

Go 4000 70%
w
w
FILLED
cz:
u
3000

55%RH
2000
UNFILLED
1000

0
10 100 1000
CREEP RATE (xl0· 8/ sec)

Figure 3 - Effect of Humidity on Creep of Epotek H-74 Systems.

5000
'iii
a.
V) 4000
II )
W
cz:
I-
V) 3000 60% RH
Go
w
W
cz:
U 2000

1000

o 10 100 1000
CREEP RATE (l(10·8 / sec)

Figure 4 - Effect of Humidity on Creep of LCA4.

EFFECT OF MOISTURE ON DIMENSIONAL STABILITY 143

Epotek H-74 epoxies, changing the humidity level of the test

environment was seen to have a much greater effect in the un-
filled epoxy than in the two filled systems. The data for the
filled LCA4 at the higher humidity level were anomalous. As
shown in Figure 4, there was a distinct change in slope of the
log (creep rate) versus creep stress curve at the higher stress
levels in three of the four specimens tested. At the lower
stress levels the slopes for these three specimens were similar.

Effects of Temperature

Creep tests were made with a filled LCA-4 adhesive at three

temperatures (90 0 , 130 0 , and 150 0 F) using the same applied
stress (3100 psi) and relative humidity (43 percent). The speci-
mens were conditioned for 100 hours at the test temperature
and humidity level before the creep tests were made. Typical
creep curves are shown in Figure 5. The creep rates (y) obey
an Arrhenius-type relationship as shown in Figure 6:

• K -Q/RT
Y = e (1)

where
Q activation energy for creep

R gas constant

T absolute temperature

K constant

DISCUSSION

Effects of Filler Content on Moisture Absorption

Although the creep properties are dependent on the humidity

of the test environment, the magnitude of the effect is not
directly related to the amount of filler present. For example,
in the Epotek systems the volume of epoxy resin in the bonds
is in the ratio 100:65:30 for the unfilled, 35 percent, and
70 percent filled specimens, respectively. However, at constant
humidity and creep stress level, che changes in creep rate are
measured more in orders of magnitude than in these simple
ratios. From Figures 3 and 4, the creep behavior can be de-
scribed by an equation of the form
.
logY= AT + B (2)
 ~
~

50

150°F
40

I:
<:::.
I:
.- 30 I
'"0.... 130°F
0
Z
;;:
a:
I-
en 20
Q.
w
w
a:
(.)
gO°F

10
r
o m
1000 L
o 2000 3000 4000 5000 6000 7000
I
TIME (SECONDS) C
G)
I
Figure 5 - Temperature Dependence of Creep of LCA4.- Stress m
Ul
3100 psi, Humidity 43% RH. m
-l
»
r
EFFECT OF MOISTURE ON DIMENSIONAL STABILITY 145

100
90

80

70

60

50

--.
U.,
c 40
:.::::.
.,.=
...0b 30
w
l-
e:(
a:
II..
w
w
a:
u
20

10~--~~--~----~----~----~
2.9 3.0 3.1 3.2 3.3 3.4
1000/T
Figure 6 - Logarithm of Creep Rate vs liT for LCA4.
146 E.J. HUGHES ET AL.

where
y creep rate

T creep stress

A,B constants

The slopes of the log (creep rate) vs creep stress plots are
given in Table II. It appears that the presence of a filler
reduces the susceptibility of the epoxy adhesives to effects
due to moisture absorption. This is not surprising since adding
filler to the system reduces the amount of moisture absorbing
epoxy in the adhesive bond. What is somewhat more surprising
is that after an initial relatively large change in the creep
behavior of Epotek H-74 on the addition of 35 percent by weight
of filler, further increases in the filler content had a much
smaller effect on the subsequent creep properties.

Effects of Temperature on Moisture Absorption

The creep stress obeyed an Arrhenius-type relationship

as shown in Figure 6. The activation energy for creep is calcu-
lated to be 7250 cals/mole which is about three times higher
than that for nylon epoxide adhesives measured at the same tem-
perature but at a lower stress level. 8 The high activation
energy for creep is consistent with the high strength, high
modulus properties of filled gyro-grade adhesives.

Combined Effects of Thermal and Moisture-Induced Stress

The generation of residual stresses in adhesively bonded

joints can arise from: differences in thermal expansion coeffi-
cients between the adhesive and adherend~~10 and from moisture
absorption by the adhesive. ll The presence of internal or resi-
dual stresses in any material is always a potential source of
dimensional instability. When external stresses are applied
to the material during normal operation, their effects can be
greatly modified by the presence of existing internal stresses.
Previous work on several filled epoxy adhesive systems 5 has
shown that the relationship between creep stress and creep rate
is not monotonic. The type of relationship observed is shown
schematically in Figure 7. This unusual relationship was inter-
preted in terms of the thermally-induced residual stresses which
exist in bonded joints and which are relieved as a result of
plastic flow during creep -- accounting for the initial decrease
in creep rate with increasing stress. A finite element analysis
of the residual stresses in the bonds 5 showed that although
they were significant, their magnitude was too low to fully
account for the observed behavior. In that treatment, no effort
was made to include the effects of moisture-induced stresses.
EFFECT OF MOISTURE ON DIMENSIONAL STABILITY 147

These latter stresses can be determined in a manner similar

to that of the thermally-induced stresses; provided that the
humidity and temperature of the test environment is known toget-
her with the coefficient of moisture absorption of the epoxy
at the test temperature. The latter quantity, which is equiva-
lent to the coefficient of thermal expansion, is the change
in length per unit length per percent relative humidity of the
test environment. An apparatus has been developed by Singer-
Kearfott 1 which enables this property to be measured as a func-
tion of temperature. Typical values for the LCA4 and Epotek
H-74 systems are given in Table III. To determine whether the
combined effects of thermal and moisture-induced residual
stresses can account for the non-monotonic slope of the creep
relationship shown in Figure 7, a finite element analysis was
performed (using an isoparametric method) to determine the
residual stresses arising from the combined effects of the
two phenomena. The average values of the radial (ox), longitudi-
nal (Oy)' and tangential (oz) components of the residual stress
derived from the analysis are shown in Table IV for filled
Epotek H-74. Included in Table IV are the components of the
residual stress arising from the temperature difference between
the cure and the test temperatures. The stresses were determined
for a typical 0.003 inch bond line specimen at a test tempera-
ture of 120 0 F and relative humidity of 23 percent. The combined
effects of the applied and the internal stresses can be repre-
sented by the octahedral shear stress, T , where
oct
2
Tact = 1/3 (ox-Oy) + (Oy-oz) + (oz-ox) (3)

If the non-monotonic shape of the creep rate curve in Figure

7 is due to residual stresses, Tact calculated at the threshold
stress (TT in Figure 7) should be equal to the value of Toct
calculated from that value of the applied stress above the
knee of the curve which results in the same creep rate as the
threshold stress (TA in Figure 7). At the threshold stress,
both the residual and the applied stresses are operative, whereas
at TA it is assumed the residual stresses have been relieved
by plastic flow and only the applied stress is operative. In
Equation 3, Txy, T ZX ' and the a terms are contributed by the
residual stresses while the Tyz term is the applied stress.
In napkin-ring type specimens Txy and TzX are either zero or
close to zero and can be ignored. By substituting the appropriate
stress values in Equation 3, the octahedral shear stress can
be determined at TT and TA' The results are shown in Table V
where the two values of Tact derived from the analysis are in
reasonable agreement, lending support to the hypothesis that
148 E.J. HUGHES ET AL.

Table III - Values of the Coefficient of Moisture

Absorption for Epotek H-74 and LCA4.

Test
Epoxy Temperature Coefficient of Moisture
Identification (OF) Exp. (in/in/% RH x 10- 6 )

LCA4 (filled) 160 18.22

140 14.11
120 8.55

Epotek H-74 160 51.9

(filled) 140 48.88
120 45.0

Table IV - Average Values of the Residual Stresses in

an Epotek H-74 Adhesive Bond Derived from
the Finite Element Analysis. Humidity
23% RH, Temperature 120o F.
CYz
Adhesive lb/in 2

Combined Thermal 2385 -14 2327

and Moisture-
Induced Stresses

Thermally-Induced 1174 -11 1214

Stress

Table V - Comparison of Octahedral Shear Stresses

at LT and LA in Epotek U-74.

Threshold Applied Octahedral Shear Octahedral Shear

Stress LT Stress LA Stress at LT Stress at LA
(psi) (psi) (psi) (psi)

950 2660 l360 2170

EFFECT OF MOISTURE ON DIMENSIONAL STABILITY 149

'A

~
w
a:
I-
en
a..
w
w
a:
u

CREEP RATE

Figure 7 - Schematic of Creep Rate-Creep Stress Relationship

in Filled Epoxy Adhesive Systems.

the non-monotonic slope of the creep curves is due to residual

stresses. The values of TT and TA shown in Table V were obtained
from Refer~nce 5. In this analysis no consideration was given
to other sources of residual stresses such as shrinkage stresses
during cure. Inclusion of these additional stresses would result
in even closer agreement in the values of ~ct calculated at
TA and ):.

CONCLUSIONS

1. The torsion-shear creep of filled and unfilled epoxy ad-

hesives is increased by the presence of moisture.

2. Unfilled systems have a greater susceptibility to moisture-

induced creep effects than do filled systems.

3. In the filled systems examined, increasing the amount of

filler had a minimal effect on the creep behavior in humid
environments.

4. The activation energy for creep at approximately 3000 psi

stress was comparatively high, probably reflecting a high
degree of cross linking and the innate stability of these
epoxy systems.
150 E.J. HUGHES ET Al.

5. The non-monotonic creep behavior of this type of epoxy

adhesive system can be accounted for by considering the
combined effects of thermally-induced, moisture-induced,
and externally applied stresses.

ACKNOWLEDGEMENTS

The authors would like to express their appreciation to

LaVerne Dunham who performed much of the creep testing described
in this paper. We are also indebted to Joseph Rygelis who per-
formed the finite element analyses. Finally, we would like to
thank Professor Norman Brown, University of Pennsylvania,
Philadelphia, Pennsylvania, for his many helpful ideas and
suggestions during the course of this work.

REFERENCES

1. J. Boutilier, E.J. Hughes, and J.L. Rutherford, in "Proc.

26th National SAMPE Symposium," Los Angeles, 781 (1981).
2. Y. Weitsman, "Residual Stresses in Adhesive Joints," Mater-
ials Lab., Wright Patterson AF Base, Final Rep., AFWAL-
TR-81-4121 (1981).
3. J.P. Sargent and K.H.G. Ashbee, J. Adhesion, 80, 175 (1980).
4. W. Althof, in "Proc. 11th National SAMPE Tech. Coni. ,"
309, Boston 1979.
5. E.J. Hughes and J.L. Rutherford, Materials Sci. Eng., 44,
57 (1980).
6. E.J. Hughes, J.L. Rutherford and F.C. Bossler, Rev. Sci.
Instrum., 39 (5), 666 (1968).
7. F.C. Bossler, M.C. Franzblau and J.L. Rutherford, J. Phys.
E, Ser. 2,' 1, 820 (1968).
8. H.Shen and J.L. Rutherford, Materials Sci. Eng., ~, 323
(1972).
9. H. Dannenburg, Soc. Plast. Eng. J, 21 (7), 669 (1965).
10. W.T. Chen and C.W. Nelson, IBM J. Res. Develop. 23 (2),
179 (1979).
11. S. Gazit, J. Appl. Polymer Sci., 22, 3547 (1978).
THE DIMENSIONAL STABILITY OF EPOXY ADHESIVE JOINTS

J.P.Sargent

H.H.Wills Physics Laboratory

University of Bristol
Royal Fort, Tyndall Avenue
Bristol BS8 lTL, England

An interferometric technique which employs

Moire patterns has been used to study the deformation
that occurs when an adhesive joint undergoes water
uptake. It has been demonstrated that very large
swelling stresses can be generated as a result of the
inhomogeneous swelling. The magnitude and position of
these stresses can be obtained by application of
equations developed from beam and elasticity theory.
A consequence of this analysis is the prediction of
damped normal displacement waves in the cover slip in
regions of negligible water concentration. This is
observed expe~imentally. Measurements are reported on
the irreversibility of dimensional changes when an
adhesive joint is subjected to uptake and subsequent
removal of the water responsible for swelling - on the
assumption that the epoxy based adhesive is free 'from
shear stresses during the cure - and on the influence
of dissolved inorganic salts upon the swelling
behaviour.

151
152 J.P. SARGENT

INTRODUCTION

The swelling inhomogeneity that occurs when a resin adhesive

absorbs water may be conveniently demonstrated by making model
joints consisting of an adhesive layer sandwiched between a rigid
substrate and a flexible microscope cover slip. If such a joint
is used as one of the components of an interferometer, the
resulting interference pattern may be analysed to give information
about the deformation of the cover slip and hence of the
underlying adhesive. The experimental technique, reported by
Sargent and Ashbee l makes use of photographs of the interference
pattern to generate Moire patterns in order to precisely follow
the development of swelling in the adhesive layer.

EXPERIMENTAL METHOD

Figure la is a schematic diagram of an adhesive joint

manufactured from a microscope cover slip as one adherend and a
rigid block of glass as the other. Figure lb shows a schematic
representation of the apparatus and optical path. The adhesive
joint is mounted in close proximity to an optical flat and when
illuminated with monochromatic light a complex pattern of
interference fringes is formed. These fringes precisely reflect
the topography of the cover slip and therefore of the underlying
adhesive. Should the shape of the cover slip change, as happens
when the resin swells during water uptake, the pattern of
interference fringes changes. If images of the interference
pattern photographed during swelling are superimposed onto an
image obtained at the start of the experiment, a Moire pattern is
generated, the development of which faithfully follows any
changes in shape of the cover slip.

19mm
"
:.". tttt tt
'"
.l "., *l50)Jm

If 25mm "

lQI
Figure lao Schematic diagram showing
25mm

the test specimen.

DIMENSIONAL STABILITY OF EPOXY ADHESIVE JOINTS 153

Lamp
Micc"

/ : -_.
! ____rU LfJ
Camera

n_n_E~_~:
~ UV ---~
Hal f s l vPred
m irror

Filter

Optical f1al
Spec i men
chamber

Spec i men

Figure lb. Schematic of the optical path.

Application of the Moire technique to these interference

patterns is valid so long as there is no change in sign of the
gradient of the surface, at the point of interest, between the
initial photograph and any subsequent photographs.

Specimens usually took the form of 19mm diameter circular

glass cover slips bonded to a rigid block of glass, some specimens
however, were manufactured using cover slips which were square in
outline, some in which the glass substrate was replaced by
anodised titanium, and some in which all the· glass components
were replaced by mica.

Figure 2a. Predicted water Figure 2b. Moire contours

concentration contours. of displacement.
154 J.P. SARGENT

RESULTS

1. Swelling Behaviour 1 ,2,3

Figure 2a shows the predicted water concentration contours

obtained by solution of Fick's diffusion laws for a specimen
having a square outline. Figure 2b is a Moire pattern photograph
for an FMI000 specimen immersed in distilled water for 4 hours at
60C. The Moire finges are the circumferential lines running around
the rim of the specimen and which are superimposed upon the
interference pattern. For small times, when both the displacement
and penetration distance of the advancing water front is small,
then the correlation between the experimentally derived
displacements and the water concentration contours is good. Since
the Moire fringes are displacement contours an additional effect
is apparent at longer times of immersion, namely the presence of a
negative displacement ahead of the swelling front. This is seen as
closed Moire loops towards the centre in Figure 2b. This negative
displacement is a consequence of the cover slip compensating for
the swelling around the rim and as a result a complicated normal
stress system is introduced into the adhesive of a magnitude which
eventually leads to fracture of the cover slip.

Equations derived by Love 4 show that the normal displacement

(w) of a thin plate rigidly supported at its edge and subjected to
a normal pressure p is given by p = -Da 4w/ax4, where D is the
flexural rigidity. Using this equation in conjunction with the
equations of linear elasticity and beam theory to describe the
deformation of the cover slip, it is possible to predict the
behaviour of an adhesive joint when it undergoes water uptake. An
important consequence of this analysis 3 was the prediction of
damped waves of normal displacement in the region of negligible
water concentration. The predicted waves are sketched in Figure
3a.

Figure 3b shows the results of a microdensitometer scan

across a radius of the interference pattern for the specimen shown
in Figure 2b. The specimen had undergone 4 hours exposure to
distilled water at 60C. The initial depression and first wave are
clearly resolved. The height of the first peak is approximately
1/26 of the initial depression (V), which is in good agreement
with the theoretical prediction of Figure 3a.

The normal displacement may be differentiated graphically in

order to obtain the normal stress acting across the cover slip. A
more satisfactory method, since it incurs less error upon
differentiation, is to obtain a functional form for the normal
displacement. Analysis is reported in reference 3 which enables
this to be achieved ~nd Figure 4a shows a plot of WIV(x) (the
DIMENSIONAL STABILITY OF EPOXY ADHESIVE JOINTS 155

Figure 3a. The predicted damped normal displacement waves.

50~------------------------,

E
c:

c
'"
~
u
-100
o
0.
II>
"0

'" - 200
!::
~
~
Vl
o 5 10
Distance across a diameter (mm)

Figure 3b. The normal displacement for the specimen from Figure 2b
showing the damped normal displacement waves.
156 J,P, SARGENT

0-5
VI
C \..
0 c

=
'iii .0
VI
4>
VI

~
VI
....4>
0 - _.!-' VI

c: '0
0 E
....
VI 0
c Z
~

I I -0-5
o 2 4
Distance across a diameter (mm)

Figure 4a. The form of WIV(x) together with the normal stress
distribution.

fourth differential) obtained using this technique. Physically the

form of this may be understood by considering the time dependence
of the pressures exerted on the cover slip by the waterfront as it
progresses into the resin. This is shown in Figure 4b.

cover sUp
~wet adhesive
o dry adhesive

Figure 4b. A schematic explanation for the form of the WIV(x)

curve in Figure 4a. In regions 1 and 3 the cover-slip is pulled
down by dry and saturated regions of the resin. This counteracts
the upward stress exerted in region 2 due to the pressure of the
waterfront.
DIMENSIONAL STABILITY OF EPOXY ADHESIVE JOINTS 157

water
removed
, , I

2 ~A

c...
E
~
'"u
Cl.
III
"'0

"0
E
~ O~--------~~M-------------~~~----------~· ~
z
:8 :8'
water
replaced
~--~----~_~,----~
I ----~~~~~----~--~
o 2 7 8 9 13
Time (days)

Figure 5. Repeated cycles of water uptake and expulsion.

2. The Irreversibility of Dimensional ChangeS

The Moire technique may also be used to observe the behaviour

of an adhesive joint when it is subjected to alternate periods of
water uptake and expulsion. Figure 5 is a plot of the percentage
displacement of a point immediately adjacent to the edge of an
adhesive joint, and extended to cover several cycles of water
uptake and expulsion. Creation of Moire patterns from images
photographed at identical amounts of swelling, and which are
separated by one or more water uptake/expulsion cycles permit
immediate detection of any irreversible changes in the adhesive
film dimensions. Figure 6a is a Moire pattern formed between
photographs taken at the points referred to as A and A' in Figure
5 (representing states of maximum water uptake). Figure 6b is
another pattern corresponding to Band B' in Figure 5
(representing states of m1n1mum water uptake). Although the
absence of any circumferential Moire patterns in either of the
Figures 6a or 6b demonstrates that repeated cycling has so far not
led to any overall irreversible changes in the displacement field,
small changes are evident. One such change is indicated by an
arrow in Figure 6b and shows the appearance of a small Moire loop
indicating that a permanent irreversible local swelling
displacement has occurred. Longer immersion times eventually lead
to a complex debonding pattern. However, it was found that
substitution of adhesive FM 1000 in place of FM 73M, led to a well
defined debonding pattern in which the development of the
debonding edge could be carefully monitored.
158 J.P. SARGENT

Figure 6a. Moire image formed between photographs taken at points

indicated as A and A' in Figure 5.

3. The Influence of Dissolved Inorganic Salts

When an adhesive joint manufactured from an anodised

titanium/epoxy adhesive/glass cover slip sandwich is exposed to a
solution that contains sodium or potassium ions, it shows a marked
reduction in the magnitude of the swelling relative to that when
exposed to distilled water 6 • Figure 7 shows the normal

Figure 6b. Moire image formed between photographs taken at points

indicated as Band B' in Figure 5.
DIMENSIONAL STABILITY OF EPOXY ADHESIVE JOINTS 159

displacement across a diameter for three specimens which were

immersed in distilled water and saturated solutions of NaCI and
KCI solution for 71 hours at 8lC. The data presented in Figure 7
corroborates earlier evidence (Ashbee and Wyatt 7 , Farrar and
Ashbee 8 ) that the cured resins behave as semipermeable membranes.
If for each of the immersion media the diffusion species in the
adhesive is pure water, then when it reaches either the oxide
surface, the glass surface, or impurities present in the adhesive,
water solubles will be dissolved with a consequent decrease in
chemical potential. To compensate for this and hence maintain
chemical equilibrium between pockets of aqueous solution and the
water or water solution to which the joint is exposed, the
pressure of the liquid increases, thereby causing the pockets to
expand and contribute to the swelling. When, instead of pure
water, the aqueous environment is a solution that contains sodium
or potassium ions, the difference in chemical potential and hence
the magnitude of the osmotic pressure and the magnitude of the
swelling, are all diminished.

Experiments in which the rigid titanium substrate was

replaced by its glass equivalent showed that immersion of the
adhesive joint in saturated NaCI or KCI solution had a similar
mitigating effect upon the swelling. Examination of the specimen

30r------------------------------------.
FM l()()(), 81·C, 71 Hours

debonding
edge
E
3- .-d ist ill ed
water
c
<1>
20
E
<1>
u
o
Q.
1/1
"0
01
C
:::: 10

Distance across a d ia meter (mm)

Figure 7. Displacement normal to 3 adhesive joints after immersion

in distilled water and saturated solutions of NaCI and KCI
respectively.
160 J.P. SARGENT

Figure 8. Pressure filled cavities observed between crossed polar-

izers for adhesive FM 1000 after immersion in distilled water.

with a transmission polarising optical microscope revealed the

presence of numerous pressure pockets contained within the
adhesive layer 9 , (Figure 8). These showed patterns of stress
birefringence consistent with their having a large internal
pressure. Comparison between two specimens, one of which was
immmersed in distilled water and the other immersed in a saturated
solution of NaCI, showed that there was a reduction in the number
of cavities per unit volume for the NaCI immersed specimen.

4. Curing Stresses 9

Non-uniform changes in glue-line thickness during cure may

also be measured using the Moire technique by monitoring the
normal displacements of the deformation of the cover slip in an
adhesive butt joint as the underlying adhesive cures.

The accelerated cure reactions that take place during the

elevated temperature curing recommended by manufacturers for
proprietary resin-based adhesive films promote cross-linking and
thereby give rise to shrinkage. There are no superimposed
dimensional changes attributable to such processes as chain-
scissioning, as could be the case in polyesters that contain
diffused water, or to the release of volatiles, as might be the
consequence of condensation reactions in polyamides for example.
The shrinkage is assumed to be homogeneous because any tendency to
create shear stress is thought to be relieved by viscoelastic
flow. This assumption is based on the premise that before it gels,
the fluid resin behaves in a Newtonian fashion and, in particular,
is unable to support shear stress.
DIMENSIONAL STABILITY OF EPOXY ADHESIVE JOINTS 161

Figure 9. Moire fringes ( M) generated by superimposition

of interference patterns ( I ) in order to reveal the
deformation of the cover slip after reaching the cure temperature
(l20C). The cover slip was 150 fJll thick.

Each specimen was mounted in a specimen chamber so that the

free surface of the cover slip was in close proximity to an
optical flat. This assembly was then mounted on an optical bench.
In order that a uniform temperature distribution was maintained
across the specimen it was mounted on a thick disc of copper, this
in turn was held in good thermal contact with an aluminium
specimen chamber. Figure 9 shows a sequence of Moire patterns
taken during the cure of a specimen manufactured from Redux 312/5
(manufactured by Ciba-Geigy). Figure 10 shows the normal
displacement at different points across a diameter of the same
specimen after 5 minutes at the cure temperature, where to = 30
minutes is the total time at the cure temperature. The curve fits
the equation
162 J.P. SARGENT

w = .073 (x - .0095)2 (1)

where w is the normal displacement

x is the distance measured from the edge of the specimen

Since the data are described by a parabola, the fourth

differential o4w/ox4 and hence the normal stress azz are
evidently zero. Hence it is concluded that the deformation is
caused by radial stresses transmitted from the resin to the cover
slip and not by stresses created normal to the joint. To check
that curing really does cause the parallel sided disc of resin to
transform into a concave lens shape, a specimen was manufactured
using cover slips for both adherends. As expected, curing caused
this sandwich to deform into a double concave lens as might have
been produced by the application of edge tractions.

Figure 11 shows a sequence of Moire patterns for the specimen

from Figure 9 on cooling from the cure temperature. In addition to
a small increase in the concave deformation during cooling ( when
the joint presumably behaves as an elastic sandwich with
contraction in the resin exceeding that in the glass cover slip),
the Moire patterns reveal a generally complex pattern of
deformation, as shown by small loops and perturbations on the
circumferential fringes. These may be associated with local
inhomogeneities such as air bubbles and the carrier cloth present
in the adhesive.

E 5
~

~
<D

-
0
4
c
(\J

E
(\J
u
0
3
-"0.
VI
-u
2
0"1
C
'-
::l
U

o 5 10 15 19
Distance across diamete r 103x (m)

Figure 10. Normal displacement across one diameter of the cover

slip after 5 minutes at the cure temperature i.e.(To-25mins).
DIMENSIONAL STABILITY OF EPOXY ADHESIVE JOINTS 163

Figure 11. Moire fringes generated by superimposition of

interference patterns in order to reveal the deformation of the
cover slip on cooling down from the cure temperature to room
temperature. The cover slip was 150 ~ thick.

CONCLUSIONS

1. Using an optical interferometric technique it has been

demonstrated that the swelling of a model epoxy adhesive joint
exposed to aqueous environments is strongly inhomogeneous, and
that the stresses generated can be sufficient to fracture one of
the adherends.

2. The application of beam and elasticity theory to the swelling

of the adhesive joint has resulted in the prediction of a negative
displacement of the cover slip and damped normal displacement
waves. Both have been observed experimentally.

3. An adhesive film subjected to alternate periods of immersion

in distilled water and exposure to dry air, both at 62°c undergoes
reversible changes in linear dimensions when the period of water
immersion is less than 1 day, but that irreversible changes occur
after longer times.

4. Dissolved salts in the immersion water lead to reductions in

the swelling displacement.

5. An optical interference method has been used to look for non-

uniform shrinkage during the curing of an epoxy based adhesive
film. Inhomogeneities have been measured, the magnitudes of which
indicate that stresses of the order of 10 5 pa are established and
maintained throughout curing.
164 J.P. SARGENT

ACKNOWLEDGEMENTS

The author is grateful to Dr. K. H. G. Ashbee for his helpful

comments in writing this paper.

REFERENCES

1. J. P. Sargent and K. H. G. Ashbee, J. Adhesion, 11, 175

(1980).
2. J. P. Sargent and K. H. G. Ashbee, Polymer Composites, 1, 93
(1980).
3. D. E. Jesson and J. P. Sargent, J. Adhesion, 14, 119 (1982).
4. A. E. H. Love, "A Treatise on the MathematicalTheory of
Elasticity", Dover Publications, Inc., New York, 1944.
5. J. P. Sargent and K. H. G. Ashbee, Polymer, 23, 327 (1982).
6. J. P. Sargent and K. H. G. Ashbee, J. Phys. ~ 14, 1933 (1981).
7. K. H. G. Ashbee and R. C. Wyatt, Proc. Roy. Soc~ Lon., A312,
553 (1969).
8. N. R. Farrar and K. H. G. Ashbee, J. Phys. D., 11, 1009 (1978).
9. J. P. Sargent and K. H. G. Ashbee, To be published.
10. J. P. Sargent and K. H. G. Ashbee, I&EC Product R&D, 1l, 302
(1982).
STRENGTH CHARACTERISTICS OF MONO AND MULTIPLE-WIRE STEEL TO STEEL

JOINTS BONDED WITH AN EPOXY ADHESIVE

R. W. Hylands
Portadown College of Further Education
Lurgan Road, Portadown
Craigavon BT63 5BL, U.K.
*Formerly of Ulster Polytechnic

Strength characteristics of mono-wire and

multiple-wire specimens were investigated as support
work for a research project concerned with the pre-
stressing (energising) of structural materia1s. 1 ,Z,3,4,5
The mono-wire joints were composed of cylindrical steel
to steel interfaces, the adhesive used being Ara1dite
AY 103 with Hardener HY 991. Compression and tension
joints were concentrically constructed and after curing,
axially loaded to failure. Nine glue line thicknesses
and five bond lengths were selected ranging from 0.001
in to 0.100 in (0.0254 mm to 2.54 mm) and 0.50 in to
2.50 in (12.7 mm to 63.5 mm) respectively. Each test
was performed five times. The strongest joints were
obtained with the thinnest glue lines and increases up
to 0.060 in (1.524 mm) diminished the strength by
approximately 32% for both compression and tension
specimens. For glue line thicknesses from 0.60 in to
0.100 in (1.524 mmto 2.54 mm) the joint strength was
almost constant. Test results displayed a coefficient
of variation for the joint strength of between 6.89%
and 13.41%. A relationship was found among mean break-
ing force, glue line thickness and bond length. The
multiple-wire joints consisted of Bridon prestress
wire manufactured by British Ropes Limited 6 of 0.276
in (7 mm) diameter 'patented' plain cold drawn and
produced to BS 2691 7 • Four of these were placed in-
side a standard square steel tube together with a
central 0.1875 in (4.76 mm) diameter wire which pre-
vent the four larger wires from undue warping or
twisting during use. Five bond lengths were selected
165
166 R.W. HYLANDS

for compression and tension joints ranging from 0.50 to

2.50 in (12.7 mm to 63.5 mm) respectively and after
curing all joints were axially loaded to failure. Each
test was performed five times. Test results displayed
a coefficient of variation for the joint strength of
between 5.54% to 12.47%. In general, compression
specimens were slightly stronger than the corresponding
tension ones for all adhesive tests.

INTRODUCTION

The main research project compared the bonding of the cables

of a poststressed timber beam with a similar unbonded beam. The
cable for the posts tressed timber beams consisted of the multiple-
wire and standard steel tube arrangement shown in Figure 1.

Preparatory to this main area of research a suitable adhesive

was selected and its strength properties determined. The chosen
adhesive required had to possess the following essential character-
istics: ability to bond to a variety of different materials, ad-
hesion, cohesion, low shrinkage during cure, set 100% solid, dur~
ability, resistant to moisture, resistant to dilute acids alkalies
and many solvents, ability to be flexible, low creep, possibility
of room temperature cure, low viscosity, ability to solidify for a
range of glue line thicknesses and an assembly period of about
four hours. The epoxy resin selected for the research project
satisfied all of these characteristics.

The strength properties of this adhesive had to be determined

to establish strength behaviour of joints with varying glue line
thicknesses. This information was necessary as the glue line thick-
ness within the cross-section of the duct is variable. For these
tests mono-wire cylindrical joints were used and the range of glue
line thickness chosen to be compatible with those within the duct.
Tests were also required to be carried out using a multiple-wire
arrangment similar to the one which was used in the posts tressed
beam. Consequently, joints were constructed using steel to steel
interfaces; these results were a useful prelude to the design and
analysis of the posts tressed beam with bonded cables.

For all adhesive tests both tension and compression bonded

joints were used.
STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 167

0.62" (15.75 mm)

~I

I
-+-
I
'i 'i
'"C;
~

'"r-.
0
'""-
'"
N

'"
0

-+-
I
EB I
---+--
I I

Figure 1. Cross-section of multiple-wire arrangement for post-

stressed timber beams.

Mild Steel Specimens used in Mono-Wire Tests

The glued joint which was constructed consisted essentially

of a steel rod placed concentrically inside a hole which was
drilled along the centroidal axis of a circular steel bar. The
steel rod had part of its length geometrically similar to the
0.276 in (7 mm) diameter prestress wire and had a polished bright
surface finish. A hole was drilled and reamed at the centroid of
the bar cross-section along the length of the specimen. These
specimens were used to enable accurate control of the bond length
and glue line thickness of each joint. Nine different hole dia-
meters were selected to give glue line thicknesses ranging from
approximately 0.001 in (0.0254 mm) to 0.100 in (2.54 mm) and five
different bond lengths, viz., 0.5 in (12.7 mm), 1.0 in (25.4 mm),
1.5 in (38.1 mm), 2.0 in (50.8 mm) and 2.5 in (63.5 mm) used to
increase test data. Figure 2 shows the test specimens.
168 R.W. HYLANDS

Design of Jigs fOr Alignment of Mono-Wire Joints

It was necessary to design jigs which would enable both

proper alignment and a concentric joint of uniform thickness to
be maintained. Two types of jigs were designed and manufactured
to accommodate compression and tension specimens. Five complete
sets of jigs for each type of test were made so that the testing
programme period could be shortened. Figure 3 shows typical
alignment jigs for compression and tension test specimens.

NOTF.S

I Rond lengths are 0 . 5 in

(12.7 mm) , 1.0 in (25 . 4 0>:,) ,
1.5 in D8 . I mm) , 2 . 0 in
(50 . 8 ~) and 7 . 5 in
( 63 . 5 mm) .

lJ iame t e r of hole s range from

0 . 277 in (7 . 04 mm) to

rt -.
D -
0 . 377 in (9 . 58 rom) 0 enable
bonded joints to be
~ith glue line thicknesses

\170- ranging from 0 . 001 in to

0 . 100 in (2.54 nun to 0 . 0254 mm).

Ib 8.
I~"
I 1llIIl>"' I
(f) 8----------
-{±)-
+
I
--- --- -

2"
\" (50 . 8 mm) ~ I
-<t- r--------t
(f) 4-
---------
- . _- - --

Figure 2. Compression and tension specimens for adhesive tests.

 Ul
-l
::D
m
Z
G')
-l
( I
(")
.. j i g retaining bolts I
• »
::D
f- - -
l - »
(")
-l
J m
II ::D
- - --- en
:::!
l (")
Ul
o"T1
compression t est specimens (mono - wire) Ul
-l
m
m
r
grub screw used in determining -l
bond lengt h
o
Ul
[ -l
m
- - - m
II- r
C-
' / / / / / / / / / / / // /..l O
--- - - - Z
'////////// / /LLd -l
Ul
- - -
[ J
"
tension test specimens (mono-wire)

Figure 3. Typical alignment jigs for compression and tension

test specimens.

0)
to
170 R.W. HYLANDS

To ensure easy separation of the jigs after the adhesive

had been cured, a film of grease was smeared on the portion of the
test specimens adjacent to the jigs. Great care was taken to en-
sure that no grease came into contact with the surfaces being
bonded or indeed with the adhesive. When assembled, both parts of
the jig and joint specimen were held firmly in position using hex-
agon screws as indicated in Figure 3.
The jigs for the compression joint specimens were made to re-
ceive the complete range of bond lengths, the bond length being
determined in each case by the length of circular steel bar with
the concentric hole along its length. The length of these bars
was cut to agree with the correct length of bond.

For the compression tests a special gripping device was de-

signed and manufactured to fit the top jaws of the Avery testing
machine and also the steel specimen which was to be tes~ed. A
circular steel block was made to receive the lower portion of the
jointed specimen being tested and to maintain an axially applied
load during testing.

The jigs for the tension joint specimens were so designed to

facilitate the assembly of the five different bond lengths. This
was achieved by means of a small grub screw attached to the fe-
male portion of the jig which allowed the joint specimen to be
positioned at the desired bond length.

Steel Specimens used in Multiple-Wire Tests

The purpose of these tests was to examine the strength

characteristics of a group of wires which were bonded to a
standard steel sheath. The glue line thickness varied from 0.001
in (0.0254 mm) to approximately 0.100 in (2.54 mm). For the ten-
sion tests standard steel tubing and prestress wires were used.
Five complete sets of test specimens were designed and manufac-
tured to accommodate the five selected bond lengths. This
allowed twenty-five tension tests to be carried out on these
multiple-wire joint specimens.
For the compression tests the steel sheath was embedded in a
concrete cube surround, the multiple-wire arrangement being then
placed inside the steel sheath and bonded at the required bond
length. It was previously calculated that the joint would fail
before slip would take place between the steel sheath and con-
crete cube surround. Twenty-five concrete cube surrounds with
embedded steel ducts and five sets of multiple-wire specimens
were constructed to facilitate a total of twenty-five compression
tests to be performed. Figure 4 shows the multiple-wire test
specimens.
 en
-l
:0
m
Compress ion joint specimen for multiple-wire tests (steel to steel) Z
Gl
-l
I
()
~ ABC I
»
:0
»
()
-l
m
:0

-=- =- ==----=- =------=-=-=-

- I" -
r ~ ~"- ~ - -
r 3 -" -~ " - Cii
-l
c=5
= en
o
"'T1
CONCRETE LA la I I
lc
_"fl en
. ". -l
m
m
r
-l
o
en
. :1: -l
m
m
r
L
o
Z
-l
ffi c o en

3- 3"
blrtmm - r1(..1-mm

Figure 4. Tension joint specimen for multiple-wire tests (steel to steel).

-..J
172 R.W. HYLANDS

Requirements of Testing Apparatus

The main objects of this part of research was to determine:

(a) Strength characteristics of a selected adhesive for

the mono-wire and multiple-wire tests (compression and
tension), steel to steel.

(b) How the strength of a joint varies with glue line

thickness.

(c) How the strength varies with bond length.

(d) The mode of failure for the adhesive tests.

The testing machine satisfied the following requirements:

(a) Accuracy - The test machine shall be verified according

to the requirements of Grade A, BS 1610, 'Verification
of testing machines', Part I, 'Methods of load
verification requirements for elastic proving devices
and verification of machines for tension and compress-
ion testing,.8

(b) Rate of loading - with the test piece in position and

under load. The rate of increase in load shall be be-
tween 300 and 600 1bf/minute, according to BS 1203 :
1963 Appendix B.9

(c) Type of jaw assembly - The jaws shall grip the test
piece with a wedge action. Each pair of jaws shall be
attached by loose-fitting pin joints which in turn are
fitted by ball-and-socket joint to the straining heads.
In a machine mounted horizontally the dead weight of
the jaw assembly shall be carried by freely moving
cross-members. This shall not affect the freedom of
rotation of the jaws.
Alternatively, one of the pair of jaws shall be
attached rigidly to the tensile machine and the other
jaws shall be suspended from a ball-and-socket joint
allowing sufficient sidesway movement of the jaw to
permit self-alignment of this jaw while the test pieces
are being pulled.

(d) Axial load for compression testing - The machine will

require adaptability for the addition of suitably de-
signed portions of apparatus to facilitate proper com-
pression tests to be executed.
STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 173

(e) Be capable of adjustment to operate with a scale not

exceeding 20 kN.

(f) Be capable of producing a graphical recording of load

against change in length to enable observation of the
mode of failure for a particular test.

The machine used for the adhesive tests was the Avery-
Denison Universal Testing Machine.

The number of tests to be repeated for each joint specimen

was another important consideration. BS 1203 : 1963, Appendix B9
specifies that five tests should be carried out for each specimen
and the mean of the results reported for the failing load. It
was also decided to consider five different bond lengths so that
such things as edge effects, Poisson's effect, etc, could be con-
sidered.

Test Procedure

The following is a generalised procedure for carrying out

adhesive tests.

All steel specimens were thoroughly degreased using carbon

tetrachloride on a fluff-free cloth. A close inspection of the
surfaces to be bonded was carried out to check the presence of
any undesirable foreign particles that would interfere with the
glue line of the joint. The surfaces were also checked to en-
sure that they were free from damage.

The adhesive was prepared in accordance with the manu-

facturers instructions and due regard given to handling pre-
cautions for Araldite epoxy resin materials. lO Araldite AY 103,
a plasticised liquid epoxy resin was mixed with Hardener HY 991,
a light-brown liquid in the recommended mix proportions of 100
parts Araldite to 40 parts Hardener by weight or 100 parts
Araldite to SO parts Hardener by volume. The adhesive was pre-
pared at room temperature, ie 200 C (6S o F) using volumetric pro-
portioning and was thoroughly stirred to give a uniform liquid
of low viscosity. The mixed adhesive was then poured into the
female specimen and the male portion inserted to form the re-
quired joint. The excess glue displaced was removed from the
joint.

The jointed samples were placed in the Standards Laboratory

and left for twenty-four hours to cure. This laboratory had a
controlled room temperature of 20 0 C (+ 10C) ie 6S o F (+ 1.SoF);
a certificate having been issued to this effect by the British
Standards Institution.
174 R.W. HYLANDS

After curing, each joint specimen was carefully assembled

between the jaws of the Avery-Denison Universal Testing Machine.
The specimens had been identified by letter or number and were
tested at approximately ten minute intervals to agree with the
twenty-four hour curing period. An axial load was applied to the
jointed specimen, the rate of increase in load being kept within
the limits of 300 and 600 lbf/minute as stated in BS 1203 :
1963 Appendix B.9 The rate of loading was controlled using the
load pacer on the testing machine. Careful records were kept of
the quality of the joint specimen, mode of failure, etc, for
each test sample.

After testing the specimens were soaked in chloroform and

placed in a fume cupboard for twenty-four hours to break down the
epoxy resin on the specimen. The specimens were then spotlessly
cleaned and stored in methylated spirit preparatory to further
testing.

Mono-wire Tests

The steel specimens to be jointed were designed to give nine

glue line thicknesses: 0.001 in (0.0254 mm), 0.004 in (0.1016 rom),
0.012 in (0.3048 mm), 0.020 in (0.508 mm), 0.035 in (0.889 mm),
0.040 in (1.016 rom), 0.060 in (1.524 mm), 0.080 in (2.032 rom) and
0.100 in (2.54 mm).

Provision was also made to test each glue line thickness over
five different bond lengths, viz; 0.5 in (12.7 mm), 1.0 in (25.4
mm), 1.50 in (38.1 mm), 2.0 in {50.8 rom) and 2.5 in (63.5 mm).
The order of carrying out the tests on the specimens was as
follows: Day No 1. all 0.5 in(12.7 mm) bond lengths (nine tension
and nine compression tests), Day No 2, all 1.0 in (25.4 mm) bond
lengths, etc. This testing procedure was adopted to retain con-
sistency within each group of results.

In the compression tests, the male specimens were designed

to ensure that there would be no buckling of the member when the
joint was being loaded to failure.

The length of specimen between the upper jaws of the testing

machine and the glued joint was calculated to be 1.875 in
(47.63 mm), this value being the greatest possible load antici-
pated, viz 6000 1bf (26.7 kN). In the tension specimen the length
of member was not a ruling factor.

The specimens whose glue line thickness was 0.001 in

(0.0254 mm) provided some difficulty in ensuring that the adhesive
completely covered all parts of the joint. As previously stated
alignment jigs were used to assemble the concentrically jointed
specimens. A film of grease was smeared around the surfaces of
STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 175

the parts of the jointed specimen not being bonded and also on the
adjacent surfaces of the alignment jigs. This facilitated easy
removal of the jointed specimen from the steel jigs after the
twenty-four hour curing period.
The analysis of tests results show an approximate spread of
12% between the lowest and highest readings after five repeated
tests for a particular joint. This spread is satisfactory for
adhesive tests. ll In general, it was found that the tension joint
specimens were slightly weaker than the corresponding compression
joint specimens. It is believed that this is because of the
Poisson's ratio effect when the jointed specimen is under axial
load. The compression male member is obviously subjected to a
small lateral increase in dimension whereas with the tension
arrangement the member is subjected to a small contraction. This
is a possible reason for both types of tests having differing
results. Compression and tension specimens failed with a marked
explosive bang with a well defined reading being recorded on the
testing machine. The results of 450 tests using the mono-wire
arrangement revealed information regarding varying glue line thick-
ness and bond length for both compression and tension test specimens.

Graphs of mean breaking force versus glue line thickness were

plotted for both compression and tension experiments using a mean
of five readings for each point plotted and are as shown in Figures
5 and 6. Additional graphs of mean breaking force versus bond
length were plotted to check if there was a linear relationship be-
tween breaking force and each equal increment of bond length. An
approximate linear relationship was found to exist for test
specimens having a similar glue line thickness. Figures 7 and 8
show graphs of mean breaking force versus bond length with all
results plotted.

These give a pictorial representation of the spread of re-

sults for each bond length. Calculations for the coefficient of
variation indicate a spread of values ranging from 6.89% to 13.41%.
From the graphs of mean breaking force versus glue line
thickness, Figures 5 and 6, it can be seen that the strongest joint
is obtained with a thin glue line. As the glue line becomes
thicker, up to about 0.060 in (1.524 mm), the joint strength
diminishes for both compression and tension tests, and for glue
line thicknesses in the range 0.060 in (1.524 mm) to 0.100 in
(2.54 mm) the joint strength is approximately constant.

A probable reason for the diminution in strength with the

thicker glue line joints is the slight shrinkage of the adhesive
from the interfaces during curing. The shrinkage property for the
adhesive under consideration is approximately i.5%. Self straining
in the adhesive is more likely during shrinkage in a thick glue
176 R.w. HYLANDS

o
oo

--- -- - t-
0.5 i
(12 . 7 mm)
~
• • .-+
-+--.,
~--- I +'t

.-.--- ............
1.0 i n
(25 . 4 mm) t
-
• •
1. 5 1 l .- r-~--!f\
.",

W
(38 . 1 mm) 1 + I
T I

-+-.
~.

I".
2.0 1
(50 . 8 mm
- +-
·1- -.-.~. ~
1'\
I• I
1 I
· .
-- ~ ~f
2.5 i
(63 . 5 mm)
----:---
.-.~ ~\-
+ +
,~ \
•
~-•
\ 4

Figure 5. Mean breaking force versus glue line thickness for mono-
wire tests on compression specimens ,steel to steel.(5 bond lengths)
 STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 177

a
N

.
'\
-'1
z
-'"
w
::!: -\
'\l "K.-.
. ~. · 2 . 5 in
(63 . 5 nun)

'"
U N

'"0
\
:...
<.:>
z.... '''---L.
·
~~
2 . 0 in
~
..: ~
· (50 . 8 run)
'"'"
'"z
..:
'"
:>::
co
-"'" ~--.
· · 1.5 in
(38 . I mm)
\\

~.
'-.....
1. 0 in
~.-. · · (25 . 4 mm)

--- --
......
. ~
0 . 5 in
'-- · ( 12.7 nun)

o
o 0.254 0 . 5080 . 7b2 1.96 1.270 1.524 1.778 2 . 032 2 . 286 2 . 54 mr.!
o 0.010 0 . 0200 . U30 0 . 0400.050 O. O(,O 0 . 0'/00 . 080 0 . 090 0 . 10 in

GLUE LINE THICKNESS

Figure 6. Mean breaking force versus glue line thickness for mono-
wire tests on tension specimens, steel to steel. ( 5 bond lengths)
178 R.W. HYLANDS

SO 0 LeNGTH

12 . 75 25 .4 38 . I 50 . 8 63 . 5 (mm)
0 .5 1. 0 1. 5 2 .0 2 .5 (in)

Z <Xl
.>f.

til
U
'"
0
:..
t.:> ~
....Z
'"<:
'"x
'"
z
<: N
-
'"

~ ~-------r------~--------+---

0 . 020

~ r--------r-------+--------+-------

~ ~------4--------+--------+-------~--

~ ~ ______ ~ ______ ~ ________ ~ ______- L______ ~ ______ ~

Figure 7. Mean breaking force versus bond length for mono-wire

tests on compression specimens, steel to steel.
 ~
STRENGTH CHARACTERISTICS OF STEEL T a STEEL JOINTS 179

o
( in)
o (mm)

BOND LENGTH

Figure 8 . Mean b to bsteel.

tests on t ens ion reaking
specimens,
f orce versus
steel ond length for mono-wire
180 R.w. HYLANDS

line than a thin one in a very thin one it would be negligible.

From test results it would appear that the shrinkage of adhesive
during curing period affects the joint strength to varying degrees
up to a glue line thickness of approximately 0.060 in (1.524 mm);
beyond this joint thickness the adhesive shrinkage does not
significantly diminish the strength of the jointed steel specimen.

From the graphs of mean breaking force versus bond length,

Figures 7 and 8, it is seen that there is an approximate linear
relationship in both compression and tension tests. There is how-
ever a tendency for the joint strength to deviate from this linear
relationship and diminish slightly in strength as the bond length
increases; it is believed that if the bond length range was sub-
stantially extended this linear relationship would not necessarily
prevail.

A comparison of compression and tension test results for

mean breaking force versus glue line thickness is shown in Figure 9
and it may be generally seen that the compression tests are
slightly stronger than the corresponding tension tests. The
following may provide a possible explanation. When the 0.276 in
(7 mm) diameter steel bar is loaded there is a change in its
lateral dimension which would tend to be more significant in the
higher load range and almost negligible in the lower load range.
In both compression and tension tests the shrinkage of the ad-
hesive during curing and Poisson's effect must contribute to the
joint strength of corresponding test specimens. It is reasonable
to argue that the shrinkage is the same for both compression and
tension tests and hence the difference in joint strength in the
corresponding tests is caused by the Poisson's effect alone. The
Poisson's effect in the compression tests would tend to make the
joint stronger because of the additional frictional force de-
veloped by the increased lateral spread of the 0.276 in (7 mm)
diameter bar. In the tension tests the bar would contract a
small, amount laterally and tend to pull away from the joint inter-
face thus reducing the strength of the joint. The graphs in
Figure 9 support this reasoning.

From the results of the test series a relationship was de-

rived between the three variables mean breaking force (Y) kN, glue
line thickness (Xl) mm and bond length (X 2 ) mm, a separate
equation being determined for compression and tension test situ-
ations. Using a computer program obtained from the University of
Kansas,12 a facility exists to express a dependent variable Y in
terms of two independent variables Xl and X2 using a po1ynomina1
equation up to the power of 7. However, a po1ynomina1 equation of
degree 4 gave sufficient accuracy for adhesive test results, viz,
STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 181
N
N

o
N
.
\
..
\\
>-\t.~
\ .
~

-
\~~~~
~'-'t c"
\to I'-'~! 1 t
·
i
.
2 . 5 in
(63 . 5 nun)

:\:r~
~. ~.,,+-. I I Vi I
I
z
-'"
-
I
N

'"
U
It II
'"0
:...
<.:l
Z ~
'~. '----L~. : ·
2 . 0 in
(50 . 8 nun)
....
:..:
< .~
'" "-:::1 ::::-.-t-
--=:::::
X
«l

1 ·
:'\ · 1. 5 in
Z
« co 1 (38 . 1 nun)
'"
\:
~..::---
::::::;:.~
.----.:::::::::: · 1. 0 in
· I (25 .4 ClID)

.~
~.
'~. 0 . 5 in
·
.~

N
~:~.~+
· I (12 . 5 nun)

o
o 0 . 2540 . 5080 . 7621 . 96 1. 270 1. 5241 . 7782 . 032 2 . 28 6 2 . 54 !DID
o 0 . 010 0 . 020 0 . 030 0.040 0 . 050 0 . 060 0 . 070 0 . 080 0 . 090 0 . 10 in

GLUE LINE THICKNESS

Figure 9. Mean breaking force versus glue line thickness for

mono-wire tests on compression and tension specimens, steel to
steel. ( 5 bond lengths)
182 R. W. HYLANDS

Y = Cl + C2Xl + C3X2 + C4X2l + C5Xl X2 + C6X22 + C7X3l + C8Xl2X2

2 3 4 3 3 4 22
+ C9Xl X2 + ClO X2 + CllX l + C12Xl X2 + C13Xl X2 + C14 X2 + C15Xl X2

This program is an outgrowth of research in geological trend

analysis. The method consists of expanding the desired linear
regression into a matrix of normal equations, which is then solved
by inversion, giving coefficients of the regression. A simple
Gaussian elimination is used for inverting the matrix.

The coefficients of the equations for compression and tension

specimens are as follows:

Compression joints (maximum error = 3%)

Cl =-0.8156 C6 = -0.7032 x 10-2 Cll = 0.7864

C2 = -3.5370 C7 -4.1023 C12 -0.2121 x 10-1

C3 = 0.4710 C8 = 0.1010 C13 = -0.7808 x 10-5

-3 C14 = -0.7563 x 10-6
C4 = 6.8191 C9 = 0.3828 x 10

C5 =-0.1873 ClO = 0.1236 x 10-3 C'5 = 0.1673 x 10

-3

Tension joints (maximum error = 6%)

-1 C6 = 0.6253 x 10-3 Cll = 0.6724
Cl = 0.4274 x 10
C2 -2.7993 C7 -3.5703 C12 = -0.1933 x 10-1
C3 = 0.3185 C8 = 0.1060 C13 = 0.7919 x 10-5

C4 = 5.9071 C9 = 0.1014 x 10-2 C14 = 0.3780 x 10-6

-4
C5 = -0.2065 ClO = -0.4092 x 10-4 C15 = -0.7295 x 10

MultiEle-Wire Tests
The steel specimens to be jointed were designed, constructed
and tested for five different bond lengths, viz; 0.5 in (12.7 mm),
1.0 in (25.4 mm), 1.50 in (38.l mm), 2.0 in (50.8 mm) and 2.5 in
(63.5 mm). The glue line thicknesses within the sheath varied
from 0.001 in (0.0254 mm) to approximately 0.100 in (2.54 mm).
To retain consistency within each group, the order of testing was
as follows: Day No 1, all 0.5 in (12.7 mm) bond lengths (five
tension and five compression), Day No 2, all 1.0 in (25.4 mm)
bond lengths, etc.
STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 183

The main problem with multiple-wire tests was to ensure that

all specimens were spotlessly clean particularly when being re-used
for subsequent tests. The top portion of the jointed specimen,
which is common to both tension and compression arrangements, was
steeped in chloroform for twenty-four hours to destroy the ad-
hesive, cleaned with a cloth and then stored in methylated spirit
for a further twenty-four hours, removed and placed in acetone for
half an hour and allowed to dry preparatory to further testing.

The lower portion of the tension jointed specimen was treated

in a similar manner. With the lower portion of the compression
specimen it was slightly more difficult because of the concrete
surround. However, the steel sheath which was embedded within the
concrete was thoroughly cleaned prior to constructing this portion
of the specimen. During curing of the concrete surround the ends
of embedded steel sheath were 'stopped' with plasticene to prevent
unwanted materials from entering the previously cleaned duct.
Prior to constructing the adhesive joint, the plasticene was re-
moved and a fluff-free cloth containing carbon tetrachloride was
dragged through the duct to retain inner cleanliness. This lower
portion of the compression test specimen was not re-useable after
testing.

There was a difference in mode of failure with tension and

compressive specimens. The tension specimens failed with a marked
explosive bang with a well defined reading being recorded by the
testing machine, whereas with the compression specimens there was
little or no noise at the point of failure although the testing
machine recorded a definite failure load.

For all tests the jigs and general arrangement for testing
proved to be satisfactory.

The results of 50 tests using the multiple-wire arrangement

give information with regard to the variation of breaking force
for differing specimen bond lengths for both compression and ten-
sion test specimens.

Graphs of mean breaking force versus bond length were plotted

to check if there was a linear relationship between breaking force
and each equal increment of bond length. An approximate linear
relationship was found to exist for both compression and tension
situations as shown in Figures 10 and 11. Each point on the
graphs was an average of five similar experiments. Additional
graphs of breaking force versus bond length were plotted to give
a pictorial representation of the spread of results for each bond
length. These are shown in Figures 12 and 13.

Calculations for the coefficient of variation indicate a

spread of values ranging from 5.54% to 12.47%
184 R.w. HYLANDS

BOND LENGTH

o 12.7 25.4 38. I 50.8 63.5 (nun)

o 0.5 I I 5 2 2 5 (in)

1\
o

\i
\
co "

z
~
~
\J
t
r.:
u
""
0
""'
'-' 0
Z N
H
:<:
«
'"""
~

z
« ..;t

'"
~
N

co
N
\f

N
C"l

of

Figure 10. Multiple-wire tests (compression specimens) steel to

steel.
STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 185

•
N
0')

00
N .
-.j"
Z N
-"

'"
U
•
'0"
~
0
'-'
Z
N

H
:.:
.,,;
.:L:

'"
<"'l
Z
.,,; -0

'":;:: •

00
.

o
o 12.7 25.4 38. I 50.8 63.5 (rum)
o 0.5 1.5 2 2.5 ( in)

BON]) LENGTH

Figure 11. Multiple-wire tests (tension specimens) steel to steel.

186 R.w. HYLANDS

BOND J.ENGTH

u 12 . 7 25 .4 38 . 1 50.8 63.5 (mm)

o 0 .5 1. 0 1.5 2 .0 2.5 (in)

~ ~----+--

:z.
'"
.><

,.;
u
"",..0
0
<:>
:z.
....
:.:
<
,.;

""
'"
:z. .....
<
~
N

co

~:~----+----~------+--------4---

~ L- _ _ _~______-L_______L-______~______________~

Figure 12. Multiple-wire tests (compression specimens) steel to steel.

STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 187

~ .-------r-------~------~-------r-------r------~

~ ~------~------+-------4-------~----

N
~ ~------+-------+-------~-------r

~ ~------r-------+-------4

Z
...
N

""
'"
U
::.::
0
:...
0
l) N

....0.::Z
..:
'"::.::
'"z
..:
'"::;:

~ L- ______ ~ ______J -_ __ _ _ _ ~ _ _ _ _ _ _-L______-L______~

o 12 . 7 25 . 4 38 . I 50 . 8 63. 5 ( mm )
o 0.5 I 1. 5 2 2. 5 (in)
BOND LENGTH
Figure 13. Multiple-wire tests (tension specimens) steel to steel.
188 R.W. HYLANDS

From the graphs of mean breaking force versus bond length,

Figure 14, it can be seen that the compression specimens are
slightly stronger than the tension specimens particularly in the
higher load range. This is probably caused by the very small
increase in lateral dimension of the multiple-wire group which
gives rise to a small additional force which thus increases the
breaking load. The glue line thickness in these tests varies
from 0.001 in (0.0254 rom) to about 0.100 in (2.54 rom) within the
cross-section of the multiple-wire arrangement. The mu1tip1e-
wire specimens generally were found to fail between the outside
periphery of the wire group and the inside surface of the steel
sheath; it is therefore reasonably simple to evaluate the
approximate surface area of the glue line at failure. Consequently,
if we consider the mean shear bond stress of the adhesive for glue
line thicknesses> 0.060 in (1.524 rom) from the mono-wire test re-
sults and compare with the mean shear bond stress at failure of the
multiple-wire tests, it is found that the values obtained are in
reasonable agreement.

Consider the following examples:

Compression specimens (mono-wire) 2.5 in (63.5 rom) bond

length, glue line thicknesses> 0.060 in (1.524 rom)
see Figure 5.
mean breaking force
Mean shear stress of adhesive shear area
1.29 x 2240
7(x 0.276 x 2.5

= 1333 lbf/in2 (9.22 N/rom2 )

Compression specimens (multiple-wire) steel to steel 2.5 in

(63.5 rom) bond length see Figure 10.
mean breaking force
Mean shear stress of adhesive =
approx. shear area
3.55 x 2240
2.4984 x 2.5
= 1273 1bf/in2 (8.78 N/rom 2 )

Tension specimens (mono-wire) 2.5 in (63.5 rom) bond length,

glue line thicknesses> 0.060 in (1.524 mm)see Figure 6.
1.25 x 2240
Mean shear stress of adhesive
71 x 0.276 x 2.5
= 1292 lbf/in 2 (8.91 N/mm2)
STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 189

'"""
If
<Xl
•
'" o-T

z
.;.: '"""" C- 0

0
r.:l
U

'" .
I

1
~

0
0 f/
Z
H
'"
~
..:
0'1

'"
«l
Z
..:
-~
'<"' ~ :

<Xl
•

o
o 12.7 2:5.4 38. I 50.8 63.5 (rnm)
o 0.5 1.5 2 2.5 (in)

BOND LENGTH

Figure 14,. Multiple-wire tests , steel to steel (compression

and tension tests).
190 R.W. HYLANDS

Tension specimens (multiple-wires) steel to steel 2.5 in

(63.5 mm) bond length see Figure 11.
3.42 x 2240
Mean shear stress of adhesive = 2.4984 x 2.5

As a result of this comparison one may suggest that it would

be feasible to use the results of mono-wire tests for glue line
thicknesses beyond say 0.060 in (1.524 mm) to determine a mean
shear bond stress which could be used in a situation involving a
bonded structural element whose glue line thickness is variable.

The foregoing comparisons indicate a reasonable degree of

consistency with the general adhesive testing.

Figure 15 shows a typical breaking line which occurred with

compression and tension specimens at failure.

CONCLUSIONS
The following conclusions are indicated from results of tests
employing an epoxy resin adhesive (Araldite AY 103 with Hardener
HY 991).

(i) The strongest joints are obtained using thin glue lines.

(ii) The strength of a bonded joint diminishes as the glue

line increases in thickness up to about 0.060 in
(1.524 mm); the joint strength being approximately con-
stant for thicker glue lines in the range 0.060 in
(1.524 mm) to 0.100 in (2.540 mm).

(iii) Jointed specimens subjected to compression loads tend

to be slightly stronger than corresponding joints sub-
jected to tension, due to Poisson's ratio effect when
load is applied to the joint.

(iv) A relationship does exist between mean breaking force,

glue line thickness and bond length for the mono-wire
joints; a separate equation being necessary for com-
pression and tension joints.

(v) The strengths of mono-wire bonded joints for glue line

thicknesses beyond 0.060 in (1.524 mm) were in reason-
able agreement with the strengths of multiple-wire
joints.
 STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 191

..,..... e
OJ
....
e
<'-!
...." OJ

....'" ....
.....'" 0)
.
0
Q)

OM
a
t)
Q)
p..
0)

Q)

OM
'"'
?
Q)
.-t
p..
OM
+-I
.-t
~
0
'"'
4-1
e
~

Q)
e ~ 0
OM
U'"I
c: .....
"' .-t
M
'" "' OM
0
.;.l
"'..... '"
N
Q)
~
:;) <:>
'"'
,.0

.-t
'"
t)
OM
p..
>.
H

.....
U")

-I I Q)

'::l"'
M
OM

I • (WW o;C 0; 1 ) " Z9 · 0 -I ~

(UDIl 0;0 · 6 1) " o;C 0

192 R.W. HYLANDS

RECOMMENDATIONS FOR FURTHER RESEARCH

The range of glue line thicknesses in this research project

has been confined to 0.001 in (0.0254 mm) to 0.100 in (2.54 mm).
From test results associated with the mono-wire arrangement it
would appear that the ultimate strength of a bonded joint is
approximately constant for glue line thicknesses between 0.060 in
(1.524 mm) and 0.100 in (2.54 mm); however, it would be a useful
exercise to extend this range to check if this constant relation-
ship still exists for thicknesses beyond 0.100 in (2.54 mm).

There is also lack of information regarding the behaviour of

joints having a wide variety of bond lengths and consequently
more work requires to be done to obtain strength characteristics
of the adhesive when subjected to an extended range of bond
lengths. For example, results of tests for bond lengths varying
from 100 mm to say 1 m in 100 mm increments of bond length would
be invaluable. From these results graphs may be plotted and an
appropriate law developed related to strength and bond length
which could be used by designers whose work involves a detailed
knowledge of epoxy resin adhesives.

The information derived relating to mono-wire bonded joint

situations was concerned with concentrically formed joints giving
a uniform glue line thickness throughout the joint. Test results
data related to a similar joint having variable glue line thick-
ness within the mono-wire arrangement would prove beneficial.
Graphs may be plotted to show how the adhesive strength varies
with glue line thickness under these eccentric conditions using
a single parameter; these results could be compared in detail
with test results of the multiple-wire joints.

A detailed study as to the surface preparation and finish of

the adherend would be useful. For example, polished bright sur-
faces compared with sand blasted chemically etched and other sur-
face finishes could be compared from the point of view of
strength variation. To ensure consistency it is suggested that
the tests be carried out using the mono-wire arrangement and a
glue line thickness kept within the constant portion of the
graphs of Figures 5 and 6; the only variable would then be sur-
face preparation and finish. From the results of this work the
designer could be given positive direction as to the choice and
standard of finish for a particular practical situation.

All of the tests reported are associated with one epoxy

resin adhesive. The author believes that similar tests should be
carried out on a number of different epoxy resins and a com-
parison made of their structural inter-relationships.
STRENGTH CHARACTERISTICS OF STEEL TO STEEL JOINTS 193

REFERENCES
1. E. H. Sidwell, Civil Eng. 64, No. 751, 120 (Feb. 1969)
2. S. R. Port and E. H. Sidwell, Civil Eng. 64, No. 752, 226
(March 1969)
3. E. H. Sidwell and B. C. Carroll, Civil Eng. 66, Nos 774, 775,
47 (Jan/Feb. 1971).
4. E. H. Sidwell and F. F. Keatley, civil Eng. Oct. 1974, p.28;
Jan. 1975,p.50.
5. R. W. Hylands, "The bonding of the cables of a poststressed
timber beam and a comparison of its performance with a similar
unbonded beam". PhD Thesis Council for National Academic
Awards, June 1978.
6. British Ropes Limited, Doncaster, England. "Bridon wire and
strand for prestressed concrete (1962)"
7. BS 2691 (1969) "Specification for steel wire for prestressed
concrete". British Standards Institution.
8. BS 1610 Grade A "Verification of testing machines, Part 1.
Methods of load verification of machines for tension and com-
pression testing". British Standards Institution.
9. BS 1203 (1963) "Synthetic resin adhesives for plywood".
British Standards Institution.
10. CIBA-GEIGY (UK) Limited, Plastics Division, Duxford, Cambridge.
"Handling precautions for aralidite epoxy resin materials".
July 1974 Sheet No. M37c.
11. R. Houwink and G. Salomon, "Adhesion and Adhesives" Vol 2 -
Applications. Elsevier Publishing Company. Amsterdam 1967.•
12. The University of Kansas State Geological Survey. "Computer
contributions No. 28 KWIKR8 - A FORTRAN IV program for
multiple regression and geological trend analysis".
A STUDY OF ADHESIVE JOINTS BETWEEN ALIPHATIC

POLYAMIDES AND METALS

S. S. Pesetskii, V. E. Starzynskii,
and S. V. Shcherbakov
Institute of Mechanics of Metal-Polymer
Systems
BSSR Academy of Sciences, Gomel, USSR

Injection-moulded adhesive joints of aliphatic

polyamides and metals were studied and showed
that intermediate thin adhesive films applied from
polymer or oligomer solutions onto metals is an
efficient means of monitoring the bond strength .
Contact events in a metal/thin polymeric film/poly-
amide system have been investigated and the
results are described here. The thickness of
intermediate films whose optimum values range
from a few tenths to 1 - 2 fm were found to be
a most important parameter that determines the
adhesive strength. The bond strength improvement
is most obvious during a thermal contact between
the intermediate film and metal in air. Adhesive
bonds are formed between the metal and the film
which are resistant to dynamic stresses and
liquids. The bond between polyamide and the inter-
mediate film is usually formed owing to molecular
interactions of polar groups in macromolecules at
the interface. The joints were found to fail cohesi-
vely through the intermediate film. The improved
adhesion to metals and cohesive strength of thin
intermediate films are believed to result from the
fact that thermo-oxidative degradation is inhibited
in these during the thermal contact with metal.

195
196 S.S. PESETSKII ET Al.

INTRODUCTION

Aliphatic polyamides contain adhesively active amide

groups, yet their adhesion to metals is not strong enough.
Moreover, the bond is unstable, it may change spontaneously
during the storage of joint structures, or under alternate
exposure to liquids and atmospheric factors, or in the course
of their thermal treatment in air, etc. 1-5. The adhesive
strength of joints is also dependent on the way they are
formed. If the process of joint making allows only a moment-
ary contact between the molten polyamide and a metallic
surface, the joints have low adhesive strength. This is parti-
cularly characteristic of the injection-moulded adhesive joints6 .
'1'0 improve their adhesive strength, it is advisable to intro-
duce intermediate adhesive layers (IL) such as thin poly-
meric films that can be apRlied onto metallic surfaces prior
to coating with polyamide 7-10. However, the properties of
thin ILs and characteristics of adhesive contacts with IL s
have not been studied in detail.

This paper describes the results of studies of adhesive

joints between aliphatic polyamides and metals which contain
ILs applied from polymeric solutions.

EXPERIMENTAL

Adhesive joints of polyamides P6 (poly- c,-caprolactam),

P66 (poly(hexamethylene adipamide», P610 (poly(hexamethy-
lene sebacamide» with steel and aluminum have been studied.
Two types of specimens were used, namely, unilateral and
bilateral adhesive joints. The first type was represented by
polyamide coatings (120mm x 12mm x lmm in size) on alumi-
num foil (A99 grade, 100 flm in thickness). The second type
consisted of two steel rods (Steel 45,IC 0.45%) and an ad-
hesive polyamide layer (2 mm thick) to form a butt joint. The
specimens were injection-moulded following the procedure
published earlier 11.

For ILs, polyamide P548 (c. -caprolactam (44% )-hexa-

methylene adi pamide ( 37%) -hexamethylene se bacamide ( 19%)
copolymer) or polyvinyl butyral (PVB) were dissolved in
organic solvents and applied onto the metals from solutions.
The succession of the steps was as follows. The surface
of the metallic substrate was placed horizontally and the
solution applied onto it. Then the solvent was evaporated at
30 - 100 0 C. The ILs thickness was calculated as well as
determined by interferometry7.
JOINTS BETWEEN ALIPHATIC POLYAMIDES AND METALS 197

The adhesive strength was determined by the normal

separation of rods and peeling-off the foil from the coating 11 •
The atmospheric resistance of the joints was tested using an
accelerated method 10. According to this method, changes in
the adhesive strength with time, of exposure to atmospheric
conditions have been determined. One cycle of the accelera-
ted test consists of the exposure of the specimens to 100%
RH at 55 0 C for 3 hours along with IR(O.S kWt) and UV
(1.0 kWt) -irradiation with subsequent exposure to the tempe-
ratures -20 up to -25 0 C for 3 hours. The test cycle was
repeated several times. The adhesive bond resistance to
dynamic loading waS determined from the breaking contact
load as measured for polyamide-coated steel rollers paired
with uncoated ones. The tests were carried out at the speed
of 7 rev.s- 1 , the load was increased stepwise by 10 kN/m
after every 2.5 . 104 cycles.

IR-spectral analysis was performed on UR-20 spectro-

photometer fitted with an attachment for a multiple frustrated
total reflection (MF'l'R) (MF'l'R unit, KRS-5, number of ref-
lections, 14; angle of incidence, 45 0 ). The kinetics of degra-
dation was studied by thermogravimetry (Derivatograph OD-
102) using a known procedure 12 based on the analysis
of the thermograms recorded at various heating rates.

Effect of Intermediate Layer Thickness on

Adhesive Strength

The IL thickness ( S) was found to be a most

important parameter that controls the adhesive strength. The
adhesive strength \ 6') is described as a function of §" and
it is obvious that a maximum in ()' is reached for 8'"
values ranging from a few tenths to 1-2 ..Lim ( Figure 1
a and b). /

The maximum adhesive strength is 2 - 3 times higher

in the presence of IL than without it. Normal separation
tests showed that it reaches the value of the tensile strength
(Figure 1 a). Joints with thin ILs appeared resistant
towards the operating conditions, namely, atmospheric factors
( Figure 2 ), liquids 6 , and dynamic contact loadings
(dynamic tests showed that the plastic deformation of the
polymeric coating, and not the breaking of adhesive bonds,
cause the failure of the adhesive joints ( Table 1 ).

The adhesive strength of joints with ILs of optimum

thickness is in fact not influenced by the dissolving
solvent ( Table II) or the solution concentration.
198 S.S. PESETSKII ET AL.

G , MPa A, k 1m

40
/~ 2.4

20 a

0 .15 0 .4 5 1.2 2 . 4
'-0-0
o '--........_--'-____---'-_-"'-_.....,'--"._
10 30
1.6

o
0.4 1.2 2 6 10 20 30
g ,jLm $'jJ- m

Figure 1. Variation of adhesive strength in (a) P6/ Steel

and (b) P6/Aluminum joints with thickness of intermediate
layers (a, b - 1, P548) and (b - 2, PVB).

Table I. Load Bearing Capacity of Polyamide Coatings

(2 mm thick) Bonded to Steel 45 Rollers.
Coat Mode of Coat Breaking Mode of
Material Bonding to Contact Stress, Coat
Metal kN/m Failure

P6,P66,
P610 Without IL 30-40 Separation
from metal

P6, P66 With IL of Plastic

P610 P548
S = 1 rm 130-140 deformation
JOINTS BETWEEN ALIPHATIC POLYAMIDES AND METALS 199

% 1.0 .« - 1

° 0
0 .6
~
• •
0 .2 2

0
50 150 500 600 't , h

Figure 2. Variation of relative adhesive strength in P6/ Steel

joints (1) with P548 layer and (2) without, as a function
of time of exposure. (00) initial adhesive strength;
( 6') adhesive strength after atmospheric action.

Table II. Effect of Dissolving Solvent Used on Adhesive

Strength of P6/Steel Joints (IL of P548, a
= 0.8)A--m;
conditions of ILls thermal contact with substrate :
temperature, 320 0 C; time, 0.9 ks ).

Solvent Solubility Parameter Adhesive

( J/m 3 ) 0.5 • 10-3 Strength,
MPa

Aniline 22.44

N-Propanol 24.28 57±2.7

Ethanol 26.11

O-Cresol 27.13

Phenol 29.58

Resorcinol 32.44 58±3.2

200 S.S. PESETSKII ET AL.

(5' , MPa
60 Ii 8
•
,Q
•
40
.- 2
0 - 1

20
0
1 50 300 600 750 t, days

Figure 3. Effect of steel storage period ( 1) without IL

and (2) with P548 layer on the adhesive strength of
joints with P6.

Ultimate strengths similar to that for polyamides were achie-

ved for substrates with different roughnesses. Therefore, it
is believed that the adhesive properties of ILs are not
affected by the time of film formation from a solution.

It is obvious from Figure 3 that the storage of metallic

substrates in air at about 25 0 C and 70% RH, i. e. labora-
tory ambient conditions, for over 2 years did not affect the
adhesive strength significantly. It is clear, therefore, that ILlS
adhesion to metals cannot result from specific interactions
between ILs and the newly formed metallic surfaces.
Atmospheric factors, temperatures, and periods of IL
formation on metallic surfaces are most important for adhesive
strengths.The improved adhesion resulting from thin IL is most
vividly manifested during thermal treatment,i.e. thermal contact
('1'C) between IL and metal in air. When an inert gas was
used instead of air, the improvement was not so obvious
(Table III).. Hence, the improved adhesiveness of thin poly-
meric films may be expected to result from their kinetic vari-
ations and,probably, a different mechanism of thermal oxidation.

'1'0 verify this idea, the effect of time-temperature

conditions of thermal contact on the adhesive strength was
studied between IL and metal in air, as well as the kinetics
of thermal oxidative degradation of ILs of different thickness-
es on substrates.
 L
0
Z
-I
en
OJ
m
-I
Table III. Effect of 'l'hermal Treatment Conditions at 280 0 C on Adhesive Strength =E
of P6/ Aluminum Joints. m
m
Z
IL Material Thermal IL's Adhesive Strength in kN/m Mode of Joint »
Contact ( 'l'C ) Thickness, Against TC Period in ks Failure C
-a
Environment I
,f'm
!:;
0.6 1.2 1.8 fi
-a
0

Air 1 2.20 1.94 1.86 Failed cohesively

»~
s:
30 0.25 0.25 0.30 through ILs 0
m
en
P548 »z
0
Argon 1 1.10 1.10 1.00 ILs separated s:
30 1.60 1.50 1.50 from metal m
~
r-
en

Air 1 1.80 2.25 2.40 Failed cohesively

30 0.90 0.80 0.75 through ILs
PVB
Argon 1 1.30 1.30 1.25 ILs separated
30 1.40 1.35 1.35 from polymer

N
0
202 S.S. PESETSKII ET AL.

Properties of ILs and Metal as Influenced

by the Conditions of Their Thermal Treatment

The mode of kinetic dependence of the adhesive

strength on the thermal contact (TC) temperature changes
with IL thickness (Figures 4, 5). The adhesive strength of
joints formed with thick ILs ( (5" = 30 flm) either decreases
(from 1.7 - 1.3 kN/m up to 0.4 - 0.1 kN/m, Figure 4 b )
on increasing the TC time (rC) or depends, in a comp-
licated way, upon CC remaining in general low (0.4 - 0.9
kN/m, Figure 4 d). With thin ILs ( § = 1 p.m) the variation
of the adhesive strength depends on the temperature
( Figure 4 a, c ).

The mode of joint failure is also influenced by TC

conditions. At the initial stages of TC (CL. L. 0.6 kS) with
ILs of (5 = 30 JAm, as well as over the range of minimum
adhesive strengths with ILs of g- = 1 jim (Figure 4 a, c),
the joints failed ( as MFTR indicated ) at the IL/metal
interface. In other cases (irrespective of 0-) a cohesive
failure in ILlS material was observed (cohesive mode of
ILls failure was observed visually and recorded by MFTR
technique ( Figure 5 ).

A, kN/m A, kN/m

2.0
1.5
1. 5 1.0

1.0 a,S

0 1.2 2.4 3,6 72 lQa 0 1;2 2:4 3.6 7.2 lGa

1", ks
A, kN/m A, kN/m A, kN/m

2.4 4 2,0 d
3 1.6 1 2 3
1.6
2 1.2
o.a 1 o.a
0,4
0 1.2 315 7.2 14:4 0 1.2 3,6 7.2 14.4
T, ks
Figure 4. Variation of adhesive strength with period of TC
in air between P548 (a, b) and PVB (c, d) on aluminum
and P6 at: (1) 240, (2) 280, (3) 340 0 C. ILlS thickness:
a, c - 11m; b, d 30 j"'m.
JOINTS BETWEEN ALIPHATIC POLYAMIDES AND METALS 203

s:::
6
o

0.

o
OJ
.0

1400 1600 1800 900 1100 1300 V, - 1

cm

Figure 5. IR-spectra obtained by MFTR for aluminum foil-

substrate (1, 2), P548 (3) and PVB (4) films 3D/-, m thick
along with spectra of aluminum foil after P6/aluminum joint
had failed (5, 6) that incorporated ILs ( 0' = lJ-tm) of
P548 (5) or PVB (6); 'l'C conditions between ILs and
substrate: temperature, 240oC; time, 3.6 ks.

Therefore, the differences in the adhesive strengths

of joints with thin and thick films result from ILls thickness
effect on the cohesive strength of ILls material : cohesive
strengths of thin films formed in contact with metals are
much higher than those of thick films.

High values of adhesive strengths with thin ILs

can be achieved under severe conditions of thermal
contact when the temperature is 100 - 160 0 C higher
than the temperature of the viscous flow of ILts material,
and the time of thermal contact was several hours
( Figure 4 a,c, curves 2 and 3 ). Under such conditions
blocks and bulky polymeric films undergo severe degra-
dation. Hence, the resistance of thin films to thermal
oxidative degradation should increase.
204 S.S. PESETSKII ET AL.

6M, %
7

80 6
5
60
4
40
3

20 2
1

0 .7 1 .0 4 8 20 40 c5 , jJ- m
Figure 6. Weight loss with thickness of P548 films on
aluminum: heating rate 3.07 0 Cjmin; temperature of film
degradation: (1) 150, (2) 200, (3) 250, (4) 300,
(5) 350, (6) 400, (7) 450 0 C .

The kinetic curves in Figure 6 plotted based on

thermogravimetric data and describing the degradation of
P548 films indicate that volatile products can be easily
removed from films of 2 - 3 ~m thick, particularly at
an early stage of degradation (curves 1 - 3). The differ-
ences are not significant in the kinetics of weight loss for
films of 20 - 30/,m and thicker.

Calculations showed 6 the effective activation energy

E* of the thermal oxidative degradation of P548 thin films
to be higher than that of the films 30)1m thick during
the process of degradation ( Table IV ). With low percent-
age of conversion ( 20 - 40 %) the difference in values
reached about 50 %.

Similar results were obtained with PVB films. For the

conversion range of 10 to 60%, E* for films of ~ = l;:m
exceeded 1.3 times E* for films of S = 30)£ m (Figure 7).
JOINTS BETWEEN ALIPHATIC POLYAMIDES AND METALS 205

Table N. Effective Activation Energy of P548 films

on Alur;ninum Thermal Oxidative Degradation.

Film E* in kJ/mole depending on conversion, %

thickness,
/,m
20 30 40 80 90

1 116.9 116.9 116.9 156.3 175.1

30 76.2 76.2 76.2 149.0 141.2

E* , kJ/mole

125 . 7 • •
• •
0
83 . 8

. - 1J.lm
o - 30jl. m

41.9
20 40 60 80 cl , o/a

Figure 7. Variations in the effective activation energy during

thermal oxidative degradation of PVB films ( 1 and 30/, m
in thickness ) on aluminum. cA , conversion percentage
(weight of the decomposed polymer in % with respect to
the initial material).
206 S.S. PESETSKII ET AL.

AM,%

70

50

30

10

o
Figure 8. 'l'G-curves of P548 films (1, 2 - 1.,/l m thick) ,
(3, 4 - 30/lm thick) on aluminum. (1, 3) original films;
(2, 4) films after thermal treatment on aluminum at 340 0 C
for 10.8 ks. Heating rate 3.o7 o C/min.

D/Dl128 0 ...... -1
....... 1650 - 1670 cm

4
2
.-
0- 1 ~m
30 ~m

.... -:- a,
A 0
•
1,2 °.......
. . . 3400 cm - 1 , V NH
0.8

0 .4
o
12 24 72 10. 8 t. ks

Figure 9. Variation of relative optical density of absorption

bands for P548 films after their thermal treatment on alumi-
num foil at 240 o C. D/Dl128 is the ratio of optical density
(D) of the absorption band under discussion to the optical
density of the absorption band at 1128 cm- 1 .
JOINTS BETWEEN ALIPHATIC POLYAMIDES AND METALS 207

Keeping in mind the direct correlation between E*

measured during dynamic heating and the thermal stability
of polymers under isothermal conditions, one may expect the
thin films to increase their resistance to the thermal oxida-
tive degradation under the thermal conditions as well. This
expectation is also supported by TGA and IR-spectroscopy
(Figures 8 and 9).

Figure 8 shows that thick films undergo more severe

changes than thin ones in the course of thermal contact
with metals. A film of 30 pm thick after thermal treatment
at 340 0 C for 10.8 ks gave a TG-curve that describes
a typical single-stage process of decomposition (Figure 8,
curve 4) which differs radically from the curve for the
original film (Figure 8, curve 3). TG-curves for films of
0- = 1?m are almost the same qualitatively before the
thermal contact and after that (Figure 8, curveS 1 and 2).

Figure 9 describes the variations in IR-spectra of

P548 films taken with MF'l'R in the course of TC with
aluminum at 240oC. It follows, therefore, that with thin films
the band related to CONH group concentration (amide I,
V NH) retains, or is quite intense after a long TC with
the metal (CC = 3.6 up to 18.0 kS). Hence, the amide
groups within thin films fail more easily than those within
thick films under the conditions of the thermal oxidative
degradation.

One can conclude from the above that the inhibition of

thermal oxidative degradation of macromolecules in thin ILs
during thermal contact incceases their cohesive strength.
The easy removal of low molecular weight products of
degradation from thin films after their TC with metal also
increases the cohesive strength and adhesion of these to
the metal 15

On the Formation of Contacts

between Metal, IL and Polyamide

Qualitative information on the type of bonds formed

duri~g polymer/metal contact may be obtained by
studying the resistance of joints to liquids. After TC with
the substrate in air, thin films would not be removed
from the surface with a solvent which may be associated
with polymer cross linking and their chemisorption on the
metal. That the adhesive IL/metal bond was not completely
broken after a prolong exposure (dozens hours) in boiling
water also indicates a probable chemisorption interaction.
208 S.S. PESETSKII ET AL.

280 300 320 T,oC

A, k Tim
2 .3

1.5

0 .7

14. 4 28 . 8 43 . 2 rH
2 O, ks

Figure 10. Variation of adhesive strength in P6/ aluminum

joints containing ILs of P548 ( 0- = 1 pm) with time of
their exposure to water (CC H20) at 100 0 C. Conditions of
IL's thermal contact with substrate when joints were formed:
temperature, 280 0 C; time of contact, 0.6 ks (1, 4); 2.4 ks
(2); 7.2 ks (3); (4) residual adhesive strength versus
TC temperature after 60 ks of joints' exposure to water.

After water exposure, the residual strength of joints is de-

pendent on the time-temperature regimes of TC between IL
and metal. The higher the TC temperature, the stronger the
residual strength (Figure 10, curve 4). At a fixed tempera-
ture the residual strength increases with TC period
(Figure 10, curves 1 - 3).

After the joints were exposed to water, the amount of

IL material retained on the substrate after separation decre-
ased. Figure 11 shows that ILs treated at 280°C for
0.6 ks on the substrate were separated completely after a
prolong exposure (10 - 15 h) to boiling water. The adhe-
sive strength of joints formed on substartes which had been
separated from coatings exposed to water for 10 - 15 h
(curve 1) equalS to the adhesive strength of the joints
that were free of ILs (dashed curve). With ILs, the bond
with metal did not fail completely in water after a prolong
(7.2 kS) thermal contact at 280 0 C. The residual IL on
JOINTS BETWEEN ALIPHATIC POLYAMIDES AND METALS 209

A, kN/m

3 2

2
1

0 14.4 28.8 43.2 LH 2 0 ' ks

Figure 11. Variation of adhesive strength in P6/ aluminum

joints formed after coating was separated from aluminum
against the time of specimen exposure to water (C[ H 0 ) at
100 0 C. IL was of P548 with S- = 1/"m. TC con~tions:
temperature, 280 0 C; time of exposure : (1) 0.6 kSj
(2) 7.2 ks. Dashed line, joints free of ILs.

the substrate could be seen with a naked eye and recorded

with IR-spectroscopy (MF'l'R). When these were reused
as ILs, high adhesive strengths were recorded (Figure 11,
curve 2).

Thus, under proper conditions of thermal contact for

ILs and the metal, adhesive bonds are formed that do not
fail in water completely. Similar results have been obtained
by the authors when adhesive joints were investigated after
exposure to ethanol.

A polyamide/IL contact may be the result of physical

adsorption, diffusion, or interdiffusion of macromolecules
along with chemical bonds formed between IL and polyamide.

Polyamide/IL contacts are usually formed when poly-

amide is in a viscous-flow state. Since the accepted
technology of specimen production allows a short period of
contact (below 10 - 20 s) between the molten polyamide
and IL, and because IL undergoes cross linking after
which the molecular (segmental) mobility is restricted in it
210 S.S. PESETSKII ET AL.

A, kN/m

2 2 3
1

",
o 20 40 60 120 180 240 L, Sec.

Figure 12. Adhesive strength of P6/ aluminum joints

against the period of P6 contact with PVB intermediate
layer ( 0' = l.flm) at 225 0 C. (1) heat untreated IL;
(2, 3) heat treated IL at 280 0 C in air for 0.9 ks and
7.2 ks, respectively•

.::
o

a.

o
!Il
.0
<l;

-1
cm

Figure 13. IR-spectra (MF'I'R) of PVB films ( b = l.ftm)

on aluminum foil. (1) heat untreated film; (2, 3) heat
treated film at 280 0 C for 0.6 and 3.6 ks, respectively.
JOINTS BETWEEN ALIPHATIC POLYAMIDES AND METALS 211

during thermal contact with metals 7,8, the diffusion mechanism

of contact formation seems almost unlikely 16,17. With short-
time contacts of thermoplastics, chemical reactions seem
unlikely between these. Hence, it is believed that the physi-
cal intermolecular relations make a major contribution to the
bond development between IL and metal. This idea is
supported by the results of kinetics study of the adhesive
contacts when PVB was used for ILs (Figure 12).

The joints produced with heat untreated PVB intermedia-

te layers possess low adhesive strengths after instantaneous
contact and fail at the P6/PVB interface. Their strengths
improve gradually with the time of contact (Figure 12,
curve 1). Heat treated ILs give adhesive strengths similar
to the ultimate magnitude after a short contact period
(Figure 12, curves 2 and 3). The heat treated ILs in
air provided a higher concentration of oxygen-containing
polar groups (carbonyl groups in particular) in PVB
surface layers which is shown through an increased IR-
absorption at 1730- 1 cm (Figure 13). Thus, the surface
oxidation of intermediate layers (ILs) improves the bonding
capacity in IL/polyamide systems.

It may be stated from the above that intermolecular

relations of macromolecular polar groups at the interface
playa major role in the development of contact between
the coating and intermediate layer.

CONCLUSION

The application of intermediate adhesive layers in the

form of thin films (applied from polymeric solutions) onto
metals appears an efficient means of controlling the adhesive
strength of aliphatic polyamides/metal joints. One of the major
parameters that determine the adhesive strength was found to
be the IL thickness, the optimum value of which ranges from
a few tenths to 1 - 2 )Am. The improvement in adhesive
strength is most obvious during thermal contact of thin IL
with a metal in air. An adhesive bond develops between the
metal and IL which is resistant to dynamic loadings and
liquidS. Polyamide/IL bonds result usually from intermolecular
interactions of macromolecular polar groups at the interface.
These investigations reveal that a cohesive mode of
joint failure through IL material is dominant. It is suggested
that the inhibition of thermal oxidative degradation in thin ILs
under thermal contact with metal improves their adhesion to
the metal, and also the cohesive strength.
212 S.S. PESETSKII ET AL.

REFERENCES

1. S.S. Pesetskii, N.I. Egorenkov, and S.V. Shcherbakov,

in "Composite Polymeric Materials", Issue II, p. 39,
Naukova Dumka, Kiev, 1981.
2. S.S. Pesetskii, N.I. Egorenkov, and S.V. Shcherbakov,
Colloid Journal, No.5, 992 (1981).
3. N.I. Egorenkov, V.N. Mlynskii, and E.N. Sokolov,
Mechanics of Polymers J., No.1, 40 (1978).
4. V.N. Mlynskii, N.I. Egorenkov, and V.A. Belyi, Physico-
Chern. Mechanics of Materials J., No.2, 83 (1975).
5. N.I. Egorenkov, A.I. Isterin, and V.L. Potekha, Mechanics
of Polymers J., No.1, 170 (1976).
6. S.S. Pesetskii, "Study of Moulded Polyamide/Metal
Adhesive Joints". Dissert. Thesis. Riga, 1980.
7. S.V. Shcherbakov, V.E. Starzhynskii, and S.S. Pesetskii,
BSSR Academy of Sciences Reports, 20, No.12, 1086
(1976) .
8. S.S. Pesetskii and S.V. Shcherbakov, BSSR Academy of
Sciences Trans., Sere Phys. & Techn. Sci., No.1, 45
(1979) .
9. W.A. Belyj, S.W. Scherbakow, und S.S. Pessezkij, Plaste
und Kautschuk, 27, No. 11, 642 (1980).
10. S.V. Shcherbakov, S.A. Osipenko, and S.S. Pesetskii,
BSSR Academy of Sciences Trans., Sere Phys. & Tech.
Sci., No.1, 127 (1974).
11. S.V. Shcherbakov, S.S. Pesetskii, S.A. Osipenko, and
O.N. Kurganova, BSSR Academy of Sciences Trans.,
Sere Phys. & Techn. Sci., No.3, 121 (1976).
12. H.L. Fridman, in Polymer Pre prints , .1, No.2, 662 (1966).
J. Polymer Sci., .2., 183 (1964).
13. M.B. Neuman, Editor, "Aging and Stabilization of Polymers':
Nauka, Moscow, 1964.
14. E.T. Gevorkyan and L.V. Baranova, High Molecular
Weight Compounds J., 196, No. 10, 758 (1977).
15. V.P. Karlivan, et al., Plastics J., No. 11, 46 (1976).
16. S.W. Stscherbakow, Herstellung verstarkter metallpoly-
merer Erzeugnisse durch SpritzguB. Sammelband
"Verstarkte Plaste-80", S. K.9.1-K.9.8, Dresden, DDR ,
(1980).
17. A.A. Berlin and V.E. Basin, "Fundamentals of Polymer
Adhesion", Khimia, Moscow, 1974.
"HONEYMOON" PHENOLIC FAST-SETTING ADHESIVES FOR EXTERIOR-GRADE

FINGER JOINTS

A. Pizzi

National Timber Research Institute

Council for Scientific and Industrial Research
P.O. Box 395, Pretoria 0001, Republic of South Africa

All fast-setting adhesive systems tested proved

satisfactory for accelerated finger joints production
as they show development of strength at ambient tempe-
rature rapid enough to allow finger joint machining
five to 30 minutes after manufacture, according to the
resin s¥stem used. The results of accelerated weathe-
ring tests showed all the adhesive systems to be
weather- and boil-proof. The main advantages obtained
were:

(i,) A considerable decrease in the cost of phenolic

adhesives for fingerjointing obtained by the use of
tannin extracts, and

(ii) improvement in production flow obtained by the

decrease of the time delay between the manufacture of
the finger joints and their machining or dispatch from
an overnight period to approximately ten to 30 minutes
at ambient temperature according to the adhesive system
used.

213
214 A.PIZZI

INTRODUCTION

Finger joints are commonly used in South Africa to produce

long boards from short-length timber. Such joints are acceptable
in structural timber and in laminations for glulam (Table 1).
The adhesives normally used for finger jointing are melamine/ureal
formaldehyde and phenol/resorcinol/formaldehyde which require
lengthy periods or temperature higher than ambient to set. There
is, therefore, a one-day delay between fingerjointing and further
processing and despatch which is inconvenient and interferes with
production flow.

Separate application adhesives capable of setting faster than

conventional adhesives were developed in the U.S.A. 10 ,1 and in
other countries 7 ,12 to glue large components where presses were
impractical 1-4. Kreibich 4 describes these "honeymoon" systems as
follows: "Component A is a slow-reacting resin with a reactive
hardener. Component B is a fast-reacting resin with a slow-reacting
hardener. When A and B are mated, the reactive parts of the
components react within minutes to form a joint which can be handled
and processed further. Full curing of the slow-reacting part of the
system takes place with time." (Figure 1).

In South Africa the basic "honeymoon" gluing system was

considerably modified by: (i) the use of tannin-based rather than
synthetic phenols for both components A and B; (ii) tae elimination
of the hardener from component B, and (iii) the elimination of the
expensive m-aminophenol from component B.

RPF resin P-F prepolymer

< ~
React with m-aminophenol
'V H
Z
H
z
rz:l rz:l
Mix with hardener Z
0
Z
0
p., p.,
~
u
~ Mix with hardener
u

Apply to surface B

Mate surfaces A and B

~------~ ~--------~
under pressure

Figure 1. .
Flow diagram of fast-set adhes1ves accor d·1ng to Kre1·b·1Ch4•
"HONEYMOON" PHENOLIC FAST-SETTING ADHESIVES 215

Table I. Finger joint dimensions.

The dimensions in a finger joint profile should be as follows (see
Figure of Finger Joint Profile, below}:
1. The finger length, L, should be at least 7.5 mm.
2. The pitch should not exceed the appropriate of the following
maxima:

Finger length, L, mm Pitch, p, mm, max.

7.5 2.5
Over 7.5 up to and including 10 L
D
Over 10 up to and including 24 L
3.0
Over 24 up to and including 35 L
3!.5
Over 35
4.0
3. The weakening ration v=b/p, where b=width of slot-base should
not exceed the appropriate of the following values:

Finger length, L, mm Weakening ratio, v, max.

7.5 O.OS
Over 7.5 up to and including 35 0.16
Over 35 0.20
NOTE: The maximum slot-base width can be calculated, the
relationship b=vxp being used.

4. The finger-tip thickness should be within the limits 1.1b-1.2b.

5. The profile dimensions should be such that the clearance, c, in
the glued joint (after assembly), is not greater than O.OSL of
4.00 mm, whichever is smaller.
NOTE: The presence of some clearance is desirable in joints of
any finger length.

Figure Joint Profile

L
L Finger length
t b Slot-base width
p Pitch
t Thickness of finger-tip

The following end pressures are recommended:

Finger length, L, mm End Press~,e, MPa, min.

Dense hardwood Softwoods and
other hardwoods

7.5 up to and including 10 15.3 7.S

Over 10 up to and incl uding 21 12.75 5.S
Over 21 up to and including 31 10.2 4.0
Over 31 up to and including 41 7.65 2.0
Over 41 5.1 1.0

I} Of density, at a moisture content of 12%, greater than 600 kg/m 3

216 A. PIZZI

This allowed the economical use of these types of adhesives.

South Africa is the first country to have used industrially and
extensively these types of adhesives for fingerjointing.

In the original U.S.A. adhesive, component A is a phenol/

resorcinol/formaldehyde cold-setting adhesive to which paraformal-
dehyde hardener has been added. Coconut shell and wood flour
fillers may be added or omitted from the glue mix. Component B is
instead a phenol/m-aminophenol/formaldehyde resin. This system has
been proved to be successful at laboratory levels. While finger
joints manufactured with this adhesive proved to be excellent, both
in speed of curing and durability, a few problems were encountered
in the availability, ease of supply and high price of m-aminophenol
chemical.

This study describes the development of alternative fast-setting

adhesives not containing m-aminophenol.

The adhesives used for components A and B of the "honeymoon"

systems are shown in Table II.

All the combinations of components A and B were used to

manufacture finger joints. Particular attention was paid to, and
extensive tests were done with, the adhesive systems A1/B1, A1/B2
and A1/B3.

To better understand the function of the different A and B

components of the different adhesive systems presented, it is
necessary to give some background about the resins they are
composed of.

Table IL Adhesive Components (A and B) used for "honeymoon" systems.

Component A Component B

A1. Commercial phenol/resorcinol B1. Pine tannin extract; no

formaldehyde cold-setting hardener; pH adjusted to 12.4
laminating adhesive + 1
hardener + fillers (pH=8,O) ) B2. Wattle tannin extract 2); no
hardener; pH adjusted to 12.6

B3. Commercial phenol/resorcinol/

formaldehyde resin2~ no
hardener; pH adjusted to 11.4

1) Cascophen RS 12 supplied by General Chemical Corporation Ltd,

Johannesburg, South Africa.
2) Mimosa extract supplied by Natal Tannin Extract Co. Ltd,
Pietermaritzburg, South Africa
"HONEYMOON" PHENOLIC FAST-SETTING ADHESIVES 217

Components A

AI. Phenol/resorcinol/formaldehyde laminating adhesives.

This is one of the many resorcinol/terminated phenolic resins
available on the market for cold-setting laminating applications.
All these classes of adhesives are weather- and boil-proof.

Components B

B1 and B2. Pine bark tannin extract and wattle bark tannin
extract. The phenolic nature and consequent behaviour of pine and
wattle tannins is evident by the schematic structure of their main
monomeric constituents.

(OH)
OH

o OH HO o

OH

Wattle Tannin Pine Tannin

These schematic wattle and pine tannin structures represent

the two major types of flavonoid tannins known, wattle tannins
being composed of mainlYSEgsorcinolic A-rings and pine tannins by
phloroglucinolic A-rings . The pyrogallic and catecholic B-rings
do not take part in the reaction with formaldehyde to any noticeable
extent, up to pH=+10, due to their low nucleophilic character S- 8 •
A further difference between these two tannins is the mode of
linkage of the flavonoid units 4,6 in the case of wattle tannins and
4,8 in the case of pine tannins.

The stronger nucleophilicity of the pine tannins phlorogluci-

nolic A-rings causes pine tannin to be faster reacting, with form-
aldehyde, at comparable pH's, than the resorcinolic A-rings of the
wattle tannins. As a consequence pine tannin extract, with the pH
adjusted to 12.4 was a more logical choice as a component of the
"honeymoon" system as it should allow faster curing, hence shorter
time delay before finger joint machining, than wattle tannin extract.

B3. Phenol/resorcinol/formaldehyde without any hardener and

with pH adjusted to 11.4.

The bark extracts of the black wattle tree (Acacia mearns~~,

formerly mollissima), commercially available and containing 70 %
218 A.PIZZI

to 80 % tannins, and of the pine species Pinus patuZa, pilot plant

produced by ambient temperature countercurrent methanol extraction
were used for this study.

The influence of the following parameters on the performance of

the different adhesive systems and of the finished finger joints
were evaluated: (i) wood density, (ii) wood species, and (iii)
tolerance to quantity variations, namely variations of the component
A : component B ratio of 2 : 1; 1: 1 and 1 : 2 were investigated.

To show that the "honeymoon" systems curing speeds are consi-

derably faster than conventional adhesive systems, finger joints were
also prepared by gluing both finger joint profiles with a commercial
phenol/resorcinol/formaldehyde resin + hardener and with a commercial
wattle/resorcinol/formaldehyde resin + hardener of unadjusted pH's.
Thus, the components A1 and A2 were used to glue both sides of the
finger joints in order to evaluate the rate of strength increase on
curing of traditional fingerjointing adhesives.

EXPERIMENTAL

Adhesive Resins Preparation

The phenol/resorcinol/formaldehyde resin used was a commercial

product. However, any of the numerous phenol/resorcinol/formalde-
hyde resins commercially available can be used with similar results.

Glue Mixes

(All pH's adjusted with NaOH 30 % to 40 % solution.)

Component Parts by weight
A1. PRF liquid resin (52 % solids) 100
hardener + fillers, powder 24 (20 paraformaldehyde)
pH 8.0
pot-life 1.5 to 3.0 hours at 25°C
B1. Pine extract solution (50 % solids) 100
pH 12.4
pot-life Indefinite as no hardener
is added
B2. Wattle extract solution
(50 % to 55 % solids) 100
pH 12.6
pot-life indefinite as no hardener
is added
"HONEYMOON" PHENOLIC FAST-SETTING ADHESIVES 219

B3. PRF liquid resin 100

pH 11.4
pot-life indefinite as no hardener
is added

Fingerj ointing

Boards of South African pine (Pinus patuZa, Pinus padiata,

Pinus pinastep, Pinus taeda and Pinus eZZiotii) of dimensions
500 mm x 120 mm x 37 mm and of EUaaZyptus grandis of dimensions
500 mm x 75 mm x 25 mm were mechanically fingerjointed, using
the various adhesive systems presented. All the adhesive
systems were used to produce finger joints to be tested 30 minutes,
one. hour , two hours, four hours, 24 hours and seven days after
gluing to evaluate the rate of strength increase with time.

The experimental series of finger joints which were manufactured

were the following:

General. Finger joints were prepared using medium density

pine (approximate density range 0.450 g/cm3 to 0.550 g/cm3), using
the adhesive systems presented.

Effect of wood density. Finger joints were prepared

using high density pine (approximate density range 0.650 g/cm3 to
0.,]50 g/cm3), using the AlB adhesive systems Al/B2 and Al/B3.

Tolerance to quantitative and viscosity variations.

As it is practically impossible, under factory production conditions,
to have identical quantities of adhesive on both the two profiles of
each finger joint, it was deemed necessary to investigate the
tolerance of the adhesive systems to variations of quantities. This
was achieved by varying the glue mix viscosity of the two adhesive
components in order to have approximate proportions by weight of
A : B of 2 : 1; 1: 1 and 1 : 2. The viscosities used were as
follows:

Component A Component B Approximate mass

3.8 Pa.s 1) 2.0 Pa.s (2 1)

1.9 Pa.s 1.9 Pa.s (1 1)
1.9 Pa.s 3.5 Pa.s (1 2)

1) 1 Pa.s 1 000 centipoises

The AlB adhesive systems used, were 1/2 and 1/3. Gel times at
60°C of mixtures of different mass ratios of component A :
220 A. PIZZI

component B, namely 25 : 75; 33 :.67; 50: 50; 75: 25 and

100 : 0 were also measured. The gel times results for the AlB
adhesive system AllBl are shown in Table III.

Control. Series of finger joints in which both profiles of

each finger joint were glued wit~ component Al were also prepared.
This was deemed necessary to compare the rate of strength increase
of the "honeymoon" adhesive systems with that of more traditional
cold-setting adhesives for finger joints.

Testing
The finger joints prepared were tested to failure in center
loading, three points, bending, 30 minutes, one hour, four hours,
24 hours and seven days after gluing and assembling. Breaking
strength in bending for a span of 914.4 mm as well as percentage
wood failure of the broken finger joints was assessed according
to the South African Bureau ijf Standards SABS 096-1976 specification
for structural finger joints. The SABS specification requires a
minimum of 20 MFa in bending. Several samples for each time span
were tested. For all the finger joints series prepared with the
different adhesive systems, accelerated weathering tests according 9
to the South African Bureau of Standards SABS 096-1976 specification
for structural finger joints were carried out seven days after
gluing and assembling. The accelerated weathering tests consisted
of testing in tension, with a universal testing machine, finger
joint samples of 250 mm x 25 mm x 3 mm, dry, after 24 hours cold
water soaking and after six hours boiling. The SABS 096-1976
requirements for exterior grade structural finger joints after
24 hours cold water soaking and/or six hours boiling are the
following:

Minimum failing load Wood failure

(Newtons) (%)

1 400 90-100
1 700 70-89
2 100 50-69
2 500 30-49
2 800 10-29

The average shear strength and percentage wood failure results

obtained for the different adhesive systems and finger joint series
are shown in Table IV.
"HONEYMOON" PHENOLIC FAST-SETTING ADHESIVES 221

Table III. Gel time at 60°C of adhesive A1/B2 containing different

ratios of components A and B.

A B Mean Gel Time

Parts by Mass Parts by Mass Seconds

25 75 50.0
33 67 39.3
50 50 37.3
67 33 39.3
75 25 45.6
100 o 320.0

Statistical analysis

As the total experiment involved nearly 1 000 finger joints,

the large sample allowed meaningful statistical analysis to be
applied to the results. In this type of experiment the most
useful information concerns the relationship between the strength
of a joint and the period of time after gluing and assembling.
The strength versus time curves obtained are of considerable
importance as they show how long after gluing and assembling the
finger joint has enough strength to be machined or dispatched
without damages. This relationship can normally be expressed in
some form of regression equation or graphic curve.

The analysis of the experimental results was looked at accor-

ding to two different approaches. In the first method one adhesive
system was selected and simple regression analysis was used to
establish the relationship of strength as a function of the density
of the timber. Several regressions were performed where each
equation corresponded to a unique gluing time, viz. for 30 minutes,
one hour, two hours, four hours, 24 hours and seven days. The
strength versus timber density graphs were plotted for the 30
minutes to four hours period as this was the more interesting time
span. A certain density was then selected and the corresponding
strength value was determined from the strength versus density curves
for each time span. These strength versus time span values were
then plotted. This procedure was repeated for the rest of the
adhesive systems.

In the second and more comprehensive approach it was decided

to use multiple regression analysis to obtain, for a certain
adhesive system, an equation which would describe strength in terms
of the other variables. The independent variables considered
were the following:
 """"
""

Table IV. Average Shear Strength and Wood Failure Results of B.A. Pine and E. saZigna Finger
Joints Prepared using A1/A1; A1/B1 and A1/B3 Adhesive Systems.

B.A. Pine E. sa Zigna

Dry 24h cold soak 6h boil Dry 24h cold soak 6h boil
:>.
.j..I
:>.
.j..I
•.-1 '.-1
o:t.l ....... o:t.l
Z s-.e Z
......, s-.e
......, Z s-.e Z s-.e s-.e Z s-.e
Adhesive '-' '-' '-' '-' ~ '-' '-' '-' '-' '-' ~
. "t:I .. ......, .Z "t:I
..d .. ..d ..d .. ..d . ..d .. ..d ..
System .j..I
- -.. ..-. -.. . - -
s:: Q) .j..I s:: Q) .j..I s:: Q) Q)- .j..I
- -.
s:: Q) .j..I s::
-.. ..- -..
Q) .j..I s:: Q) Q)-
000 1-1 bOO 1-1 1-1 bO ... bOO 1-1 bOO 1-1 bOO 1-1 bO ...
S:: • .-I ;:I S::'.-I ;:I s::bOO
•.-1 ;:I Cd S::'.-I ;:I S::'.-I ;:I S::'.-I ;:I Cd a
Q) CIl 'OM Q) CIl Q) CIl 'OM Q) f/) 'OM Q) CIl Q) CIl 'OM
'OM 1-1
a0 'OM 1-1 0
1-1 0'.-1 1-1 s:: 0'.-1 1-1 s:: 0'.-1 Q) ........ 1-1 s:: 0·.-1 1-1 s:: 0'.-1 1-1 s:: 0'.-1 Q) ........
.j..I .j..I .j..I .j..I
s::Q) Q) Q) Q) .j..I Q) .j..I Q)
OO.j..l
o Cd C/l.j..l
o Cd Ul.j..l
o Cd ::> bO C/l.j..l
o Cd o Cd o Cd > bO
:31.1-1 :31.1-1 :31.1-1 <'-' :31.1-1 :oo.u :31.1-1 C/l.j..l :31.1-1 <'-'

A1/A1 3 720 96 2 132 76 2 170 87 0.500

A1/B1 5 252 90 3 474 85 3 558 89 0.512
A1/B3 4 060 94 2 658 94 2 824 92 0.531 3 985 88 4 182 78 3 538 80 0.481

»
"'tI
N
N
"HONEYMOON" PHENOLIC FAST-SETTING ADHESIVES 223

1 - curing temperature in °c
2 - curing time in hours
3 - percentage wood failure
4 - wood density in kg/m3

The dependent variable was the modulus of rupture (MOR)

expressed in MPa. As the points obtained appeared to follow a
logarithmic law for the curing time, it was decided to use both
the curing time and its natural logarithm in order to obtain more
elastic and pliable equations able to better describe the phenomenon.
The multiple regression analysis was performed with forced variables
and not-forced variables. The multiple regression equations with
the relative coefficients for the statistically significant
independent variables and the percentages of the phenomenon explained
for one of the adhesive systems for which there were enough data
to perform a multiple regression analysis are shown in Table V.

Table V. Example of dependence, by non-linear multi-regression

analysis, of MOR increase from curing time, temperature, percentage
wood failure and wood density for S A Pine finger joints glued with
a fast-setting adhesive system.

Adhesive system Al/Bl

- Unforced variables:

MOR = 2.773 temperature + 5.285 In time - 48.308

Percentage of phenomenon explained = 79.00 %

Level of significance of variables:
curing temperature = 100.00 %
In of curing time = 100.00 %

Forced variables:

MOR = 2.119 temperature + 0.047170 time + 0.075485 wood

failure + 0.029009 density + 4.026 In time - 50.583

Percentage of phenomenon explained = 80.95 %

Level of significance of variables:
curing temperature 98.97 %
curing time = 75.58 %
percentage wood failure 88.79 %
wood density 93.24 %
In of curing time 100.00 %
224 A. PIZZI

INDUSTRIAL APPLICATION

In Tables VI and VII are shown the testing results of finger

joints manufactured in a three-day production factory trial using
the adhesive systems A1/B2 and A1/B3. To investigate if planing of
the finger joints 30 minutes after manufacture would damage the glue
line and decrease its strength, finger joints were tested rough, as
manufactured, and after planing. From the bending and tensile
strength results it appeared that planing of the joints 30 minutes
after manufacture has no bearing on the strength of the joint.
The joints were tested 30 minutes, one hour, two hours, 24 hours
and seven days after manufacture. All the components A and B of
the adhesive system used had been industrially manufactured. The
viscosities of the adhesive systems used, were the following:
A1/B2 = 3 400/3 200 cps; A1/B3 = 3 800/3 700 cps. As a consequence
of this industrial production trial, the South African Bureau of
Standards accepted the new adhesive systems as suitable for exterior
structural grade finger joints of timber of grades 4, 6 and 8.
Timber of grades 4 and 6 constitute +99 % of the timber finger-
jointed in the Republic of South Afr1ca.

DISCUSSION

The results shown in the figures and tables indicate that all
the adhesive systems tested perform sufficiently well to allow
fast binding of finger joints. The interesting feature of all the
adhesives tested is that notwithstanding the differences in curing
and setting rates among the different systems, they are fast enough
in developing sufficient strength to allow short delay between
assembly and machining or dispatching. The systems A1/B1 and A1/B2
are of considerable economical importance as a proper manufactured
adhesive is used only on the "A" profile of the finger joint while
the "B" profile is coated with a considerably cheaper and easily
available tannin extract.

The curves in Figure 2 show that phenol/resorcinol/formaldehyde

(PRF) adhesives are very well suited to the function of components
A and B, and that tannin extract are as suitable to the function of
component B as PRF resins. Very little difference exists in the
results obtained with the A1/B1 and A1/B3 resin systems.

The curves in Figure 2 clearly indicate that the rate of strength

development in finger joints glued with conventional adhesives are
much slower than for joints glued with the various "honeymoon"
adhesives. The 20 MPa level is reached, for conventional phenolic
cold-setting adhesives, in 100 minutes for the A1/A1 system, hence
in periods considerably longer than with "honeymoon" systems.
"HONEYMOON" PHENOLIC FAST-SETTING ADHESIVES 225

AI/B3

/
./
I AI/AI
f-3O /
l.? /
Z /
W /
0:: /

tn 25 /
/

/
l.?
Z
~ 20~ _ _/----'l------=G.:.:..'R.:.:..~.:::o.:..::'E:......=;8___ /

-.... ....,.,'" ... ....'""" - ....

" dAy.
W
CD S'yn_
DtaIU.'Y Deaeaty DuJ.II11;Y
hlhlr. hll\lre relllH.
19/~) li/~1 C'i!~'

151--H-=~::::='-"'-7"''------ t o,l .,., '"

AI /_I
Al f.3 "
11
0 ,&76 U,2 O,SL::II
0 , $11
".'
n ,2 " O,ElC'
O. 61'J1
10.1 ; 10.1
Q-.,nEi H,I
S'I, " ,. 0,$22 Sl ,fI "
100 0,621

2 4
TIME (hours)

Figure 2. Development of finger joint bending strength as a

function of time. Systems AI/Al; AI/BI; AI/B3 = S.A. Pine.
226 A.PIZZI

Table VI. Development of Bending Strength versus Time;

Industrial Production Finger Joint Trial.

Curing time Mean Values 1)

Treatment A1/B2 A1/B3
(hours)

Ambient Temperature °c 21 24

0.5 Planed MOR (MPa) 8.6 9.6

Density (kg/m3) 538 555
Wood failure (%) 2 1

Rough MOR (MPa) 8.5 12.1

Density (kg/m3 ) 527 531
Wood failure 1 0

1.0 Planed MOR (MPa) 12.5 11.5

Density (kg/m3) 617 612
Wood failure (%) 2 5

Rough MOR (MPa) 12.9 13.4

Density (kg/m3) 598 584
Wood failure (%) 2 2

2.0 Planed MOR (MPa) 18.6 16.8

Density (kg/m 3 ) 505 501
Wood failure (%) 26 18

Rough MOR (MPa) 17.1 16.2

Density (kg/m3) 510 480
Wood failure (%) 19 18

24.0 Planed MOR (MPa) 35.4 34.2

Density (kg/m3 ) 527 640
\V'ood failure (%) 100 70

Rough MOR (MPa) 35.1 33.0

Density (kg/m3 ) 512 633
Wood failure (%) 91 47

1) Mean taken of five specimens

 ::I:
o
Z
m
-<
~
o
o
z
""0
::I:
m
Z
or
Table VII. Accelerated Weathering Test of Finger Joints Manufactured Industriall~ n
"'TI
Tension Test of 250 mm x 25 mm x 5 mm Specimens. »
CJ)
-!
Dry 24h Cold Water Soak 6h Boiling Water en
m
Failing Wood Failing Wood Failing Wood -!
Adhesive system -!
load failure load failure load failure Z
(N) (%) (N) (%) (N) (%) G)
»
c
A1/B3 Planed 5 741 97 3 467 25 3 946 47 ::I:
m
Rough 5 846 97 3 833 34 4 247 54 CJ)
<:
m
CJ)
Each value presented in this table is the mean taken of 25 specimens.
Thanks for this table are due to the South African Bureau of Standards.

N
N
--.J
228 A. PIZZI

The results of the accelerated weathering tests showed that

all the finger joints produced with the various "honeymoon" adhesive
systems satisfy with ease the requirements for weather- and boil-
proof fingerjointed structural timber.

Interesting is also to compare the equations correlating the

modulus of rupture (MOR) in bending with other independent variables
obtained by non-linear mUltiple regression analysis for adhesive
systems 13 • This comparison gives an insight on the differences in
intrinsic characteristics of the various resins which are related
to their molecular structure. The independent variables which
mostly contribute to the MOR values are those appearing in the
unforced variables equations 13 . The MOR of the conventional
control adhesive, which is quite slow-curing, depends mainly on
the length of time the adhesive is left to cure 13 • The MOR of the
adhesive system containing the very reactive m-aminophenol is
instead quite independent of the curing time as high MOR's are
achieved very quickly after assembly of the joint. The latter
adhesive system is though strongly dependent on the density of the
wood used (both the percentage wood failure and the wood density
variables appear in the equation). Such a behaviour has often been
noticed in industrial practice and it is caused by the rapid
thickening and cure, on assembly, of the m-aminophenol-fortified B1
component, the rate of setting and molecular size increase of which
causes so fast a thickening to allow only limited penetration of
the adhesive in the wood grain and adhesion only to the end-grain
of the more porous, lower to medium densities, wood species.

CONCLUSIONS

Pine and eucalyptus finge.r joints manufactured with the new

adhesive systems presented develop sufficient strength to allow
further processing within a very short period (five to 30 minutes
at 25°C to 28 °C).

The "honeymoon" adhesive systems tested are capable of such

fast curing that pine and eucalyptus finger joints comply with the
requirements of the South African Bureau of Standards SABS 096-1976
specification for the manufacture of fingerjointed structural
timber for bending strength (MOR) within 15 to 45 minutes, and for
weather- and boil-proof tensile strength within one week from
assembly.

The main advantages of these adhesives a~e economical as:

(i) the delay between the manufacture of fingerjointed timber and
further processing or dispatch is decreased from an overnight period
to anything between five and 30 minutes according to the adhesive
system used. This eliminates the accumulation of a full-day
production stock and therefore improves production flow. This is
"HONEYMOON" PHENOLIC FAST-SETTING ADHESIVES 229

achieved without altering the existing equipment and method of

production, and can therefore be easily implemented; (ii) compo-
nents B1 and B2 are simple tannin extracts with adjusted pH. B2
is commercially available at the ridiculously low price of only
13 % to 15 % of commercial phenol/resorcinol/formaldehyde adhesives
mass/mass and solids/solids. The coupling of the A2 component,
commercial phenol/resorcinol/formaldehyde adhesive, with the B1 and
B2 components also considerably decrease costs in adhesives for
fingerjointing. The A1/B3 system, all synthetic, disprove, once
and for all, the need of addition of fast-reacting and expensive
components, such as m-aminophenol, to fast-setting finger joint
adhesives as claimed by previous authors 1 ,4,11,12. Unnecessary
is also the preliminary warming of the timber while using conven-
tional phenol/resorcinol/formaldehyde adhesives as now sometimes
advocated in North America; (iii) increased independence from
oil-derived synthetic adhesives and from a scarce, expensive and
not easily available raw material such as m-aminophenol, and
(iv) use of only one adhesive for beam laminating and fingerjoin-
ting; the preparation of the finger joint adhesive systems requires
only the addition of one component, component B, to the usual
beam-laminating adhesives.

Disadvantages of these systems are: (i) preparation of the

adhesive is complicated by the fact that it consists of more
ingredients than conventional adhesives, namely two resins and one
hardener against the one resin and one hardener of conventional
systems; (ii) complete coverage of both profiled faces is essential,
and (iii) mis-matching of profiled faces after adhesive application
would lead to slow curing or no curing at all.

These disadvantages though have been found to be minimal in

industrial production practice. Sixty per cent of the South
African fingerjointing industry is now using one of these fast-
setting adhesives on a regular production basis.

REFERENCES

1. G. F. Baxter and R. E. Kreibich, For. Prod. J., 23, (1), 17

(1973) •
2. W. Caster, For. Prod. J., 23, (1),26 (1973).
3. H. Ericsson, Papper och Tra, (1), 19 (1975).
4. R. E. Kreibich, Adhesives Age, (17), 26 (1974).
5. A. Pizzi, J. Appl. Polym. Sci., 23, 2777 (1979).
6. A. Pizzi, J. Appl. Polym. Sci., 24, 1257 (1979).
7. A. Pizzi and D. G. Roux, J. Appl-.-Polym. Sci., 22, 1945 (1978).
8. A. pizzi and H. o. Scharfetter, J. Appl. Polym.-Sci.,~, 1745
(1978).
9. "South African Bureau of Standards SABS 096- 1976", Specification
for the manufacture of fingerjointed structural timber,
230 A. PIZZI

Pretoria, South Africa, 18 pp.

10. G. T. Tiedeman andM. R. Sanclemente, J. Appl. Polym. Sci.,
17, 1813 (1973).
11. ~ T. Tiedeman, M. R. Sanclemente and H. A. Smith, J. Appl.
·Polym. ScL, 17,1819 (1973).
12. P. K. van derlWesthuizen, A. pizzi and H. O. Scharfetter,
Wood Southern Africa, 3, 7 (May 1978). Paper presented at
IUFRO Wood Gluing Working Party S5.04, Merida, Venezuela,
October 1977.
13. A. Pizzi, D. du T. Rossouw, W. E. Knuffel and M. Singmin,
Holzforschung und Holzverwertung, 32, (6), 140 (1980).
AGEING OF STRUCTURAL FILM ADHESIVES - CHANGES IN CHEMICAL AND

PHYSICAL PROPERTIES AND THE EFFECT ON JOINT STRENGTH

C.E.M. Morris, P.J. Pearce and R.G. Davidson

Materials Research Laboratories

Department of Defence Support
Ascot Vale, Vic., 3032, Australia

Epoxy-based film adhesives are extensively used

in structural aircraft applications but although the
one part nature of these materials has many advantages
in terms of ease of use, the short shelf-life can be a
serious disadvantage, especially when the material
spends lengthy times in transit between manufacturer
and user. Studies on a number of epoxy and nitrile-
epoxy adhesives have shown that slow cure, hydrolysis
of the resin and specific interactions between
components can occur during storage which result in
modification of various chemical and physical
properties of the uncured adhesives. The relative
importance of these reactions depends on the adhesive
composition. These modifications are reflected in
changes in the strength of joints made with aged
adhesives. This paper presents examples of these
effects drawn from the results of a number of ageing
studies.

231
232 C.E.M. MORRIS ET AL.

INTRODUCTION

Many of the adhesives used for structural applications in

aircraft are in the form of one part films of very limited shelf
life, even when stored at low temperatures. The increasing use
of bonding in critical areas, including primary aircraft
structure, has focussed attention on the nature of the reactions
which occur during storage, chemical methods of assessing the
extent of these reactions and the effect of such reactions on
the joint strength of the cured material. These considerations
are of particular concern in Australia since all adhesives of
this type are imported and must be transported long distances
from the point of manufacture. In addition, the usage of
particular adhesives is quite small so that means of extending
the usable life are of interest.

Considerable efforts by many workers have been devoted to

the establishment of chemical methods of characterising epoxy-
based adhesives f~g composite prepregs for the purposes of
quality control. Some attention has also been given to the
(adverse) effects of a high humidity atmosphere on ~7 properties
and cure characteristics of adhesives and prepregs. While
some consideration has beeg ~iven to certain aspects of the
ageing of these materials,' little attention seems to have
been paid to an overall study of the changes in the chemical and
physical properties of adhesives during storage on the one hand
and the strength of joints made with aged adhesives on the
other.

This paper summarises the results of a number of ageing

studies on two ~lasses of film adhesives: 177°C18~I~ng, epoxy
systems and 121 C curing, nitrile-epoxy systems. Several
types of reactions have been identified and their effects on
joint strength considered.

EXPERIMENT AL

Examples of two classes of film adhesives have been

studied: (a) epoxy-based, 177°e - curing systems, designed for
metal-to-metal bonding in aerospace applications, for service up
to about 200 0e and (b) nitrile rubber modified epoxy systems,
l200e - curing, designed for aerospace metal-to-metal bonding
for service up to about lOOoe.

For ageing, adhesive specimens were stored in polyethylene

bags at about 23°e. Samples were removed at intervals for
testing or, if necessary, stored at -l8°e until required.

For high pressure liquid chromatography (HPLe) a modified

AGEING OF STRUCTURAL FILM ADHESIVES 233

Varian 8500 instrument was used with an Altex 153 UV detector

operating at either 254 or 280 nm. The equipment was operated
in the reverse phase, gradient elution mode with a Du Pont
Zorbax column and the combination of tetrahydrofuran (THF) and
water as the solvent. For gel permeation chromatography (GPC)
essentially the same equipment was employed with a Waters model
R40l refractive index detector in series with the UV detector,
and using four ~Styragel columns and THF as the solvent. ll ,14

Materials associated with selected peaks in HPLC and GPC

traces were collected and identified by infrared analysis using
a Perkin Elmer 580B, ratio-recording, double beam
spectrophotometer, interfaced with a model 3500 y~ta handling
system. Procedural details are given elsewhere.

Epoxide content wa~4determined by non-aqueous titration,

as previously described.

Adhesive flow was assessed by measurements on a disc of

adhesive cured between release sheets in a heated platen press
using the cure cycle recommended for the adhesive.

Single overlap joints were made with 1.6 mm 2024-T3 Alclad

aluminium sheet for which the surface preparation was a vapour
degrease with l,l,l-trichloroethane followed by a chromic acid
etch. No primer was used. Joints were made in a heated platen
press using the cure cycle recommended for the adhesive. Joint
geometry and test methods were in accord with standard
procedures. 15 The testing was accomplished with a model TT-C-L
Instron tensile testing machine. Results are reported as the
average of 6 replicates.

Bonded honeycomb panels were made using 9.5 mm cell size

aluminium honeycomb 12.5 mm thick with 0.13 mm 2024-T3 face
sheets. These panels (75 mm wide) were tested by climbing drum
peel.

Dynamic mechanical properties of cured adhesive samples

were assessed with a Rheovibron DDV-II-C direct reading dynamic
viscoelastometer, operating in the tensile mode at 110 Hz and a
heating rate of about 2°C/min. Samples for examination were
cured in batches in the press, using a polyester template, to
ensure equivalent thermal history.

REACTIONS OCCURRING DURING AGEING

Slow Cure

The adhesives in this study all contain latent curing

234 C.E.M. MORRIS ET AL.

systems (dicyandiamide, dicyandiamide/Monuron or the 2,4-

to1y1ene di~isocyanate/dimethy1amine adduct). Room temperature
(about 23°C) is far below the recommended cure temperature for
these systems (177°C for the former and 121°C for the latter
two). In addition, these curing agents are only very slightly
soluble in epoxy resins below about 80°C. In spite of this,
over a period of weeks at ambient temperature some
po1ymerisation (chain extension and cross1inking) can occur.

Polymerization can be expected to result in a reduction in

the epoxide content (increase in epoxy equivalent weight, EEW)
and a reduction in the solubility of the adhesive in organic
solvents. Where a substantial increase in molecular weight
occurs, reduced flow of the adhesive during cure would be
expected.

Hydrolysis

Certain types of epoxy resins, notably those containing

the N-g1ycidl1 aromatic amine moiety, are susceptible to
hydro1ysis. 1 Resins of this type are frequently incorporated
in adhesives for high temperature applications. Resins based on
the dig1ycidy1 ether of bispheno1 A (DGEBA) are regarded as
stable to hydrolysis. However, recent evidence indicates that
in certain adhesives based on y~EBA resins some hydrolysis of
the resin occurs during ageing (Figure 1).

Hydrolysis of the resin leads to a reduction in epoxide

content and increase in polarity. The latter effect may modify
the moisture uptake properties of the resin.

Triglycidyl 14-aminophenol}

DGEBA
Figure 1. Epoxy resins.
AGEING OF STRUCTURAL FILM ADHESIVES 235

Interactions Between Components (Excluding Cure)

Adhesives consist of a number of components and there is
the possibility of other reactions occurring during storage.
Nitrile modified epoxy systems contain carboxy-terminated, low
molecular weight, butadiene acrylonitrile (CTBN) or
po1ybutadiene (CTB) polymers; or higher molecular weight,
carboxylated nitrile rubber; or a combination of these
(Figure 2). A reaction can occur between the carboxyl and
epoxide groups. In some nitrile-epoxy adhesives these groups
are prereacted (that is, deliberately reacted as part of the
manufacturing process) whereas in other systems they are not.
In the latter case this reaction can proceed slowly during
storage of the adhesive.

Occurrence of the carboxy1-epoxide reaction during storage

results in a reduction in the epoxide content. In addition,
depending on the rubber molecular weight, a decrease in adhesive
solubility and in adhesive flow during cure may occur.

EFFECT ON JOINT STRENGTH

The effect of the reduction in the epoxide content during

ageing on the properties of the subsequently cured adhesive
depends chiefly on modification of the crosslink density. If
the epoxide loss results from the normal cure reactions
proceeding very slowly, the chemical structure of the cured
adhesive may not differ greatly from that achieved from a normal
cure of fresh adhesive. It would be anticipated that the joint
strength would then be little affected. On the other hand,

-f' CH -CH-CH-CH I -
2 2x
(CH - CH I -
2 I y
CN
(CH -CHI
2 I
CDOH
Z
t
m

HOOCtICH,CH- Ctruy,- (C",- d:'f COOH

n
Figure 2. Nitrile rubbers.
236 C.E.M. MORRIS ET AL.

epoxide loss from non-cure processes potentially can lead to a

cured resin of lower crosslink density. This can be expected to
result in reduced joint strength, especially at high
temperatures.

A reduction in flow during cure would be most significant

in the bonding of honeycomb panels, where good flow is needed
for adequate filleting and thus the formation of strong bonds.
In the case of overlap joints, the pressure applied during the
cure cycle may be sufficient to overcome deficiencies in
substrate wetting ability of an adhesive of moderately
diminished flow. For systems of greatly altered flow, or in
circumstances where only light pressures are employed during
cure, the applied pressures may be insufficient for this purpose
and defective joints result.

AGEING STUDIES

Examples of the above reactions and their effect on joint

strength for two types of epoxy adhesives are provided from the
results of extensive ageing trials conducted at, or slightly
above, ambient temperature.

High Temperature Curing, Epoxy Adhesive

A study was conducted on two 177°C curing, epoxy adhesives

designed for service up to about 200°C. These systems contained
trig1ycidy1-(toar~nopheno1) as well as other, glycidy1 ether
epoxy resins.' Results obtained from just one of these
adhesives is described in this paper.

The HPLC analysis of fresh and aged samples is shown in

Figure 3. As the extinction coefficient of trig1ycidy1-
(4-aminopheno1) at 254 nm is much larger than that of the other
components, the chromatograms are dominated by the peaks due to
this compound, its oligomers and derivatives. Fractions
corresponding to the peaks designated 1, 2 and 3 were collected
and examined by IR. The spectra were consistent with
trig1ycidy1-(4-aminopheno11 2having 1, 2 and 3, respectively,
epoxide groups hydrolysed. An indication of the rate of
hydrolysis is provided in Figure 4 which shows the increase in
peak heights for the material with 2 and 3 epoxide groups
hydrolysed.

Table I summarises the changes in epoxide cor~ent,

solubility in ch10robenzene, and flow during cure. These
quantities altered slowly over the first 16-20 weeks of the
investigation and thereafter changed rather more rapidly.
AGEING OF STRUCTURAL FILM ADHESIVES 237

ft - - InitiQI mQter'QI

I, ------ Aged 36 weeks

~ fl
.'.:
.,~.'
.'
h

"I: .'
,
"
"
I,
"

60

SO ,
:l
:, \
, :
\
. jD
40
: ,, " \
\
% THF
) V\... i{\i\J ;

o S 10 1S
ELUTION VOLUME, ml
Figure 3. HPLC analysis of fresh and aged epoxy adhesive.
Material from labelled peaks was collected and analysed by IR.

Table 1. Change in Epoxide Content, Solubility, and Flow during

RT Ageing of a High Temperature Curing Epoxy Adhesive

Ageing Time EEW Chlorobenzene Flow

weeks g/eq. Solubility %
wt %

0 229 66 370
10 245 65 410
20 284 60 300
30 474 38 0
238 C.E.M. MORRIS ET AL.

30

---
- - Peak 2

---- Peak 3
/
./
"
I
/
.J!! 20
'1:
'" I
I
.d
'-
ID
l-
:I: I
!:2
..... I
/
:I:
~
c(
..... /
'l. 10 /
/
/
/
/

--
/
/'
...... . /
0 10 20 30 40
TIME weeks

Figure 4. Change in peak height for peaks 2 and 3 in Figure 3.

Results of dynamic mechanical analysis of cured, aged

samples are shown in Figure 5. By 16 weeks substantial changes
had occurred, especially in the region above 120°C.

Tensile-shear strength of lap joints, tested at ambient

temperature, 132°C and 177°C, are given in Figure 6. High
temperature joint strength was substantially reduced by ageing
long before any effect was discernible for joints tested at
ambient temperatures. The increased strength observed after
several weeks ageing in the 132°C and ambient cases is
attributable to increased flexibility (lower crosslink density)
of the adhesive. 12

These results are consistent with a combination of

hydrolysis and slow cure. For times up to about 15 weeks
ageing, hydrolysis effects predominate, as indicated by the HPLC
AGEING OF STRUCTURAL FILM ADHESIVES 239

'.::-.'~-" -' .. -... .. .......

'- "

-' ......

",
' ..........
......
'\
"- ....
10,0 .... , 1010
,,
"- , ..,
'\ ~
'\
\
~
X
Initial rna ter ial \
'\ :J:
8 weeks
\
,, 0
8r
16 weeks '-... 9
10 c:
26 weeks
'-- til

Cl.
'<
:J
ID
III

n
3
...
/
/
0'10 /
10 6
..- /

.... ..-
/
/
I
I
I / ,/
I ,. - .. .. /
I ./ ' .. _ .. .....
I / '

40 80 120 160 200 240

TEMPERATURE '(

Figure 5. Dynamic mechanical analysis of aged epoxy adhesive.

Upper curves - complex modulus; lower curves - tan O.

analysis and the increase in EEW but only minor changes in the
solubility and flow. Loss of epoxide content by hydrolysis
leaves the way open for a reduction in crosslink density, and
thus a reduction in the high temperature performance, of the
cured system. This is reflected in the reduced joint strength
at high temperatures. For ageing times longer than 15 weeks the
effects of slow cure (chain extension) are more evident, namely
reduced solubility and flow and a large fall in epoxide
content. These effects are ultimately reflected in joints of
very poor strength.
240 C.E.M. MORRIS ET AL.

30

....,
:I:
/",. \
/ \
---------------_ .... " \
\

...:z:'"
Q:

11\
I

~ 10
iii
.......
z
- - RT
------ 112'(
--._-- 1" ·c
o 10 20 30
TiME, ",uks

Figure 6. Tensile-shear strength of Al-AI lap joints made with

aged adhesive and tested at the temperatures indicated.

Nitrile-Epoxy Adhesives

Two 12loC curing systems were studied, both of which are

based on DGEBA epoxy resins. One system, designated Adhesive A,
contains a carboxylated nitrile rubber of molecular weight about
30,000. The nitrile rubber and epoxy resin have not been
prereacted. 16 The other system (Adhesive B) contains the same,
or a similar, rubber and some carboxy-terminated polybutadiene
of molecular weight about 4,000. The rubber and epoxy have been
prereacted. 16

Tap!es II and III show EEW and solubility data for the two
systems. A substantial and rapid change is obvious in
Adhesive A which is not evident in Adhesive B. More gradual
changes in these properties are seen in both adhesives over
longer periods of time.

GPC studies on the ageing of Adhesive A at 40°C, carried

out using both refractive index (RI) and ultraviolet (UV)
detectors, showed that a carboxyl-epoxide reaction occurred ll
(Figure 7). This is indicated by the increased UV absorbance in
the high molecular weight region (nitrile rubber) and, on
further ageing, the reduction in solubility of this high
molecular weight material. The corresponding GPC analysis of
fresh Adhesive B showed that the UV and RI traces were very
similar at the high molecular weight end indicatip~ that the
carboxyl-epoxide reaction had already taken place •
AGEING OF STRUCTURAL FILM ADHESIVES 241

Table II. Change in Epoxide Content during RT Ageing of Two

Nitrile-Epoxy Adhesives.

Ageing Time EEW

weeks g/eq.
Adhesive A Adhesive B

0 425 442
5 516 446
10 535 526
15 572 676
20 598 817

Table III. Change in Solubility during RT Ageing of Two Nitrile-

Epoxy Adhesives.

Ageing time Solubility in Chlorobenzene

weeks wt %
Adhesive A Adhesive B

o 84 88
5 60 88
10 57 80
15 54 63
20 50 52

The flow of Adhesive A dropped sharply over a short ageing

timr1wr~le that for Adhesive B fell more slowly (Figure
8).' These changes1Jere mirrored in the peel strength of
bonded honeycomb panels (Figure 8).

HPLC/IR ar~lysis indicated slow hydrolysis of the resin in

both adhesives. Figure 9 shows the analysis for Adhesive B.
Material giving rise to the peak designated B gave a spectrum
consistent with the half hydrolysed monomeric epoxy compound.
The spectrum of A was consistent with the fully hydrolysed
monomer while those of C and D corresponded to the 1§artly
hydrolysed n = 1 and n = 2 oligomers respectively. These
reactions occur slowly in the initial stages of ageing and more
rapidly in the later stages, and are rather faster in Adhesive B
than in Adhesive A.13
 242 C.E.M. MORRIS ET AL.

(
NI TR IL E RU BB ER
,( EPOX Y OliGOMERS MONURON

~
..... n
VI
Z
o
Cl.
r
VI
.....
0::
IS-Q ---"""","==
'-=='=="=='=="~
' ." ! - - - UV
•••••• II I

0::
o
I-
W
.w
o
~ &O ----~·-~·~-=-=-=·-=·~-··~·~-
...
!
..... CI
I-
.....
o VI
£ .. -.. . . . . . . . . . . _-
~ 2'0

""
Cl

(~1
. - ... _- ..
~..

\ (\/
'J

20 30 40
ELUTION VOLUME. ml

Figure 7. GPC analysis of a nitrile-epoxy adhesive (Adhesive A)

aged at 40°C.

Tensile-shear strength of lap joints tested at ambient

temperature and at 100°C showed only small changes on ageing 13
(Figure 10). Joints made with adhesive aged for more than 15
weeks were more significantly reduced in strength when tested at
elevated temperatures. Dynamic mechanical analysis indicated a
downwards shift of about 15°C in the Tg of cured samples over
the time of the ageing trial.

These results are consistent with the following

interpretation. In Adhesive A a carboxyl-epoxide reaction
occurs over the first two weeks of ageing which results in the
formation of a small amount of gel incorporating the nitrile
rubber. The presence of this gel is responsible f8r1§he rapid
fall in adhesive solubility and flow during cure.' The
latter, ,in turn, results in poor peel strength of bonded
honeycomb panels. No equivalent reaction occurs in Adhesive B.
AGEING OF STRUCTURAL FILM ADHESIVES 243

FLOW
PEEL 0·4
::E:
o
z
"'
-<
n
o
3:
0:1

O'2~
"'r-
-
z
SI
SI

~--~~--~----------~--------~----------~o
o 5 10 15 20
TIME, weeks

Figure 8. % flow and climbing drum peel strength of aged

samples of two nitrile-epoxy adhesives.

Over longer ageing times, slow f~re and hydrolysis

reactions take place in both systems. Slow cure results in
material of reduced flow. The pressure applied during the
formation of lap joints is evidently sufficient to overcome the
effect of reduced flow and the tensile-shear strength of such
joints is only slightly affected for ageing up to about 15
weeks. In the case of bonded honeycomb panels the different
configuration prevents the applied pressure from compensating
for the effects of reduced flow and the peel strength of such
joints is severely reduced after even 3 weeks ageing. As
indicated above, hydrolysis can lead to the formation of cured
material of lower performance, especially at elevated
temperatures. This effect is apparent in these systems after
prolonged ageing.
 <> N N
II
- - Initia l ma~erial .p.
c
" e .p.

----- Aged 22 weeks

:1
B I
~
, :' II
~, \ e
•• "
"
" ""
I'
"
01 \1 "
"
" "
", "
"' ,
",
70 \\, "
',
I
I
I
,,
60 I
y

~
50 ,,
,, ,"I
,, I
,
% THF I' I
I'
,, ,, . ,
, I, I,
,, ,, \ '''',, :' I ,
,, I ,, " I ,
,, , f'\,, ,,
I
,,
, ,, ,, ',I \
A
,, , ,, ", , \
I
\ '-
r, I
( ,
~
\,
1\
," , ~ I'
I , ~J
,,
I , ,.
: \...... J ()
m
s:
s:
o 5 10 lS 20 25 o
:0
:0
ELUTION VOLUME mls
Ui
m
-I
Figure 9. HPLC analysis of fresh and aged nitrile-epoxy adhesive (Adhesive B). Peaks labelled n are »
due to epoxy oligomers; A, B, C and D were analysed by rR; Y is due to the curing agent. r
AGEING OF STRUCTURAL FILM ADHESIVES 245

-RT
B ----100 ·c

30~ ____
A - - - - - - - - - - - - - -__-______-~~~~==

B
------------------------------ ---
---- --
---- -----
A
----------_ ... _-------- ----- ------ --------
..........
.........
---........
5 10 15 20 25
TIME weeks

Figure 10. Tensile-shear strength of Al-Al lap joints made with

aged samples of two nitrile-epoxy adhesives. Test temperatures
as indicated.

REFERENCES

1. J. F. Carpenter and T. T. Bartels, SAMPE Quart., I (1), 1

(1976) •
2. C. A. May, D. K. Hadad and C. E. Browning, Polym. Eng. Sci.,
19,545 (1979).
3. ~ E. Twichell, J. O. Walker and J. B. Maynard, J. Chromatogr.
Sci., 17, 259 (1979).
4. G. L. Hagnauer, Polym. Composites, 1, 81 (1980).
5. D. L. Paradis, in "Structural Adhesives and Bonding", p. 110,
Technical Associates Conference, El Segundo, CA, 1979.
6. G. W. Lawless, in "Proc. 24th Natl. SAMPE Symp.", Vol. 2,
p. 979, 1979. Available from SAMPE, Azusa, CA.
7. R. A. Pike, F. P. Lamm and J. P. Pinto, J. Adhesion, 11, 229
(1982) .
8. D. J. Crabtree, in "Resins for Aerospace", C. A. May, Editor,
p. 459, American Chemical Society, Washington, D.C., 1980.
9. R. A. Pike, F. P. Lamm and J. P. Pinto, J. Adhesion, 12, 143
(1981). -
10. C. E. M. Morris, A. G. Moritz and R. G. Davidson, in "Adhesion
and Adsorption of Polymers", L.-H. Lee, Editor, Part A, p. 313,
Plenum Press, New York, 1980.
246 C.E.M. MORRIS ET AL.

11. J. J. Pitt, P. J. Pearce, T. W. Rosewarne, R. G. Davidson,

B. C. Ennis and C. E. M. Morris, J. Macromol. Sci., Chem.,
A17, 227 (1982).
12. P. J. Pearce, R. G. Davidson and C. E. M. Morris, J. Appl.
Polym. Sci., ~, 4501 (1982).
13. P. J. Pearce, R. G. Davidson and C. E. M. Morris, J. Appl.
Polym. Sci., 28, 283 (1983).
14. P. J. Pearce, R. G. Davidson and C. E. M. Morris, J. Appl.
Polym. Sci., 26, 2363 (1981).
15. "Adhesives, Heat Resistant, Airframe Structural, Metal-to-
Metal", U. S. Federal Aviation Specification MMM-A-132.
16. C. E. M. Morris, P. J. Pearce and R. G. Davidson, J. Adhesion,
..!2, 1 (1982).
STRUCTURAL ADHESIVES BASED ON DIALLYL PHTHALATE

A. M. Usmani
The Research Institute
University of Petroleum & Minerals
Dhahran, Saudi Arabia

This paper describes dial1y1 phthalate resins as a

candidate replacement for the moisture sensitive epoxies
used in formulating structural adhesives. Polymerized
dia11y1 phthalate had most of the performance proper-
ties desired but was excessively brittle. F1exibi1iz-
ing monomers were evaluated. Some promising copolymers
were defined. Dia11y1 phthalate/epoxy interpenetrating
polymer networks were also explored.

INTRODUCTION

Surfaces are involved in many physical processes. Chemical

interaction across interfaces are studied as "pure" science or as
a way to investigate applied problems, e.g., adhesion.! Conversely,
adhesion can be studied as a basic phenomenon or as a way to in-
vestigate surface interactions.

Use of adhesively bonded structures will become more prevalent

in Air Force and Naval systems, e.g., satellites, missiles. air-
craft, weapons. and ships in the present decade and beyond. Growth
in the usage of adhesively bonded structures in the future will
result primarily from improved structural integrity of such systems.

Epoxy type. 125°C curing adhesives which represents the cur-

rent state-of-the-art in structural adhesive bonding have several
shortcomings. Their most serious deficiency is the loss of mecha-
nical properties after exposure to elevated-temperature,high-humi-
dity environments. Another problem with the currently available

247
248 A.M. USMANI

adhesives is moisture sensitivity during layup of bonded parts.

Absorbed moisture in the uncured adhesive can dramatically alter
curing behavior and final properties. There is a pressing need for
low-moisture-absorbing structural adhesives for use on aircraft and
portable tactical shelters. What is required is a chemically well-
balanced adhesive resin (aromatic or cycloaliphatic groups for
thermal and mechanical stability, alkyl groups for moisture resis-
tance, few ether linkages and hydroxyl groups for adhesion and
multifunctional reactive sites for curing) that will provide
prolonged retention of high mechanical properties when exposed to
elevated temperature in a humid environment.

This paper describes the exploratory and development work that

was performed on flexibilized diallyl phthalate (DAP) and DAP/epoxy
interpenetrating polymer network adhesive resins. Optimization of
the composition and formulation, evaluation, and product scale-up
studies are still needed before the material will be ready for
commercial application. Schematic of DAP adhesive optimization has
been provided, however.

FORM AND COMPOSITION OF STRUCTURAL ADHESIVES

Structural adhesives are generally classified by the physical

form in which they are used. Both solids and spreadable pastes
are commonly employed. 2 Examples are:

1. Solid adhesives
1.1 Film (unsupported)
1.2 Tape (supported)
1.3 Solid powders and melt-applied crosslinking types
1.4 Solvent-based
2. Solventless and high solids pastes
2.1 One-component, long shelf life
2.2 Two-component, short shelf life

Structural adhesives based on epoxies and phenolics are most

popular. Other resins used to make structural adhesive formula-
tions are epoxy-phenolic, nylon-epoxy, nitrile modified epoxy, and
polyimide (for tape and film adhesives) and one-component heat-
cured epoxy, two-component room-temperature curing epoxy, and amine-
cured polyisocyanate (for paste and liquid adhesives).

EXPLORATORY WORK ON DAP ADHESIVE

Requirements of Metal Bonding Adhesives

The adhesives should cure fully at 125°C, have good adhesion

STRUCTURAL ADHESIVES BASED ON DIALLYL PHTHALATE 249

and bonding properties, have low moisture absorption, provide

excellent bonded strength over a wide temperature range, and retain
mechanical properties when exposed to elevated temperature and high
humidity for prolonged periods.

Property-Moiety Relationship in a Resin

The properties obtained from specific chemical groups in a resin

are shown in Table 1. For low moisture absorptio~ a resin should
contain few hydroxyl, amide, and ether linkages and many aromatic
and alkyl groups. For bonding properties it is important to have
some hydroxyl, carboxyl, or nitrogen in the resin. A judicious
choice of the various groups in a resin can lead to an adhesive
meeting all the requirements outlined in the earlier paragraph.

Creating an interpenetrating polymer network (IPN) is a use-

ful approach for modifying a polymer '8 architectural design. An
lPN produces permanent entanglement that restricts the motions of
segments and thus mimics covalently bonded crosslinks. Since
property maxima and minima are obtained in lPN's, a valuable
development tool is available to develop low moisture absorbing
adhesives that will provide high retention of mechanical proper-
ties under elevated heat/high humidity conditions.

Table I. Properties versus Groups in a Resin.

Group Property Obtained

Aromatic Thermal and mechanical stability,

stiffening effect

Cycloaliphatic Thermal and mechanical stability,

stiffening effect

Alkyl Moisture and chemical resistance

Alkyl, long chain Moisture resistance, plasticization,

and toughening effect

Ether linkages Chemical resistance, flexibilizing

effect, moisture sensitivity

Amide Toughening and adhesion

Hydroxyl Adhesion, moisture sensitivity

Di- and multi-functional Curing, three-dimensional network

vinyl
250 A.M. USMANI

Based on the above considerations, we have developed flexi-

bilized diallyl phthalate and DAP/Epoxy IPN adhesive systems.

Flexibilized Diallyl Phthalate Adhesives

The excellent properties of polymerized DAP are well known. 3

These include high heat and moisture resistance, and exceptional
dimensional stability, Neat DAP monomer tends to crack during
polymerization due to shrinkage and produces a rigid product. To
avoid cracking, mineral or fiber filling is commonly used in DAP
molding compounds. In an adhesive formulation, mineral filling
can be used. However, the use of a plasticizing monomer appears
to offer an attractive alternative.

Copolymerization of diallyl phthalate with alkyl methacrylates

has been treated earlier by the author. 4 The copolymerization of
DAP and an alkyl methacrylate by a free radical mechanism at low
polymerization temperatures is less practical due to unfavorable
reactivity ratios. We considered the rates and reactivity ratios
dependence on temperature to get an insight into the method neces-
sary to prepare such a copolymer. Copolymerization of DAP with an
alkyl methacrylate can be done at elevated polymerization tempera-
ture, e.g., 125°C.

To modify rigid polymerized DAP we evaluated butyl methacry-

late (BMA), 2-ethylhexyl methacrylate (EHMA), isodecyl methacrylate
(IDMA) , lauryl methacrylate (LMA), stearyl methacrylate (SMA), and
hydroxypropyl methacrylate (HPMA) plasticizing comonomers. BNA
polymerizes quite readily with DAP but is too volatile to be use-
ful in the open. EHMA plasticized DAP, and only slight monomer
loss was encountered during curing. IDMA, LMA, and SMA were found
to polymerize satisfactorily with DAP. IDMA on copolymerization
with DAP, gave clear products. LMA and SMA on copolymerization
with DAP gave from clear to opaque resins depending upon the amount
of LMA (SMA) and the polymerization conditions.

Various ratios of DAP to comonomers, different type and

amounts of initiator, and various methods of preparing the resin
were investigated. The preferred mixtures were found to be DAP
EHMA/HPMA (80/15/5) and DAP/LMA (80/20). HPMA copolymerized
readily with DAP, and the hydroxyl groups should contribute to
increased adhesion.

Flexibilized DAP can be developed into suitable adhesive forms,

e.g., film form in which the complete formulated adhesive is car-
ried on a lightweight woven scrim fabric or mat, and paste form,
in which the adhesive and curing agent are maintained as separate
thick pastes until ready for use.
STRUCTURAL ADHESIVES BASED ON DIALL YL PHTHALATE 251

Three methods of paste (syrup) preparation were investigated.

In bottle polymerization (DAP + flexibilizing monomer + 0.2 per
cent benzoyl peroxide under nitrogen blanket at 65-70 o C for 24-48
hours), viscosity was difficult to control from batch to batch
and occasionally gelation was encountered during syrup preparation.
Another disadvantage of the bottle method is that the prepared
syrup drifts up in viscosity due to catalyst residue remaining in
the syrup. Use of DAP prepolymer (Dapon 35) with DAP monomer
produced straw colored syrups which are suitable for adhesive for-
mulations. The resin kettle method was very reproducible. It
involved. first preparing a high viscOsity DAP syrup and then dilut-
ing it subsequently with EHMA/HPMA or LMA. Slightly more uniform
copolymer was made with a slower and delayed addition of flexibi-
lizing monomer to a heated and stirred DAP. To 1600g DAP monomer,
3.3g benzoyl peroxide was added and polymerization conducted under
nitrogen blanket with stirring at 80±2°c. In about five hours the
viscosity of DAP reached approximately 10 Pa sec. At this point
the contents were quickly cooled to less than 45°C and 400g of
flexibilizing monomer, e.g., LMA was sequentially added and mixed.
A typical DAP/LMA (80/20) syrup had a viscosity of 1.13 Pa sec.

The following cycles were developed for DAP/EHMA/HPMA (80/B/~

and DAP/IMA (80/20) syrups:

° Slow cycle (in air and using 0.2 percent each of benzoyl
peroxide and DiCup catalyst): 90 o C/12 hrs. + 130°C/12 hrs.

° Medium fast cycle (in nitrogen and using 0.2 percent each
of benzoyl peroxide and DiCup catalyst): 90 o C/2 hrs. +
100 o C/1 hr. + 110 o C/1 hr. + 120 o C/2 hrs. + 130 o C/2 hrs.
(Total cure time = 8 hrs.)

° Medium fast cycle (in air and using 0.3 percent each of
benzoyl peroxide and DiCup): 125°C/12 hrs.

° Fast cycle (in air and using 1 percent DiCup): 125°C/4 hrs.

° Fast cycle (in nitrogen and using 1 percent DiCup):

125°C/l hr.
° Super fast cycle (in air and using 1 percent DiCup):
ISOoC/l hr.

° Super fast cycle (in nitrogen and using 1 percent DiCup):

I50 o C/30 min.

Other Allylic Resins

Closely resembling DAP are two other monomers; namely, diallyl
adipate (DAA) which polymerizes very slowly like DAP, and triallyl
252 A.M. USMANI

citrate (TAe), a hydroxyl-containing trifunctional monomer which

polymerizes about ten-fold faster than DAP or DAA. We have deve-
loped a DAP/DAA (80/20) resin wherein DAA provided flexibilizing
action. Also developed were DAP/TAC (80/20) syrup. by polymeriza-
tion at 80°C with 0.2 percent benzoyl peroxide, under constant
stirring and with a nitrogen blanket. This syrup resin was cured
by 0.3 percent benzoyl peroxide and 0.3 percent DiCup. DAP/DAA/TAC
(80/10/10) was another promising resin we developed. The curing
cycle for these resins was not optimized.

Optimization of DAP and Allylic Adhesive Systems

The flexibilized DAP and allylic adhesive optimization can
consist of the following components:

° Type and level of flexibilizing monomer.

° Preparation of DAP/Alkyl Methacrylate (or other) syrup.
° Determination of catalyst type (e. g •• benzoyl peroxide
and dicumyl peroxide) and concentration (0.06 to 2.0
percent) and development of suitable curing cycles.
° Water "getter" fillers (e.g •• aluminum and glass powders)
formulation study,
° Flow control agent. adhesion promoters, and other additive
formulation study.

In Figure 1 a schematic of the DAP adhesive chemistry and techno-

logy is presented.

DAP/Epoxy IPN Adhesives

IPNs have been known for many years and have been apparently
"rediscovered" several times. 5 - 7 IPNs are mixtures of two or more
distinct polymeric phases that cannot be separated by conventional
physical resolution techniques. They offer another method of
modifying a polymer~s architectural design, like graft and block
copolymerization. By varying the network composition it is possi-
ble to effect the resulting properties. Optimum properties (maxima
or minima) can be obtained with binary or ternary IPNs.

Most lPNs, on the molecular scale, are heterogeneous. with

phase domains rich in one or more of the polymer components. Most
IPNs. on a colloidal scale, appear to be compatible (i.e., homo-
geneous) blends of the component networks. Thus IPNs are a unique
class of molecular topology that represent different ways of embed-
ding molecules in three-dimensional space. 5- 7 Because permanent
STRUCTURAL ADHESIVES BASED ON DIALL YL PHTHALATE 253

Flexibilizing Comonomers
BMA SMA
DAP OMA HPMA TAC
ERMA DAA
IDMA DAI
LMA

Quick Bottle or Aluminum

Dish Polymerization

Flexibilized DAP Adhesive Resins

DAP/BMA (70/30, 80/20, 90/10, 95/5)
DAP/OMA (70/30, 80/20, 90/10, 95/5)
DAP /EHMA (70/30, 80/20, 90/10, 95/5)
DAP/IDMA (70/30, 80/20, 90/10, 95/5)
DAP/LMA (70/30, 80/20, 90/10, 95/5)

DAP/SMA (70/30, 80/20, 90/10, 95/5)

DAP/DAA (70/30, 80/20, 90/10, 95/5)
DAP/DAI (70/30, 80/20, 90/10, 95/5)
DAP/ERMA/HPMA (70/25/5, 80/15/5, 90/5/5)
DAP/IDMA/HPMA (70/25/5, 80/15/5. 90/5/5)
DAP/LMA/HPMA (70/25/5, 80/15/5, 90/5'/5)
DAP/TAC (80/20)
DAP/TAC/ERMA (75/5/20)
DAP/TAC/IDMA (75/5/20)
DAP/TAC/LMA (75/5/20)

I
Initial Screening: Adhesive Property, Moisture Resistance,
Retention of Mechanical Properties when Exposed to Heat/Humidity

Selected Flexibilized
Adhesive Resins

I
Intensive Testing and Formulation
I
Optimized DAP
Adhesives

Figure 1. Schematic of DAP adhesive optimization program.

254 A.M. USMANI

entanglements will restrict motions of segments, they would be

expected to nearly mimic covalently bonded chemical crosslinks.
We took advantage of these mimicking cross links to develop superior
adhesives.

We have developed flexibilized DAP/polyepoxy type IPNs. The

rate of DAP/LMA or DAP/EHMA/HPMA polymerization by a free radical
initiator is about the same as the rate of an epoxy reacting with
an anhydride. When both mixtures are mixed and allowed to cure,
an IPN was formed. In our work we used an aromatic epoxy with
hexahydrophthalic anhydride. Cycloaliphatic epoxies were less
suitable. DAP/LMA (80/20) with 20, 30, and 40 percent epoxy resin
plus anhydride were studied. However, no property optimization
was done. In Figure 2, a schematic of DAP/Epoxy IPN optimization
is presented.

CONCLUSIONS

Flexibilized DAP adhesive resins were developed that have

potentially superior properties over the currently used epoxy
materials. DAP/EHMA/HPMA (80/15/5) and DAP/LMA (80/20) are good
compositions. DAP/epoxy IPNs were successfully explored. None of
the compositions were optimized but schematics of optimization have
been presented.

Since these materials represent a relative new technology of

potential importance, investigators are encouraged to study these
systems.

ACKNOWLEDGMENTS

The author appreciates help and support of Dr. A.G. Maadhah

and Mr. P. F. Cooke (both of the Research Institute) in prepara-
tion of this chapter. Technical discussions provided by
Drs. J. M. Butler and I. O. Salyer (University of Dayton) is
acknowledged. Special thanks to Nihal Ahmad (Research Institute)
for typing this work.

REFERENCES

1. K. L. Mittal, Pure Appl. Chem., 52, 1295 (1980).

2. J. C. Bolger, in "Treatise on Adhesion and Adhesives",
CR. L. Patrick, Editor), Vol. 3, Chap. 1, Marcel Dekker, Inc.,
New York, 1973.
3. S. H. Hamid and A. M. Usmani, Polym. Plastics Technol. Eng.
in press.
STRUCTURAL ADHESIVES BASED ON DIALLYL PHTHALATE 255

Flexibilized DAP
Development

. I
Screen1ng Procedure
I
Aromatic Epoxy/ Selected (3 to 5) Cycloaliphatic
Anhydride DAP Systems Epoxy/Anhydride

I I
I IPN-I I IPN-II ~_--..lT
L IPN Composition/
Property Study
r
Property Mapping at 0, 25, 50, 75 and 100%
Flexibilized DAP (for maxima and minima)
Water absorption
A typical mechanical property
A typical mechanical property as a
function of high heat and high
humidity
A surface property, e.g., critical
surface tension
Special emphasis will be placed on
morphology. Morphology of IPN
depends on curing cycle. Thus,
besides composition, curing
cycles should also be studied.
0% '-------------,-----------.! 100%
Flexibilized DAP in IPN Composition +

Selected IPN
Compositions

Intensive Testing and Formulation

-I
Optimum IPN
Adhesives

Figure 2. Schematic of IPN optimization.

256 A.M. USMANI

4. A. M. Usmani, J. Mater. Sci. Letters, l, 75 (1982).

5. D. Klempner, T. K. Kwei, M. Mutsuo, and H. L. Frisch, Polym.
Eng. Sci., lQ, 327 (1970).
6. L. H. Sperling and D. W. Friedman, J. Polym. Sci., A-2, 1, 425
(1969).
7. A. M. Usmani, J. Elastomers Plastics, 11, 170 (1981).
POSSIBILITIES OF INTEGRATING SURFACE TREATMENT OF BONDING PARTS IN

THE ADHESIVE BONDING PROCESS

Klaus Ruhsland

Central Institute of Welding

Engineering of the GDR
4030 Halle (Saale) GDR

The conventional technique of adhesive bonding of

metals is characterized in most cases by a time-consum-
ing and expensive surface treatment of bonding parts
before applying the adhesive and involves industrial
safety and environmental problems.

The research results of ZIS Halle described here

show that it is possible by means of chemical and/or
physical methods to integrate the hitherto separate
surface treatment, to a great extent, into the bonding
process. It can be done, in principle, by the adhesive
itself and/or the adhesive bonding technique.

257
258 K.RUHSLAND

INTRODUCTION

According to the conventional technique for the adhesive bond-

ing of metals, the surface treatment of the bonding parts is always
done before applying the adhesive in order to clean, roughen and
activate the areas to be bonded. However, the methods that can be
applied for this purpose, depending on the bonding parts, require
much time and expenditure and involve industrial safety and
environmental problems.

On the other hand, there is no doubt that the surface treat-

ment of the bonding parts decisively determines the strength and
stability of an adhesive-bonded metal joint.

Therefore, it has not been the object of the research work

carried out by ZIS Halle within the last few years to call in
question the importance of surface treatment, in principle, but
merely to integrate it as much as possible into the process of
joining with adhesives. It can be accomplished, in principle, by
the adhesive itself or with the help of the adhesive bonding
process.

The methods developed for this purpose, which should be useful

in making a rational choice for the surface treatment of adhesive
bonding of metals shall be presented in the following.

STATE OF DEVELOPMENT OF TECHNOLOGY

At the beginning of the research work at ZIS Halle it was

merely known that a few groups of adhesives, due to their chemical
composition and/or their setting and hardening conditions, were
able to dissolve and to absorb the oil or grease films existing on
the surface of the bonding parts.

Thus, for example, it has already been pointed out l that

solvent-containing adhesives exhibit oil and grease-dissolving
properties. This is especially ev~dent when PVC plastisols are
applied on oily material surfaces. Here, the plasticizer present
in the PVC plastisol can, in principle, also be considered as a
solvent, and the grease-dissolving ability of these adhesives
is intensified by the high setting temperature required.

For the latter reason, a better oil and grease compatibility

can undoubtedly be found also with hot-setting epoxide resin
adhesives, hot-setting thermo-plastic hot-melt adhesives, and also
with cold-setting3~goxide resin adhesives, if they are hardened at
high temperatures • However, as our own investigations have
shown, a high setting or hardening temperature alone is not
sufficient to make high-grade adhesive-bonded joints, because their
water resistance is poor.
INTEGRATING SURFACE TREATMENT OF BONDING PARTS 259

Cold-setting unmodified epoxide resin adhesives are especially

sensitive to an imperfect surface treatment, particularly to
impurities caused by oil or greases. The loss in stability compared
to an optimum surface treatment amounts to approximately 50 to 100%
(depending on the type of adhesive and oil) for this group of
adhesives at a degree of greasing of X = 24. A degree of greasing,
X, of 10 corresponds approximately to 1 ~m thickness, so X = 24
signifies about 2.4 ~m layer of grease.

It has been the object of the joint research work of ZIS Halle
and VEB ASOL-Chemie Berlin to decisively improve the efficiency of
a cold-setting epoxide resin adhesive on greased surfaces by
chemical modification. In addition, how far the cleaning effect can
be increased by physical methods has also been investigated.

MODIFIED EPOXIDE RESIN ADHESIVE

The adhesive developed (trade name: Epasol FV/ZIS 939) is a

modified cold-setting two-component adhesive on the basis of
epoxide resin which, because of its composition (GDR Patent WP C 09
j/123672), is able to satisfactorily wet oily surfaces, to dissolve
and to absorb oil t~~ms already present and to activate the joint
component surfaces •

Its efficiency on an oily Al Mg 3 alloy compared to an unmodi-

fied epoxide resin adhesive can be seen from Figure 1. Apart from
the oils mentioned in Figure 1, its excellent adhesion has also
been shown on the following greasing agents: ASTM test oil No.1,
engine oil, machine oil, hydraulic oil, drilling oil emulsion,
silicone spray 200 (silicone separating agent), diesel fuel and
Siliron (industrial soap). Its absorbing capacity for greases is
small; however, it can be considerably improved by hardening at
high temperature,or a "preliminary cleaning" of the bonding sur-
faces with wash or flushing oils, which means a coarse cleaning
with oils of low viscosity.

The adhesive Epasol FV/ZIS 939 is standardized and is con-

tinuously being controlled on greased surfaces and on surfaces
pretreated in an optimum way, e.g., pickling process for aluminum.
The result of the quality control of the first 10 production
batches is given in Table I showing that the difference in strength
varies from 0 to 20%. The standard deviations are almost the same
and the coefficient of variation,is in all cases,below ± 10%.

In case of hot setting of the adhesive the difference in

strength between optimum surface treatment and oily surface is
almost negligible or is within the measurement errors (see Figure
260 K. RUHSLAND

f
!N/mmZ]

30
~ 1

---
Z
20 3

.../..B
10
3

6 12 18 2* "[Xl
Degree of greasing
Figure 1. Tensile shear strength at 20°C as a function of greasing
agent and the degree of greasing.

A = Epasol FV/ZIS 939

B = unmodified epoxide resin adhesive
Hardening conditions: 24 h at 20°C + 2 h at 90°C
v = greased
x = degree of greasing (X - 10 ~ 1 ~m grease layer thickness
on the bonding parts

Greasing agent =
1 paraffin oil;
2 silicone oil NM 4-500
3 gear oil GL 125.

2), the mechanical and thermal properties of the adhesive-bonded

joints are distinctly improved (See Figure 3) and the already very
good water resistance (see Figure 4) is further improved (see
Figure 5).

The adhesive Epasol FV/ZIS 939 can be applied Sor adhesive

bonding on oily surfcces of the following materials :

aluminum and aluminum alloys, structural steels and Cr-Ni

steels, grey cast iron, glass, ceramics, concrete, asbestos
cement, glass-fiber reinforced polyesters, Sprelacart,
INTEGRATING SURFACE TREATMENT OF BONDING PARTS 261

Table I. Tensile Sheer Strength as a Function of Surface Treatment

and Batch Number.

f t s(N/mmzl LIt
Batch-No. e+g v (%1
1 27,1t 0,6 21,9tO,8 -19,2
2 26,5tO,5 21,8t 1,1 -17,7
3 25,4t1,1 25,7t 1,1 + 1,2
4 27,Ot 0,7 24,8t 1,3 - 8,1
5 23,6t1,O 21.6:1: 0,9 - 8,5
6 23.7:1: 0.7 22,0:1:0,9 - 7,2
7 27.9t 0,7 23.6:1: 0,6 -15,4
8 24,Ot 1,0 24,5:!: 1,1 + 2,1
9 23,6t 0,5 23,Ot 1,7 - 2,5
10 27.0 ± 0,8 21,8 t 1,1 -19,3

Adhesive = Epasol FV/ZIS 939

Hardening conditions = 24 h at 20°C + 2 h at 90°C
e + g = degreased and pickled for 30 min. at 60°C in a
pickling bath containing 65% distilled water, 27.5% H2S0 4
and 75% Na 2Cr 20 7 .2H 20.

Greasing agent = gear oil GL 125.

pressboard, phenoplasts and aminoplastics, clutch facing,

wood, Ecotal sheet, unplasticized PVC, acrylonitrile-
butadiene styrene, polystyrene and polyurethane foamed
plastics together or combinations of any of these materials.

The maintenance sector, the building industry, shipbuilding,

machine and apparatus construction and structural adhesive bonding
in various branches of industry are the main fields of application
of this adhesive. Apart from this, it can also be applied as a
casting and laminating resin, as a coating and as an adhesion
primer.
262 K. RUHSLAND

Hardening conditions: 1h at 160°C

20 0~
7:
" -
j~ ~
10
;~
~ t'~
-...
e+g e 6 12 18 24 v[Xl
Degree of greasing

Figure 2. Tensile shear strengh at 20°C as a function of surface

treatment.

Adhesive = Epasol FV/ZIS 939

e + g = degreased and pickled
e degreased
v greased
X degree of greasing
Greasing agent = gear oil GL 125

VIBRATIONAL ADHESIVE BONDING

Another possibility for joining contaminated materials by

adhesives without prior surface treatmI~t is provided with the
so-called vibrational adhesive bonding (GDR Patent WP C 09
j/120042).

The term "vibrational adhesive bonding" signifies an adhesive

bonding method according to which a "hard" filler (added to the
adhesive) in combination with a mechanical relative movement of the
bonding parts (e.g. by means of ultrasonics) causes cleaning and
roughening of the bonding surfaces after the application of the
adhesive and joining. To do this, the hardness of the filler must
be greater than that of the bonding material or its surface
contaminants.
INTEGRATING SURFACE TREATMENT OF BONDING PARTS 263

f
[N/mm 2] Hardening conditions: 1~24h at 20"( + 2h at 90·e
2 ~ lh at 16Q·C

30

20

~~
10
~~ "
20 50 80 T{°Cl

Figure 3. Tensile shear strength as a function of test temperature

and hardening conditions.
Adhesive = Epasol FV/ZIS 939; Surface treatment = greased (X = 24);
Greasing agent = gear oil GL 125.

T
[Nlmm'] Hardening conditions: 24h ct 20°C lh at go·e
30

20

10
-
~ ./etg
--
'V

2 4 6 12
Storage in distilled water, months

Figure 4. Tensile shear strength at 20° C after storage in dis-

tilled water as a function of surface treatment.
Adhesive = Epasol FV/ZIS 939; e + g = degreased and pickled;
v = greased (X = 24); Greasing agent = gear oil GL 125.
264 K. RUHSLAND

't
[Nlmml ) Hardening conditions: lh at 160°C

30~-4---4---+---+--~--~

10~--~--r---r---r---r-~

2 4 6 12
Storclge in distilled water, months

Figure 5. Tensile shear strength at 20°C after storage in distilled

water as a function of surface treatment. For explanations, see
Figure 4.

The investigations made by ZIS Halle have shown that already

after 2 seconds of ultrasonic treatment maximum stability data for
the adhesive bonds on oily surfaces are attained which exceed those
of the initial adhesive using conventional surface treatment of the
bonding parts by approximately 45% (curve 2) and 75% (curve 3)
respectively (See Figure 6). Compared to the efficiency of the
initial adhesive on an oily surface (X = 24) this means an increase
in stability of about 400 or 500%.

As can be seen in Figure 7, the waster resistance of such

adhesive bonds is also much improved.

Considering the present capacity of the ultrasonic welding

equipment, the application of the technique of vibrational adhesive
bonding is first of all possible for adhesive bonding of 2small and
medium structural members (bonded area up to about 10 cm ) in mass
production, e.g., electrical engineering/electronics, precision
mechanics/optics, precision instrument engineering, the toy
industry, the jewelry trade and similar.

Making of longitudinal seams by means of continuously working

ultrasonic equipment also seems possible.
INTEGRATING SURFACE TREATMENT OF BONDING PARTS 265

rmings
[N/mm'1 '3
15 2

'"
10 17-:
~~
5
t/ t':
'1 f--
~~
et9 e 6 12 18 2~ v[X]
Jegree 01 c;redsing

Figure 6. Tensile shear strength at 20°C as a function of surface

treatment.

Adhesives: 1 unmodified epoxide resin

2 1 + 40% filler
3 2 + 8 silane adhesion promoter

Hardening conditions: 24 h at 20°C + 2 h at 90°C

e + g = degreased and pickled
e = degreased
v = greased

Degree of greasing X 24
Greasing agent gear oil GL 125
Ultrasonic treatment 2 s (for 2 and 3)
t min 95 t - 2 s (mean value - 2x standard
deviation

fmings
[N/mm'1 ~
15 ~ 16
~

10 ~ ~
~
5 ~~

~~ Z 2 3 3
e+g v v
Figure 7. Tensile shear strength at 20°C before and after 60 days
of storage in distilled water as a function of surface treatment.
a, before storage; b, after storage. For explanations, see Figure 6.
266 K. RUHSLAND

REFERENCES

1. H. Baumann, "Glues and Contact Adhesives", Springer Verlag,

Berlin, 1967.
2. G. Gierenz, "Adhesive Bonding of Thin Sheet-metal Structures
with PVC Plastisols", Notice of Deutsche Forschungsgesell-
schaft fur Blechverarbeitung und Oberflachenbehandlung E.V.,
.!2., No.7, 95-98 (1968).
3. "Rational Joining with Adhesives", Publication of 3 M Co., St.
Paul, MN, USA.
4. K. Ruhsland and B. Winkler, "Process for Making High-grade
Adhesive Bonds without Surface Treatment of Bonding Parts",
GDR Patent WP 112 658; C 09 j, 5102.
5. S. Semerdjiev and P. Panov, "On Adhesive Bonding of Metal
Parts on Unclean Surfaces", Paper presented at the meeting for
experts for adhesive and casting resin technology, Halle, GDR,
1969.
6. K. Ruhsland and B. Winkler, "Adhesive Bonding of Metals with-
out Surface Treatment of Bonding Parts", Adhasion, 1!., No.1,
6-9 (1977).
7. K. Ruhsland, "FV Adhesive Bonding Technique and its Possibili-
ties of Application", Fertigungstechnik und Betreib, 28, No.
5, 296-298 (1978). --
8. K. Ruhsland, "Results of Application-technical Testing of
Epasol Fv/zIS 939", ZIS Notice, 21, No. 10, 1070-1078 (1979).
9. K. Ruhsland, "Comparison of Properties of Oily Metal Adhesive
Bonds", ZIS Notice, 22, No. 10, 1172-1178 (1980).
10. K. Ruhsland, "Vibrational Adhesive Bonding", Adhasion, 23, No.
6, 184-188 (1979). -
 Part III
Evaluation and Characterization
EVALUATION OF ADHESIVE TEST METHODS

G.P. Anderson,* K.L. DeVries** and G. Sharon**

*Morton Thiokol/Wasatch Division

Brigham City, Utah 84302
**University of Utah
Salt Lake City, Utah 84112

Three approaches to bond strength analysis are

discussed in this article: those based upon linear
elastic stresses, those based upon linear fracture
mechanics parameters, and those based upon nonlinear
analyses. The first two approaches are then discussed
with respect to standard adhesion test methods. The
stresses in linear lap shear test specimens are shown
analytically and experimentally to be concentrated at
the bond termination and to be strongly dependent upon
adherend thickness for standard tests. The bond stress
intensity factors and energy release rates near the
bond termination are given and discussed. Adherend
thickness should be at least 4 times larger than that
recommended by standard test techniques. The stresses
in butt tensile tests are also discussed. It is shown
that test specimen alignment problems can lead to low
debond forces and excessively large data scatter.
Stress intensity factors are also determined for butt
tensile tests.

INTRODUCTION

Most designs require that component parts be joined. Adhesive

joining has several advantages over mechanical connections (e.g.,
bolts, rivets, screws, etc.). However, some factors have tended to

269
270 G.P. ANDERSON ET AL.

limit wide usage of adhesives. These include problems associated

with: (1) predicting joint strength, (2) identifying and detecting
factors that affect the quality and reliability of an adhesive bond,
(3) evaluating strength of an adhesive, and (4) determining how the
parameters obtained from the measurements in (3) can be used to
design an "optimum joint." These problems are largely mechanical in
nature and have been the subject of extensive testing and analysis.
Testing is important in all aspects of material science and engi-
neering but in no area is it more important than with adhesives.

A large number of adhesive tests have been proposed and used.

These might be categorized into two rather general classifications.
In the first of these, which we will call "average stress criteria,"
there is an implied assumption that failure is controlled by the
magnitude of the stress and tests are devised to measure this stress
(generally an average) at which failure occurs. The other classifi-
cation of tests makes use of the concepts of fracture mechanics to
evaluate the quality of adhesives and to design bonded joints.
Fracture mechanics puts special emphasis on the presence of stress
risers in nucleating failure. It is hypothesized that failure is
more dependent on the magnitude of local stresses and energy dissi-
pation mechanisms than on an "average stress" in the bond.

Most standard tests (by ASTM or other standardizing groups)

fall in the first of these categories. However, as we will show
later, these standard tests can be used to obtain fracture mechanics
properties. In addition, there has recently been a method adapted
(ASTM D3433) based on fracture mechanics and another (ASTM D3762)
that has some fracture mechanics ramifications. Standard tests
might be further divided into three groups: (1) tensile tests, e.g.,
ASTM D897; (2) shear tests, e.g., ASTM D1002 and ASTM D316S; and (3)
peel tests, e.g., ASTM D1876. Experimental results for the first
two of these are usually given as the force per unit area. The
stresses for such samples are very nonuniform and this average
stress mayor may not be closely related to the failure stress in a
practical design. In fact, while the failure stress given for lap
shear is usually the average shear stress, failure may often be more
closely related to the cleavage stresses that generally occur near
the ends of the overlap.

This paper contains numerical and analytical analyses of some

features of stresses and stress intensity factors in adhesive
joints. Particular emphasis is placed on lap shear joints and to a
less extent on a butt tensile joint. Experimental evidence for the
existence of complex stress states in such joints is also included.
Peel tests were discussed in earlier papers by the authors. 1
EVALUATION OF ADHESIVE TEST METHODS 271

FACTORS AFFECTING ADHESION

The geometries for the adhesive butt joint and lap shear test
configurations appear deceptively simple. Superficially, one might
anticipate that the calculation of stress (or energy required to
produce a unit of new surface) would be comparatively straight for-
ward. Such is not the case. Even leaving aside for the moment
intricacies of molecular structure and the singularities at
imperfections or cracks, one finds that the differences in moduli
between the adhesives and adherends, the physical constraints they
place on each other, the effects of free edges, and the influence of
minute misalignment can result in extremely complex stress states.
When we add to these largely geometric effects the influence of
macroscopic and microscopic material properties, we find that inter-
preting test results in too simplified a manner can be more mislead-
ing than enlightening. This paper treats several aspects relating
to these factors.

Adhesive thickness has been shown by a number of investiga-

tors 1 ,2,3 to have a pronounced effect on joint "strength." The
effect (even its sign) is found to be a function of the joint geome-
try; e.g., for a given adhesive, a butt joint's strength might be
found to decrease with increasing thickness, while the strength of a
lap shear joint increases to a maximum as the thickness is in-
creased. These effects and some of the possible causes are reviewed
in Reference 1. It would be too lengthy to review them here. Suf-
fice it to note that the adhesive thickness effect may be due to any
or a combination of a number of geometric factors and material prop-
erties. As a case in point, the energy required to create a unit of
debonded surface is generally much larger than that required to
rupture a "plane" of bonds. One reason for this lies in the demon-
stratable fact that in all except for perfectly elastic systems,
energy is dissipated at distances well removed from the crack tip.
In some cases, the thickness effect is closely related to the size
of this damage zone near the crack tip. If the adhesive is not
thick enough for this zone to fully develop, less energy is dissi-
pated with an associated decrease in the adhesive fracture energy.

Alternatively, the relative magnitude of the cleavage and shear

stress at or near interfaces depends not only upon the difference in
moduli and Poisson's ratios for the adhesive and adherend but also
on the exact joint configuration. While they have not been as
extensively studied, adherend thickness, stiffness, etc., can also
have a pronounced effect on bond "strength."

LAP SHEAR TEST, BACKGROUND

There has been no dearth of research in an effort to analyze

the stresses in adhesive joints. As early as 1938 Volkersen 4
272 G.P. ANDERSON ET AL.

obtained expressions for the stresses in a lap shear joint by con-

sidering the differential displacements of the adherends and neg-
lecting bending. This was followed in 1944 by the now classical
treatment of Goland and Reissner. 5 They used standard beam the-
ory and other strength of material concepts to obtain expressions
for the joint stresses. Plantema 6 combined the results of these
two earlier investigations to include shear effects in the system.

Since the stress state of the lap shear joint is so complex and
doesn't lend itself to closed form solutions, it is only logical
that as numerical methods become available, researchers would apply
these to analyze adhesive joints.

Wooley and Carver,7 for example, used finite element methods

to calculate the joint stresses. They compared their results with
the results obtained by Goland and Reissner and reported very good
agreement. Adams and Peppiatt 8 used a two dimensional finite
element code to analyze the stresses in a standard lap shear joint.
They too reported good agreement with Goland and Reissner. These
authors also investigated the effect of a spew (triangular adhesive
fillet) on the calculated stresses. A nonlinear finite element
analysis of the single lap joint was completed by Cooper and Sawyer 9
in 1979. They too compared their results with Go 1and and Reissner
as well as with experimental results from photoe1astic models. They
concluded that while nonlinear behavior can at times have a sizable
effect on joint stresses, the Goland and Reissner method is usually
sufficiently accurate for prediction of midsurface stresses.

The double lap shear joint has the advantage of greatly reduc-
ing the bending stresses. This joint has also received analytical
attention. Amijima, Yoshida, and Fujii 10 used finite element
methods to calculate the stresses. They were able to claim some
agreement between predictions of joint stresses obtained in their
numerical analysis and experimentally observed strength.

Sen and Jones 11 ,12 also used finite element techniques for
the analysis of double lap joints. They considered a viscoelastic
adhesive. They report that the magnitude of the maximum adhesive
stress decreases as the ratio of the adhesive modulus to adherend
modulus decreases from one. They also investigated the effect of
adhesive thickness and length of overlap on the stress distribution
along the joint. They report that decreasing the thickness results
in a more uniform stress distribution.

The measurement of the exact stress state in an adhesive joint

is also a difficult task. There have, however 1 been several inten-
sive efforts in this respect. Sharp and Muhal~ used a laser
technique to measure the adhesive shear strain. The technique is
EVALUATION OF ADHESIVE TEST METHODS 273

only suitable for optically clear linearly elastic adhesives. They

reported stresses in good agreement with Goland and Reissner.

Several investigators have suggested modifications to the

standard ASTM lap shear test geometries. Leaving an adhesive s~ew
and rounding adherend corners as analyzed by Adams and Peppiatt
(noted previously) can eliminate the mathematical singularity at
the end of the overlap thereby reducing the maximum stress, in-
creasing the joint strength, and reducing test variability.
Renton 14 suggested thickening the adherends, over those recom-
mended in ASTM D3165, thereby reducing bending and hence cleavage
stresses in the lap joint. He describes a technique of determining
adhesive strain by measuring changes in a magnetic field resulting
from the displacement of the steel parts of the lap joint and sug-
gests optimum specimen designs for various adhesive-adherend com-
binations. In the most thorough study of this type with which the
authors are familiar, Guess, Allred and Gerstle 15 present finite
element and experimental evidence that failure in standard lap
shear specimens is often controlled by the tensile strength of the
adhesive rather than its shear strength. As will be discussed in
more detail later, they report an apparent increase in shear
strength with increasing adherend thickness. River 16 also de-
scribes methods, and test results, for reducing the stress con-
centration at the end of the overlap and obtaining relatively ac-
curate measurement of the shear properties of an adhesive.

Because of the singularity and other problems associated with

stress criteria of failure, several groups have recently applied
the principle of fracture mechanics to adhesive jOints. 1 ,3,17,18
Of particular interest to the lap shear joint is the attempt by
DeVries, et al. 19 to apply a Griffith type energy balance. They
were able to claim reasonable agreement with experimental data and
noted an apparent dependency of the adhesive fracture energy on
mode of stress at the crack tip.

LAP SHEAR TEST, STRESS ANALYSES AND TEST DATA

For practical reasons, the designer usually attempts to load

adhesives in shear. As a consequence, the most commonly used
adhesive test is the lap shear test. Specifications published by
ASTM and other sources describe a variety of different tests but
the most commonly used lap shear bond strength tests are ASTM D1002
and ASTM D3165. The adherend thickness recommended in these
standards for most metal sheets is 0.16 cm (0.0625 in.) with 1.3 cm
(0.5 in.) overlap. Generally the quantity reported as the "lap
shear strength," or "tensile shear strength," is the breaking load
per unit bond area. However, failure initiation in materials
is usually a localized phenomenon that is more dependent
on maximum stresses (or energy) at a point reaching
274 G.P. ANDERSON ET AL.

some critical value than on the average induced values. As men-

tioned previously and reinforced by plane strain finite element
stress analyses in this study, the standard lap shear specimen shows
a variation in shear stress within the adhesive layer and in fact
can exhibit a singular behavior at the bond termination.

The authors have recently completed linear elastic stress

analyses of lap shear joints using a plane strain finite element
computer program. These analyses have particularly concentrated on
the stresses near the bond termination. Lap shear specimens with
ASTM D3165 geometry, adherend thicknesses of 0.16 cm (0.0625 in.)
to 1.3 cm (0.50 in.) and adhesive thicknesses of 0.025 cm (0.01
in.) and 0.16 cm (0.06 in.) were analyzed. The specimens were 16.5
cm (6.5 in.) long with 1.3 cm (0.5 in.) overlap. The plane strain
approximation is justified since Adams and Peppiatt 20 have
shown that the tensile and shear stresses in the direction of the
applied load are not significantly influenced by transverse
stresses caused by Poisson's ratio strains in the adherends.

The finite element code used quadrilateral elements which were

internally divided into four linear displacement triangular ele-
ments. A grid network consisting of 2,016 quadrilateral elements
was used in each analysis with elements concentrated near the adhe-
sive-adherend interface termination as shown in Figure 1. The smal-
lest element width was 0.000025 em (0.00001 in.) for stress and
stress intensity factor analyses and 0.048 em (0.019 in.) for energy
release rate analyses. Axial displacements were applied and trans-
verse displacements set equal to zero at the specimen ends to simu-
B
A I

'. II_mIn_lift".' mIIf: ~,HI '''.HI .11IJUttlJ1tR IlIlll• •Z

ADHEREND THICKNESS = 1.6 MM (0. 0625 IN. )

.ii.......
A B
- ~----;-! -:- 1-- ,------;--,--1 . -

. t r-i---t --+-+--+- ---+ t-+---r+--+--+-----I----.-~--;.--I.~ . ~.---,

~"-ii-~'.I~''.1 ~_~.~Ti;';TT, hi iiriTtiiiiWH.,.#iiiif$fii±ifrf

-t+-+++~-+-+~~-++-- ", r4t·~~ttt-;"T; j-++<-t-+-++-:--t -+t-++--t-l_r-t~--r-+-t+++--:+~~
',Hj.tltt-C.I-t)j-· H·-~t Hj--Tt-t+-tt-+----<-H-t--~"+t+-t--d:f--T-I-~
, j-I-+ + ~+-~+-+-+-t-+ ++-t-r1 --+--<-+-+-_ ...1-+-+-++-:
L-----'-_ .•.. 1 ~ __ L_L_~ L-'-~-'---.l ___ "--_ ~_~-" __ -i ___ L_'_-L_1

ADHEREND THICKNESS = 12.7 MM (0.50 IN.)

BOUNDARY CONDITIONS ALONG FACES A AND BARE

II LATERAL DISPLACEMENTS = 0
2) LONGITUDINAL DIS PLACEMENTS = 0.076 MM (0. 003IN.)
LENGTH = 165 MM (6.5 IN.) OVERLAP = 12.7 MM (0.50 IN.)

Figure 1. Lap shear deformed grids,

displacements magnified ten times.
EVALUATION OF ADHESIVE TEST METHODS 275

late clamped ends. The following material properties were used:

Modulus, MPa (ksi) Poisson's Ratio

Adherend (Steel) 207,000 (30,000) 0.30

Adhesive (Epoxy) 2,760 (400) 0.34

The adhesive ends were terminated at right angles to the adher-

ends. Often a fillet is formed at bond edges by adhesive squeeze-
out. However, the right angle geometry is obtained when the lap
shear specimens are cut and notched from bonded plates as described
in ASTM 03165. The effect of adhesive squeeze-out is discussed in
References 8 and 21.

The computed deformed shapes for lap shear specimens with

adherend thicknesses of 0.16 cm (0.062 in.) and 1.3 cm (0.50 in.)
are presented in Figure 1. An axial elongation of 0.016 cm (0.006
in.) was applied and the resulting displacements multiplied by a
factor of 10 to obtain the deformed grids. The amplified displace-
ments provide a visual indication of the bending which takes place
near the bond termination for thin adherend specimens. One would
thus expect relatively large bond stresses (cleavage) at these
points induced by adherend bending in thin adherend specimens. In
addition, for linear elastic stress analyses, both the normal and
shear stresses become singular as the bond termination is ap-
proached. 22 ,23
2.0
A~HERENDI THICKN~SS
I
I
I
I: "-"- 1.6MM m.0625IN. l
I\ - - - - 3. 1SMMIO.125IN.1
I\ ------ --- 6.35 MM 10.250 IN. I
,I- 9.52 MM 10.375 IN. I
,.., I \ .................. 12. 7 MM m. 5OIN .l
... >< \ \
v; 1.5
V>
.....
ex:
1\ \
\
l-
V> I '.
ex:
« \ \.
__---'
....
I
V>
\ '\'.
....
""
\ "-

----
N '--- ~
:::;
--- -- -- - --- ----- .....
--
--
1.0
<>:
::;::
r ' -"...................... --- --- --- ---
----

V
ex: I-
0
'.......'
--- -- -
2:
'. '
"- "- -"- "- .. - - "-

0. 5
0.00 0. 05 0. 10 0. 15 0. 20 0. 25 0.30 0. 35 0.40 0.45 0. 50
x/a

Figure 2. Bondline shear stresses in lap shear specimens.

276 G.P. ANDERSON ET AL.

The adhesive stresses adjacent to the nonterminating adherends

are plotted in Figures 2 and 3 as a function of distance from the
bond termination. The distances were normalized by dividing by the
bond length, a. and the st~esses by dividing by the average applied
shear stress, T = pIA. Adhesive thickness was 0.025 cm (0.01 in.).

4 I I I
~ I

_.. -
.
ADHEREND THICKNESS
;>- \\
II--
.. L6 MM (0.0625 IN. ) -
vi
3 ---- 3. 18 MM 10.125 IN.)
\ '"
\.
V>
UJ ------- 6.35 MM 10.250 IN. )
cr:
l- ,\ 9.52 MM 10.375 IN.) -

,
V>
UJ 2 ....... .... ..... 12.7MMIO.50IN.)
==
V>
Z
.....
I-
\\~.. r"-..
0
.....
N r--- .;..
:::;
« ~,-- ~~
::E
oe:
0
Z
0

-I
'- --
"
.. 1
- -- ---- -- -- --
-1--..- -_ - --
.. 1---.. - -- -
"--.
I--
-"-" -
.. . . . . . . .. ..

O. 00 O. 05 O. 10 0. 15 0.20 0. 25 0.30 0.35 0.40 0.45 0. 50

x/a

Figure 3. Bondline tensile stresses in lap shear specimens.

As expected, the stresses change radically as the bond termina-

tion is approached. The shear and tensile stresses are nearly con-
stant and equal to the average values over 80% of the bond area for
adherend thicknesses greater than 0.64 em (0.25 in.). For thin
adherends, the stresses vary continuously along the bondline.

The character of the stress singularity was examined by plot-

ting the tensile and shear stresses in log-log form (Figure 4). The
tensile and shear stresses exhibit a l/rn singularity behavior
(log linear) for all elements within 0.0076 em (0.003 in.) of the
adhesive edge except for the final element. This element typically
varies somewhat from the singularity pattern l due to the
coarseness of the grid network relative to the stress changes in
this area. As the distance from the bond termination increases
beyond 0.076 em (0.003 in.), the stress behavior deviates from the
singular (log linear) form as expected.

Stress intensity factors and the singularity exponents are

obtained from the straight line portion of the log-log plots. The
"singularity exponent," n, in the equation, stress = K/r n , is the
slope of the straight line portion of the Figure 4 curves. This
value was approximately 0.34 for tensile as well as shear stresses
and does not depend upon adherend thickness. The stress intensity
EVALUATION OF ADHESIVE TEST METHODS 277

factors (intercepts of the Figure 4 curves), however, were found to

depend upon adherend thickness as 8110.",n til Tabl e I.

Table 1. Stress Intensity Factors for Lap Shear Tests

Shear Tensil e
Adherend KniT' Krtl
Thickness
cm (in.) cm O• 34 (in. 0 • 34 ) Kn/Kn 0.5 cm O• 34 (in. 0 • 34 ) Kr/Kr 0.5 Kr/Krr

0.16 (0.0625) 0.448 (0.326) 2.59 0.914 (0.666) 1.77 2.04

0.32 (0.125) 0.272 (0. 198) 1.57 0.660 (0.481) 1.28 2.43

0.64 (0.250) 0.189 (0.138) 1.10 0.560 (0.408) 1.09 2.96

0.95 (0.375) 0.178 (0.130) 1.03 0.526 (0.383) 1.02 2.95

1.27 (0.500) 0.173 (0.126) 1.00 0.516 (0.376) 1.00 2.98

LEGEND
<> TENS lLE STRES S
'"
n.. oS HEAR STRESS
~6~--------~--~~~~~------~------------------+-------~
'"
'"....
0::
>-
'"g<!>

5 ~-------4---------------+----------~~~~----~

-4 -3 -2
r. em
LOG
Figure 4. Log stress plot, lap shear test.

Roth the Mode I stress intensity factor, KI' and the Mode II
stress intensity factor, KII' per unit average shear stress, T,
increase with decreasing adherend thickness, see Figure 5. Thus
one would expect higher debond forces in thick adherend specimens.
The decrease in KII with increasing adherend thickness is
greater than that for KI. For all adherend thicknesses, the Mode
I stress intensity factor is greater than that for Mode II. As
shown in the last column of Table I, the tensile component is near-
ly three times the shear component for large adherend thicknesses.
This tendency was not expected but was consistently observed for
widely varying finite element grid networks and for either plane
strain or pl·ane stress conditions.
278 G.P. ANDERSON ET AL.

IN.
0.1 0.2 0.3 0.4 0.5
0.7

0.6

!!::
;Z
0.5
'"
0
>-
:::
u

>-
~
V'> 0.4
~
~
V'>

~
>-
0.3
V'>

~
N
...::E
::l
0.2
KII/i'
'"0z
0.1

0.2 0.4 0.6 0.8 1.0 1.2

ADHEREND THICKNESS, em

Figure 5. Lap shear stress intensity factors.

The nature of the singularity near the termination of a two

material interface has been investigated by several authors. 22 ,23
For linearly elastic materials, the stresses sufficiently near the
bond termination are shown to be proportional to r-n , where r
is the distance from the bond termination and the exponent n is a
function of boundary conditions, material properties, and bond
termination geometry. If one material is rigid and a wedge of a
second material with included angle S is bonded to the rigid mate-
rial, the exponent, n (from Reference 23), can be determined from
the plot in Figure 6. For a 0.34 Poisson's ratio and a 90 degree
bond termination angle, n = 0.32. Thus the stresses near the bond
termination are singular and proportional to l/r O• 32 • This is
in essential agreement with the Reference 22 work and the 0.34
exponent obtained numerically during the present analyses.

For some combinations of Poisson's ratio and bond termination

angles, the characteristic equation in the analytical solution has
imaginary as well as real roots. 1 The imaginary roots lead to
stresses which oscillate as the bond termination is approached.

An energy release rate, G, was determined for each of nile

different adherend thicknesses using debond depths of 0.048 cm
(0.019 in.) and 0.10 cm (0.038 in.) (average x/a = O.O~7). As
shown in Figure 7, the square root of energy release rate is •
essentially constant for adherend thicknesse8 great'~r than 0.64 cm
EVALUATION OF ADHESIVE TEST METHODS 279

-0. 1

-0. 2 ~-+--\\--k-+-+----1I---"-~<--"k:-t-

-0. 31---I---+-~h-++~d--jI--+----f"-.~r--1

~50 Q~ QU Qn Q~ QTI QH QC
POISSON'S RATIO, v

Figure 6. Real roots of the charact

a free-fixed wedge
Z3 istic equation for
280 G.P. ANDERSON ET AL.

IN.
0.1 0.2 0.3 0.4 0.5
4r---~~--~-----r----~--~

o~--~--~~--~~--~--~~~~~
0.2 0.4 0.6 O. 8 1. 0 1.4
ADHEREND THICKNESS, em

Figure 7. Effect of adherend thickness on energy release rate,

G for lap shear tests.
EVALUATION OF ADHESIVE TEST METHODS 281

(0.25 in.) and approKimately 1/3 this value at the ASTM recommended
thickness, 0.16 cm (O.062~ in.). Thus the change in energy release
rate with adherend thickness for a small initial debond has the
same basic characteristics as observed for stress intensity
factors, i.e., nearly a constant value for adherend thicknesses
greater than 0.64 cm (0.25 in.) and increasing by a factor of
approximately 3 as adherend thickness is reduced to 0.16 cm (0.0625
in.).

An analysis for a 0.025 cm (0.01 in.) thick adhesive (1.3 cm,

0.5 in. adherend) showed stresses of basically the same form as for
thicker adhesives. However, the stress singularity is shifted
closer to the bond termination.

Testing was completed by the authors using single lap shear

specimens. Steel adherends with thicknesses ranging from 0.16 cm
(0.0625 in.) to 1.3 cm (0.5 in.) were used with a 1.3 em (0.5 in.)
overlap. The adhesive was a fairly rigid two part epoxy (Hysol EA
934). All testing was completed at 72°F and 0.13 cm/min (0.05
in./min) displacement rate. The resulting data are presented in
Table II and Figure 8. These data show a fairly constant breaking
force for specimens with adherend thickness greater than 0.64 cm
(0.25 in.). At the ASTM DI002 recommended thickness of 0.16 cm
(0.0625 in.), the breaking force drops by 63 percent.

Similar data were obtained by Guess, et aI. 15 These

authors also used steel lap shear adherends with 1.3. em (0.5 in.)
overlap. Their adhesives were EC-2214-R structural paste adhesive
and FM-I23-5 structural tape adhesive. Again, the breaking force
was essentially independent of adherend thickness for thicknesses
greater than 0.64 cm (0.25 in.). For these adherends, the EC-2214-
R adhesive strength dropped by 64% and the corresponding drop for
the FM-123-5 film adhesive was 23%. The amount of decrease in bond
strength for thin adherend specimens is essentially the same as the
increases in stress intensity factor and square root of energy
release rate as would be expected if these parameters controlled
debond initiation load.

The stress analyses discussed here assumed that the adhesive

as well as the adherends were linear elastic materials. For some
structural adhesives, particularly the elastomer based adhesives
and the rubber modified epoxies, the linearity assumption is
invalid and a nonlinear stress analysis is preferred. Such
analyses are not difficult to complete with various nonlinear
finite element computer programs which are presently available.
However, the material properties required for input to such
programs are not commonly available. Thus most analyses are still
completed using the linearity assumption. This is a major weakness
in current analytical approaches and may provide misleading data
for many bond systems.
282 G.P. ANDERSON ET AL.

Table II. Lap Shear Test Data

Adherend Average Stress
Thickness, at Debond, Reduction
cm (in.) N (psi) (%)
0.16 (0.0625) 10,500 (1,530) 63

0.32 (0.125) 16,800 (2,440) 40

0.64 (0.25) 24,800 (3,600) 12

0.95 (0.375) 29,000 (4,200) 0

1.27 (0.50) 27,600 (4,010) 0

IN.
0.1 0.2 0.3 0.4 0.5
5,000

30
.., 4,000
Q.
~
<'
....
C<:
< 20 3,000
~
Z ""C
::::l v>
....,
C<:

....u
Q.
2,000
C<:

eiO
Q
z
0 1,000
....
m
<:>

0 0
0 0.2 0.4 0.6 0.8 1.0 1.2
ADHEREND THICKNESS. em

Figure 8. Debond force for lap shear tests.

In much of the literature on linear elastic analyses of lap

shear specimens, the researchers have discussed the stress
concentration factor at the bond terminus. When linear elastic
analyses are used and the bond terminus not rounded, the stress
concentration factor is infinity because of the stress singularity
at this point.
EVALUATION OF ADHESIVE TEST METHODS 283

Even in the idealized case where the adherends are rigid and
restricted to remain parallel, the transverse stress has a singular
behavior very near the bond termination. The transverse stress
becomes negligibly small (as expected) and the shear stress remains
relatively constant (rigid and parallel adherends) for distances
greater than 1/10 the adhesive thickness from the bond termination.
Whether stress intensity factors for the initial bond geometry
actually control debond initiation is open to question since
inherent flaws such as bondline voids could exist in the high
stress area that are larger than the entire singularity region.
Thus the most appropriate criteria for debond initiation may be
obtained through the use of an energy release rate at the depth of
the largest inherent flaw. A yet to be satisfactorily resolved
question, is, by what means can this size be best determined.
Microscopic examination, extrapolation from finite flaws, etc. need
further exploration. Such methods as well as the use of stress
failure criteria are briefly discussed in Reference 1. If one
ignores inherent flaws, the problem of bond termination
singularities could be avoided entirely by using properly contoured
bond termination fillets and rounded adherend edges.

TENSILE BOND TESTS

Tensile bond tests are also commonly used for evaluating bond
strength. The tensile strength suggested by ASTM procedure D897 is
the breaking load ger unit bond area. However, the bond stresses
are not uniform 1 ,2 and, in fact, become singular at the bond
edges when analyzed using linear elastic analyses 22 as discus-
sed for lap shear tests.

If the adhesive thickness is not constant or if the grips used

to attach the specimen to a tensile test machine are not perfectly
aligned and rigid, the initial load is applied only at one edge of
the specimen. This has two adverse effects on the resulting bond
strength data: 1) The true failure load is higher than indicated
and 2) since the magnitude of nonalignment is not normally the same
on each sample, the data variability is much greater than for prop-
erly aligned samples.

The tensile test specimen consisting of two steel adherends

bonded together with a 0.16 cm (0.060 in.) thick epoxy (v = 0.34)
was stress analyzed using a finite element computer program. The
maximum principal stress distributions are shown in Figure 9. The
ratio of peak-to-average stress away from the edge singularity and
ratio of the edge stress intensity factor (using Reference 1 tech-
niques) to average applied stress for symmetric and edge loads
are:
284 G.P. ANDERSON ET AL.

6.0

5.0
VI
VI
LLI
~4.0
VI

. /~
!!I
VI

~3.0
Q
LLI
!:::!
EDG./V
~2.0
V
/
0=:
0 I

-
z SYMMETR Ie LOAD
1.0

o
o 0.25 0.5 0.75 1.0
RADIAL DISTANCE/MAXIMUMRADIUS
Figure 9. Stresses in tensile tests.

Symmetric Edge
Load Load

Stress Ratio (away from edge 1.1 3.59

singularity), ala

Stress Intensity Factor 0.291 (0.220) 0.964 (0.729)

Ratio, KIlo, cmO. 30
(in.O. 30 )

Thus whether stresses or stress intensity factors are considered,

the load carrying capability of the bond is reduced by a factor of
3.3 by edge loads. Even with symmetric loading the peak stress is
10% higher (to 100% higher for incompressible materials)1 than
the average value generally reported for such tests. The applica-
tion of a symmetric load is experimentally not an easy task and in
actual structural jQints is usually impractical. Very small sam-
ple, grip, or test machine misalignments can result in substantial-
ly larger stresses than the average. Slight differences in adhe-
sive thickness across the joint can have similar effects. Experi-
ments were conducted in which "standard techniques" were used and
compared with tests where extraordinary care was taken to obtain
EVALUATION OF ADHESIVE TEST METHODS 285

symmetric loading. In addition, a special test fixture grip was

used to reduce the effects of specimen misalignment in tension or
shear. This fixture and additional analyses will be more fully
described in a subsequent paper. However, the fixture basically
makes use of a rubber plate to reduce the effects of both shear and
bending misalignment.

Test results using this fixture and standard"grips (similar to

those recommended in ASTM D897) are shown below:

Load at Coefficient of
Failure, Variation
N (lb) ---~%)
Standard Grips 14,700 (3,310) 36
Modified Grips 25,900 (5,820) 6

These data show a very large decrease in variability as well

as an 80 percent gain in apparent load carrying capability when
most of the effects of edge loading are removed.

CONCLUSIONS

The authors have tried to summarize and discuss some of the

efforts in their laboratories as well as others to reduce uncer-
tainties associated with adhesive bond testing. The numerical
study reported here places particular emphasis on the regions very
close to the bond terminations. It is in this region that failure
is generally thought to originate. Numerical stress analyses dem-
onstrated that at regions near the bond terminus, the normal and
shear stress distributions exhibited the trend toward singular
behavior reported by others. They have, for instance, the same
singularity exponent as determined analytically by Dill, et
al. 23

I t appears fairly certain that for many "standard" lap shear

specimens failure initiation is not as much a function of the re-
ported "average".shear stress as it is the induced cleavage stres-
ses. Experimentally, higher strengths and less variability result
from using thicker specimens thereby reducing bending in the ad-
herends and hence these cleavage stresses.

For butt joints, the stress state is again more complex than
it superficially appears. Unless there is a perfect match of mod-
uli between the adhesive and adherends, shear stress will generally
accompany the tensile stresses in the adhesive. In addition, the
tensile stress along the bondline is not constant. It varies with
distance from the bond center and exhibits a singular behavior as
the bond terminus is approached. Even "more critically important,
286 G.P. ANDERSON ET AL.

however, are any misalignments of grips or testing machine, slight

variation in sample thickness, etc. Where extreme care is taken to
control these factors, it is demonstrated that typically much high-
er strength with less variation is obtained.

The authors are of the opinion that adhesive fracture me-

chanics somewhat along the lines described in References 1, 25, 26,
and 27 has the best potential of resolving many of the problems
suggested in this paper. Reference 27 is of particular interest to
the current authors since the study of this paper introduced them
to the application of fracture mechanics to adhesive bonds. Tests
based on fracture mechanics concepts provide a measure of adhesive
quality in terms of basic material parameters (adhesive fracture
energy or equivalently critical stress intensity factor, moduli of
adhesive and adherends, and induced or inherent flaw size). Per-
haps of even more importance, once these parameters are obtained,
adhesive fracture mechanics points the way in which these param-
eters might be systematically used to design practical joints,
select optimal joint configuration, and predict adhesive bond per-
formance.

ACKNOWLEDGEMENT

Portions of this effort were supported by NSF - Polymer Pro-

gram DMR~79 25390.

REFERENCES

1. G. P. Anderson, S. J. Bennett and K. L. DeVries, "Analysis and

Testing of Adhesive Bonds," Academic Press, New York, 1977.
2. J. J. Bikerman, J. Soc. Chem. Ind. 60T, 23, 1941.
3. S. Mostovoy and E. J. Ripling, J. App!:-Polym. ScL, ~, 661
(1971).
4. o. Volkersen, Luftfahrtforschung, 15, 41 (1938).
5. M. Goland and E. Reissner, J. Appl:-Mechanics, 11, 17 (1944).
6. F. J. Plantema, "De Schuifspanning in eme Lijmnaad," Report
M1181, Nat. Luchtvaartlaboratorium, Amsterdam (1949).
7. G. R. Wooley and D. R. Carver, J. Aircraft, 8, No. 10, 817
(1971). -
8. R. D. Adams and N. A. Peppiatt, J. Strain Analysis, 9, No.3,
185 (1974). -
9. P. S. Cooper and J. W. Sawyer, "A Critical Examination of
Stresses in an Elastic Single Lap Joint," NASA Technical
Report 1507, NASA Scientific and Technical Information Branch
(1979) •
10. S. Amijima, A. Yoshida and T. Fujii, "Proceedings of the
Second International Conference on Composite Materials,"
Toronto, Canada, Metallurgical Society of AIME, 1185 (1978).
11. J. K. Sen and R. M. Jones, AIAA J., ~, No. 10, 1237 (1980).
EVALUATION OF ADHESIVE TEST METHODS 287

12. J. K. Sen and R. M. Jones, AIAA J., 18, No. 11, 1376 (1980).
13. W. N. Sharpe and T. J. Muha, in Proceedings of the Army
Symposium on Solid Mechanics, Army Materials and Mechanics
Research Center, Watertown, Mass. AMMRC-MS-74-8, 23 (1974).
14. W. J. Renton, Experimental Mechanics, 16, No. 11, 409
(1976). -
15. T. R. Guess, R. E. Allred and F. P. Gerstle, J. Testing and
Evaluation, 5, No.2, 84 (1977).
16. B. H. River,-Adhesives Age, 24, No. 12, 30 (1981).
17. A. N. Gent, "Strength of Adhesive Bonds - Plastic and Visco-
elastic Effects," Proceedings - 46th National Colloid Sym.,
Amherst, Mass. (1972).
18. w. D. Bascom and R. L. Cottington, J. Adhesion, 1, 333
(1976). -
19. K. L. DeVries, M. L. Williams and M. D. Chang, Society for
Experimental Stress Analysis preprints, 21 Bridge Square,
Westport, CT, Cleveland Meeting (1972).
20. R. D. Adams and N. A. Peppiatt, J. Strain Analysis, 8, 134
(1973). -
21. A. D. Crocombe and R. D. Adams, J. Adhesion, 13, 141 (1981).
22. V. L. Hein and F. Erdogan, Int. J. Fracture Mechanics, Z, 317
(1971) •
23. E. H. Dill, A. L. Deak and W. F. Schmidt, in "Handbook for the
Engineering Structural Analysis of Solid Propellants," Ch. 5,
Chemical Propulsion Information Agency Publication 214, Ap-
plied Physics Lab, Silver Spring, MD, May 1971.
24. H. Nakayama, K. Tozawa, A. Hirano, and o. Ohkubo, "Effect of
Adhesive Layer Thickness on Strength Characteristics of Adhes-
ive - Bonded Soft Joint Specimen Under Static Tensile and
Push-Pull Fatigue Load Conditions," Proceedings 22nd Japan
Congo on Mater. Res. - Metallic Materials, Society of Mate-
rials Science, Kyoto, Japan (1979).
25. W. D. Bascom, R. L. Cottington and C. O. Timmons, J. Appl.
Polymer Sci. Applied Polymer Symposium~, 165 (1977).
26. R. J. Chang and A. N. Gent, J. Polymer Sci.: Polymer Phys.
Ed., 19, 1619 (1981).
27. M. L.IWilliams, Bulletin Seismological Soc. America
49, No.2, 199 (1959).
A CRITICAL APPRAISAL OF DENTAL ADHESION TESTING

W.J. O'Brien and S.T. Rasmussen

Surface Science Laboratory, Dental Research Institute
The University of Michigan, Ann Arbor, MI 48109 and
School of Dentistry, Case Western Reserve University
Cleveland, OH 44106

The various methods used to investigate dental ad-

hesion are reviewed. Concepts from fracture mechanics
are used to interpret adhesive and cohesive failure of
adhesives used in dentistry and to analyze bond
strength testing techniques. The practical strength of
an adhesive interface is a complex relationship be-
tween the stress distribution, specific adhesive
fracture energy and the size and shape of interfacial
defects. Bond strength tests are not designed to in-
vestigate these separate aspects related to adhesive
failure. The problem of defects in dental adhesive
failure are unique in that the number, size and shape
of defects which result in failure are mainly tech-
nique related. However. the investigation of dental
adhesive failure can generally be characterized as ig-
noring the relationship that specific adhesive frac-
ture energy and defects have on the practical
strength of adhesive systems. Based on theoretical
considerations we propose that dental adhesion test-
ing be reevaluated from a theoretical, experimental
and most significantly from a practical viewpoint.

INTRODUCTION

The importance of adhesion investigation is widely recognized.

However, adhesion is a complicated problem which cannot be thoroughly
investigated by any simple technique. Dental researchers interested
in adhesion have generally looked at: 1) wettabi1ity, 2) surface prep-
aration, 3) fracture path, 4) fracture energy and 5) bond strength.

289
290 W.J. O'BRIEN AND S.T. RASMUSSEN

Wettability is considered important for a fluid material to adhere

to a solid. Additionally, it also measures a fluid's ability to
displace other liquids and gases and to spread over the surface so
as to produce an interface with a minimum of voids. l ,2,3,4 While
wettability is necessary, it is not sufficient for long-term success
in the oral environment. For example, good wetting may be obtained
between tooth structure and a fluid resin, but the restorative resin
may separate during usage due to tensile forces resulting from poly-
merization shrinkage or the humid oral environment. l ,5

Surface preparation is also important to obtaining good adhesion

and has received extensive attention in dentistry. Figure 1 shows
a ceramic-metal crown with adhesive failure. The enameling is
achieved through firing a ceramic-powder paste to the metal. The
paste consists primarily of a feldspar powder with water which forms
a thick slurry capable of holding its shape when applied to the
metal. Surface preparation of the metal casting has been found to
improve the bonding of dental porcelain to the metal substrate.
Surface roughness; preoxidation out-gassing and metal conditioners
have all been extensively investigated for their roles in improving
bonding of porcelain to metal. 2 ,6,7,8

Figure 1. Ceramic-metal crown wich adhesive failure. According

to the classification system of Figure 6 this is a Type I failure.
A CRITICAL APPRAISAL OF DENTAL ADHESION TESTING 291

Similarly, the adhesion of resins and composites to enamel and

dentin have been improved by surface pretreatment with various chem-
icals and by mainte~rnI2 ~~ a dry, clean surface prior to applying
the resin. l ,3,9,lO, " Acid etching improves the mechanical
retention of resins. Acid is applied to tooth enamel so as to pro-
duce a porous surface to which resin is applied as shown in Figure
2. The surface roughness produced by etching also improves the
resins ability to cover the enamel surface. 3 Scanning electron
microscopy (SEM) has proved useful in the investigation of surface
morphologies of tooth surface following acid etching. Coupling a-
gents or liners have also been reported to improve bonding of resins
(composites) to tooth surfaces. 12 ,13,14,15

The fracture path approach to dental adhesive failure has re-

ceived attention in recent years. In metal-porcelain systems the
various types of failure have been specified. 16 The basic premise
is that it does little good to report bond strength if cohesive fail-
ure can be verified by analysis of the fracture path, since the
values obtained reflect the strength of the material in which co-
hesive failure occurred. It has also been noted that specification
of the actual zone in which failure occurs is not a simple matter. 17
Diffusion during firing of the porcelain to metal, results in a
variation in composition at the interface which cannot be clearly
defined into different zones. Fracture-path studies can help iden-
tify the weak areas so that research can be directed at them for
possible improvement. The fracture-path approach to adhesion re-
search for resin-tooth structure systems has also received atten-
tion. lO ,18,19 The fracture of resin-enamel systems is not consid-
ered to be pure interfacial since deeply penetrating resin tags
(Figure 2) are brokenoff. 19

The application of fracture mechanics to dental adhesion has

been limited. The area under stress-vs.-strain curves has been used
to compare various porcelain specimens. 20 However,this approach is lim-
ited in scopy,since the fractured area is not known. A method call-
ed the "blister techni~ue"., has been applied to barnacle and sev-
eral dental cements. 21 , 2 This approach is particularly attractive,
since once the specific adhesive-fracture energy is known, the fail-
ure stress for any situation can, in principle, be predicted. 22
Widespread application of this technique appears unlikely due to two
major reasons. First, it is nQt clear that cohesive-failure in this
test implies that the adhesive bond is stronger than the fractured
material. Secondly, it would be difficult to apply the blister tech-
nique to adhesion between resins and enamel or dentin due to the
small size of the specimens. Adhesion to enamel and dentin has been
found to be dependent upon their surface morphology, and in order
to control this variable, the specimens must be small. 23 ,24,25

A work of fracture technique, which gives the interfacial work

292 W.J. O'BRIEN AND S.T. RASMUSSEN

of fracture (Wi)' has been applied to dental adhesion. 26 This

method shows some promise in that it can be appl5.ed to small speci-
mens and cohesive failure appears to indicate that the adhesive bond
is stronger than the material in which the failure occurred.

Wettability and surface preparation studies are generally fol-

lowed up by bond-strength measurements. The bond strength test
configurations are generally designed such that the principal stress
was claimed to be either tensile 7 ,9,10,13,15,23,24,27-30 or
shear. 2 • 7,8,14,30 However, some tests were: 1) transverse,2.6
2) torsion,2 or 3) combinations of shear and tensile. 31 Of the var-
ious means of adhesion investigation, bond-strength tests are prob-
ably the most popular. Unfortunately with this emphasis a multi-
plicity of techniques have been developed. Consequently, there has
been considerable interest in standardizing a bond strength test.

In this paper we would like to present a thesis that current

bond strength tests are, in general, inadequate for the purpose of
evaluation and selection of adhesive materials. Consequently, we
propose that dental adhesion testing be reevaluated from a theoret-
ical, experimental and most significantly from a practical viewpoint.

ADHESIVE FAILURE

From theoretical considerations, most materials are expected to

have ideal strengths approaching Ella (E=elastic modulus) which
amounts to approximately 10 6 psi for ordinary window glass and 2.8
x 10 6 psi for iron. 32 Unfortunately, these immense ideal strengths
are reached only for freshly drawn glass fibers and single crystal
whiskers of defect free ceramics and of relatively dislocation free
metals. Except for extremely ductile materials, most useful mater-
ials are notch sensitive under certain condition. This notch sen-
sitivity arises because of stress concentration about the sharp tips
of the notch or cracks. This stress can easily exceed the forces
required to break the atomic bonds. Griffith33 showed that for
brittle materials enough energy must also be supplied to equal the
surface energy of the newly formed surfaces, otherwise materials
with sharp cracks would fail spontaneously under low loads. His ap-
proach has been further modified to apply to· semibrittle materials.
The following equation gives the critical tensile stress (ocr) for
failure of an isotropic-uniform plate with an elliptical crack
(length 2c, cf. Fig. 3a) subjected to plain stress:

OCr
= (2E 'Ye)fl2
11"C (1 )
A CRITICAL APPRAISAL OF DENTAL ADHESION TESTING 293

Figure 2. Scanning electron micrograph of an enamel-resin

(sealant) system. The specimen was prepared so as to show the
resin tags of the sealant (S) which penetrated the porous enamel (E).
The enamel-sealant interface runs diagonally across the upper left
hand corner.

o b

Figure 3. Uniform plates with an elliptical crack of length 2c.

The plates' dimensions are large in comparison to the crack. a) Iso-
tropic plate b) A two component plate with a crack at the interface
between the two materials.
294 W.J. O'BRIEN AND S.T. RASMUSSEN

where E and Ye are the materials elastic modulus and effective sur-
face energy (also called specific surface energy) respectively.34,35
The term Ye includes contributions from surface energy, plastic flow,
pUllout for composites or any other means of absorbing energy during
initiation of crack growth. Equation(l)along with others developed
for different geometries and loading conditions have proved very
successful in explaining the fracture toughness of many materials. 32

In a similar fasnion adhesive failure is linked to stress con-

centration about interfacial defects (Figure 3b). The formula ex-
pressing the critical pressure for adhesive failure in the "blister"
technique 21 ,22,35 of adhesion testing is:

32
3(I";fL2)
(h~3(EYo)
c-) c
(2)

where Ya is the specific adhesive fracture energy and the other par-
ameters are defined in Figure 4. In this case the plate (Figure 4)
is considered rigid in comparison to the adhesive. A more compli-
cated formula arises when the plate properties must be considered.
Inspection of Equation(~, shows that the failure pressure is very
dependent upon the defect size. Consequently, the strength of ad-
hesive bonds can also be seen to be highly dependent upon defect
size. The larger the interfacial defect, the weaker the adhesive
interface will be. Other techniques have been developed for inves-
tigation of the specific adhesive fracture energies of adhesives to
various materials. 35 All of these techniques take into account
both the specific adhesive fracture energy and the size of the de-
fect which results in failure.

The investigation of dental adhesive failure can generally be

characterized as ignoring this close interrelationship between the
specific adhesive fracture energy and defect size. This is partic-
ularly true for "bond strength" tests where results are interpreted
as measuring, in some average sense, the strength of the actual bonds
at the interface. Yet these tests are most critically sensitive to
the wide range of defect sizes resulting from preparation and/or ap-
plication of adhering material.

When two materials are joined adhesively, the interfacial resis-

tance to fracture can be greater than that of the weaker or m~6e
brittle material involved if the joint is free of defects. 35 ,
This is because of changes in stress distribution at the adhesive
interface resulting from a mismatch of elastic modulii at the inter-
face. Additionally one of the adhering phases will likely be intrin-
sically more resistant to brittle fracture making the interface po-
tentially less brittle than the more brittle phase.
A CRITICAL APPRAISAL OF DENTAL ADHESION TESTING 295

ADHESIVE FAILURE IN DENTISTRY

Many researchers feel that a sound adhesive joint will fail co-
hesively rather than adhesively.36 Unfortunately, this state of ad-
hesive perfection has not progressed to all adhesive joints prepared
in dentistry. At present, failure of adhesive joints involving
resins(composites) and enamel or dentin can be considered to fail
adhesively.l9 Of course the environmental factors are severe for
both preparation and long term service. It has been noted that even
properly made engineering joints involving epoxy can fail adhesively
after long time exposure to water. 36 While porcelain to metal res-
torations can be prepared in the laboratory with homogeneous well
prepared surfaces, these restorations are however, subjected to de-
sign limitations of the natural shape of a tooth and the remaining
tooth structure. 37 Consequently the bulk of the porcelain cannot
be freely increased to reduce stress concentrations. The porcelain
must be thick enough to give a natural appearance but not so thick
that the supporting tooth structure is severly weakened.
The investigation of specific adhesive fracture energy (Ya ) and
defects must be considered as highly crucial to the improvement and
evaluation of resin adhesives in dentistry. Interfacial defects are
practical problems arising from adverse environmental conditions and
techniques of manipulations as well as properties of the materials
involved. Consequently, tests should be made on specimens prepared
under simulated oral conditions so that the results will properly
reflect the effect of defects on the practical success of the adhe-
sive joint. It is likely that a comparison of Ya obtained under
optimum laboratory conditions with those simulating practical con-
ditions could lead to improvement of surface preparation techniques.
Adhesion to enamel a~~ ~~ntin is thought to be dependent upon
their surface morphology.' It i~8also likely that it may be de-
pendent upon the chemical variations wnich occur on the surface
and through the depth of enamel. Certain regions of enamel or
dentin may result in poor adhesion and thus be an area from which
failure is initiated. Testing techniques which could be applied to
small areas would be highly desirable. Interfacial work of frac-
ture (Wi) technique 26 is attractive from this point of view, since
it can be applied to small specimens. The fractured surface area
in Figure 5b can be around lmm2 in size. Wi is only an approxima-
tion to the specific adhesive fracture energy (Ya ) for initiation
of crack growth since its value reflects contributions from both
initiation of crack growth and for crack propagation. However, in
absence of a Ya technique applicable to small specimens, the Wi
method appears to be an important adjunct to dental adhesion
research.
Many of the materials involved in restorative dentistry (enamel,
dentin, porcelain, composites) can be considered as brittle materials
296 W.J. O'BRIEN AND S.T. RASMUSSEN

2c BLISTER

-.;;;±::::=::;;;;:::~t:=:Y
h

. . .,. . . __,~ 'r___~---...T

PRESSURE

Figure 4. Schematic of the blister technique showing the adhesive

with elastic modulus E and Poisson's ratio ~ and the adherend
(plate). Pressure is increased until failure occurs.

\* Water

~
Porcelain Metal

t) (t
Interface between metal and porcelain
-a-

Frle'urad
Sutfecel

-b-

Figure 5. Schematic of porcelain-metal specimens. Specimens were

notched at the interface between the two materials.
A CRITICAL APPRAISAL OF DENTAL ADHESION TESTING 297

because of their low strengths 39 ,40 or low work of fractures. 25 ,26

Consequently, these materials as well as adhesive joints involving
them are likely to have strengths which are highly defect sensitive.

For our purposes, defects can be classed loosely as those which

are unavoidably part of the materials involved (Type I defect) and
those which arise during preparation (Type II defect). For porcelain,
Type I defects are those which are related to the grain size of the
quartz, alumina and etc. For dental composites, a Type I defect may
be bubbles of gases trapped during manufacture or the particles them-
selves may act as sources of stress concentration. Type I defects
determine a material's maximum strength. Defects arising from pre-
paration, Type II defects, will most likely lower the practical
strength of the materials.

It should be evident that defects play an important role in any

investigation of the strength of materials or adhesive joints since
the researcher determines the number and size of defects by his pre-
paration technique. The problem of defects in dental materials are
unique in that the number and size of Type II defects are mainly
technique related. There is likely to be a wide variability in an
adhesives lifetime under mechanical stress due to differences in
manipulation. In dentistry a material with average material proper-
ties may give the best results because its manipulative character-
istics give either smaller or fewer large Type II defects. It is
our contention that researchers of dental materials must address
themselves to this problem where applicable since it is the largest
defect or number of defects which will limit the cohesive or adhe-
sive strength of an adhesive system.

Specific adhesive fracture energy (Ya) and interfacial work of

fracture (Wi) measurements are advantageous since they are, in prin-
ciple, independent of defects. Ya measurements begin with a pre-
determined defect size 2l ,22,35 and give values independent of the
defect size (Le. Ya will be the same regardless of defect size).
Interfacial work of fracture (Wi) measurements, while in principle
are not defect size dependent,. do give in practice values which are
related to the crossectional area of the defects intersected by the
fractur~d surface. 26 The effect is proportional to the defect area
as compared to the total fractured area. If the defect area is a
large percentage of the total fractured area it could be measured
and be eliminated from the measured Wi' Ya or Wi measurements could
be coupled with statistical studies of defect distributions to give
the effect of defects on the strength of an adhesive joint or a ma-
terials strength. The defect investigation should be conducted on
specimens which are most representative of materials prepared under
practical conditions. In fact old restorations would be best from
this point of view.
298 W.J. O'BRIEN AND S.T. RASMUSSEN

Porcelain-metal restorations are likely to fail cohesively.16,39

For adhesive joints, where cohesive failure is likely, the fracture
path approach is mandatory since it does little good to report the
fracture energy or strength of the fractured material as being a
measure of the quality of adhesion. The measurements reflect char-
acteristics of the material in which the fracture occurred and not
the adhesive joint. In fact it is important to determine the frac-
ture path for all specimens tested for adhesive failure. Figure 6
shows the various types of failure which can occur in metal-porcelain
systems. While it is thought 19 that adhesive failure will occur for
resin-enamel system cohesive failure is also possible and will be-
come more so as adhesives are improved. Figure 7 shows some of the
single phase type failures which might occur for an enamel-composite
system.

An important aspect of the fracture-path approach for cohesive

failure is that under certain conditions cohesive failure indicates
that the adhesive bond is "stronger" than the material in which
failure occurred. If the stress distribution, based upon a defect
free specimen, is such that the tensile stresses at the adhesive
interface are equal to or greater than those in the region of
failure the adhesive bond can be considered sufficient. However,
specimens are not defect free and failure is initiated at defects.
One must decide how the role of defects fit into interpretations of
cohesive failure.

The aim of many adhesive investigations is improvement of the

adhesive system. From this viewpoint it is not enough to determine
if failure is adhesive or cohesive, one should also determine the
source of failure. Several potential situations involving cohesive
failure will illustrate the importance of this statement. First con-
sider an enamel-composite system in which defects resulting from
mixing lead to cohesive failure. Here a change in the technique
of mixing the composite components together might reduce the size
of the defects in the composite leading to improved strength. Con-
ceivably, the strength could improve enough that adhesive failure
could occur. That is, the adhesive system is improved by changing
the failure mode from cohesive to adhesive. Second consider a
metal-porcelain system in which failure is initiated at a large in-
terfacial defect but cohesive failure occurs ( cf. Figure 8). The
adhesive interface is the weak link of the system and efforts should
be aimed at reducing the size of interfacial defects. Consequently,
a change in the methods of surface preparation and/or a change in
the means of applying the porcelain powder to the metal could lead
to a substantial improvement of a system in which cohesive failure
occurs.
A CRITICAL APPRAISAL OF DENTAL ADHESION TESTING 299

i:: ~: ':i:!!I!~!ill:::':':' ji ! 1
I. METAL - PORCELAIN II, METAL OXIDE - PORCELAIN

f.:. . :: ..:: ..:. .'. .' .,. ':.. '::".: ...j

•:0:: :':':.:.:':':}~~~~'~!,~: :.:.:: :.:.:: :.:.:;
Metal Oxide
.:'.: ....::.:....Po'reeio'i;,:
. ....:'.: ............
.. .."::.: ::.' ::.": ..
MetalOxid,

m. COHESIVE WITHIN PORCELAIN nr, METAL - METAL OXIDE

: ~:~:~:~:~:~:~:~:~:~:}~:~: :~ ~.~~{:}?~:~.?;:~.~.
~,tol 0'l,i 41,
................ ... .................
1.0.0. M'tol 0111(/'

Figure 6. Porcelain-metal failures. Classification of failure

is based upon the region in which the interface formed.
300 W.J. O'BRIEN AND S.T. RASMUSSEN

o.
DO b. c.

d. e.
Figure 7. Enamel-composite failures: a) solid specimen showing
enamel (E), particle free resin (R), resin-particle composite (C);
b) cohesive within enamel; c) resin-enamel interfacial failure;
d) cohesive within resin; e) interfacial at resin film-filler
interface; and f) cohesive within composite. The cohesive failure
of the enamel (b) is drawn to characterize the surface found for
failure of enamel. 25 .

PORCELAIN

Figure 8 A composite plate, similar to Figure 3b, showing how an

interfacial defect could initiate cohesive failure.
A CRITICAL APPRAISAL OF DENTAL ADHESION TESTING 301

BOND STRENGTH TESTS

The investigation of dental adhesive failure can generally be

characterized as ignoring the close interrelationship between the
specific adhesive fracture energy and the defect size. This is
particularly true for "bond strength" tests where results are in-
terpreted as measuring, in some average sense, the strength of the
actual bonds at the interface. Yet these tests are most critically
sensitive to the wide range of defect sizes resulting from prepara-
tion and/or application of adhering material.

Another major problem with bond strength investigations is that

a stress analysis was not made of the test geometry taking into ac-
count the elastic properties of the materials involved. Figures 9
and 10 depict various geometries used in testing porcelain to metal
or resin to tooth structure bonds respectively. As is evident, the
test geometries are, in general, not simple and the combination of
different materials makes interpretation of the failure stress most
difficult. That is, one cannot say intuitively what the failure
stress is. Researchers claim a particular type of stress (tension
shear etc.) resulted in failure and that the failure stress was given
by the ratio of the breaking load to the fractured ar~a (references
2,6,7,8,10,13,15,23,24,27,28,29,30,31). Recently ,17 , 1 some por-
celain-metal bond-strength test configurations have been subjected
to finite element stress analysis which verified these intuitive
criticisms of bond strength testing in dental adhesive research.

Other problems with bond-strength tests can be related to speci-

men preparation or testing. The measured tensile strengths of brittle
materials can be highly dependent upon specimen preparation and a-
lignment of testing apparatus. This is particularly true for many
dental restorative materials because in addition to many of them be-
ing brittle or semibrittle, the necessity of preparing specimens
similar in size to actual restorations magnifies problems due to
preparation and alignment of testing apparatus. These problems arise
for brittle and semibrittle materials because of their sensitivity
to stress concentration about shar~ defects. Similar conditions
exist for bond-strength tests. 2l ,35

Most bond strength testing techniques have been designed for

testing in air at room temperature rather than environments simu-
lating oral conditions. It is a well-known fact that the environ-
ment can have a large effect on the strength of materials. 36 ,40

Bond strength measurements of dental cements have been found

to be highly dependent upon film thickness. Additionally it is
difficult to control this variable. 23 ,29,30 Film thickness may also
be a problem for composites placed directly on the enamel or dentin
surface since there will most likely be a particle free region next
302 W.J. O'BRIEN AND S.T. RASMUSSEN

o. b.

i
METAL PORCELAIN

PORCELAIN
METAL

c. d.

PORCE L AI~
PORCELAINJ
t t
METAL
METAL

Figure 9. Schematic drawings of various test geometries that have

been used to test the porcelain-metal interface.
o. b.
SPECIMEN SPECIMEN

RESIN
(Cement)

c. d.
SPECIMEN SPECIMEN

RESIN
(Cement)
ORTHODONTIC
LINGUAL BUTTON

Figure 10. Schematic drawings of various test geometries that have

been used to test the resin-enamel/dentin interface.
A CRITICAL APPRAISAL OF DENTAL ADHESION TESTING 303

to the surface. This particle free region could be considered a

film of resin ( cf. Figure 7a).

The use of bond strength measurements for ranking/comparing of

adhesives would be appropriate if the actual stresses were known and
the specimens were prepared and tested under simulated oral condi-
tions. The values obtained would reflect contributions from both
specific adhesive fracture energy and defects. However, if one de-
sires to improve an adhesive, independent information about both
specific adhesive fracture energy and defect size would be prefer-
able.

SUMMARY

Concepts from fracture mechanics were used to interpret adhe-

sive and cohesive failure of adhesives used in dentistry. It was
concluded that current bond strength tests are inadequate for eval-
uating and improving adhesion to the extent of emphasis currently
placed on them. Until recently,17,41 bond strength tests have not
been subjected to stress analysis and consequently the meaning of
their measurements are questionable. The practical strength of an
adhesive interface is a complex relationship between the stress dis-
tribution (determined by the load, geometry and the elastic proper-
ties of the materials involved), specific adhesive fracture energy
and the size and shap.e of interfacial defects. Bond strength tests
are not designed to investigat,e these separate aspects related to
adhesive failure. That is they are affected by all of them and con-
sequently it is difficult to determine the major source of failure.

The problem of defects in dental adhesive failure are unique in

that the number, size and shape of defects which result in failure
are mainly technique related. However, the investigation of dental
adhesive failure can generally be characterized as ignoring the re-
lationship that specific adhesive fracture energy and defects have
on the practical strength of adhesive systems. Consequently, inde-
pendent knowledge of specific adhesive fracture energy and defect
characteristics could lead to substantial improvement in the prac-
tical strength of dental adhesive systems.

ACKNOWLEDGEMENT

Dr. O'Brien wishes to acknowledge the support of Research Grant

DE02731 from the National Institute of Dental Research, National
Institutes of Health, Bethesda, MD 20014
304 W.J. O'BRIEN AND S.T. RASMUSSEN

ADDENDUM

Currently, the poorest adhesive system in restorative dentistry

is that of the resin-dentin system. However, some recen.t research
indicates potential improvement in the bonding between resins and
dentin. The references are: 1) R.L. Bowen, E.N. Cobb, and J.E.Rapson,
J. Dent. Res., 61, 1070 (1982); and 2) N. Nakabayashi, K. Kojima, and
E. Masuhara, J.~iomed. Mater. Res. 16, 265 (1982).

We have recently received an update on the blister method. It

was used successfully to measure the specific adhesive fracture
energy of a gingival cell line to agar. The diameters of the blisters
(cf. Figure 4) for the gingival cell line tested ranged from 0.4 to
1.6 mm. Considering the small diameters of the blisters used in the
gingival cell investigation, it may also be possible to apply this
technique to testing the adhesion of resins to enamel or dentin. The
blister technique deserves further consideration for its potential
use in dental adhesive investigations. The reference is: R.K.
Fletcher, S.M. Breitling, J.K. Adamich, and M.L. Williams Jr., in
Vitro, 13, 857, (1977).

REFERENCES

L G.M. Brauer, in "Scientific Aspects of Dental Materials".

T.A. von Fraunhofer, Editor, Butterworths, Boston, (1975).
2. R.C. Vickery and L.A. Badine11i, J. Dent. Res. 47, 683 (1968).
3. H.D. Moskowitz, G.T. Ward, and E.D. Woolridge, Editors, "Dental
Adhesive Materials," Proceedings of the Symposium held November
8-9, 1973, Department of Health, Education, and Welfare,
Washington, D.C., (1974).
4. W.T. O'Brien and G. Ryge, J. Prosthet. Dent. 15, 1094 (1965).
5. R.L. Bowen, J. Amer. Dental Assoc., 74, 439 (1967).
6. M.H. Lavine and F. Custer, J. Dent. Res., 45, 32 (1966).
7. J .W. McLean and LR. Sced, Trans. J. Brit. Ceramic Soc., Q,
229 (1973).
8. J.W. McLean and I.R. Sced, Trans. J. Brit. Ceramic Soc., Q,
235 (1973).
9. G.M. Brauer and D.J •. Termini, J. Dent. Res. 51, 151 (1972).
10. D.H. Retief, J. Dent. Res. 52, 333 (1973). --
11. N.A. Wickwire and D. Rentz,~. J. Orthodont. 64, 499 (1973).
12. R.L. Bowen, J. Dent. Res. 44, 903 (1965). --
13. J.B. Moser, D.B. Dowling, E.H. Greener, and G.W. Marshall
J. Dent. Res. 55, 411 (1976).
14. R.L. Patrick, C.M. Kaplan, and E.R. Beaver, J. Dent. Res. ~,
12 (1968).
15. H.L. Lee~ A.L. Cupples, R.J. Schubert, and M.L. Swartz
J. Dent. Res. 50, 125 (1971).
A CRITICAL APPRAISAL OF DENTAL ADHESION TESTING 305

16. W.J. O'Brien, in "An Outline of Dental Materials and Their

Selection", W.J. O'Brien and G. Ryge, Editors, pp. 180-194,
Saunders, Philadelphia, 1978.
17. K.J. Anusavice, R.D. Ringle, and C.W. Fairhurst, J. Prosthet.
Dent. 42, 417 (1979).
18. J.A.Rensch, Am. J. Orthodont., 63, 156 (1973).
19. D.H. Retief, J. Oral Rehab. 1,265 (1974).
20. F.J. Knap and G. Ryge, J. Dent. Res. 45, 1047 (1966).
21. M.L. Williams, K.L. De Vries, and R.R. Despain, J. Dent. Res.
~, 517 (1973).
22. P.R. Despain, K.L. De Vries, R.D. Luntz, and M.L. Williams,
J. Dent. Res. 52, 674 (1973).
23. R.L. Bowen, J.lDent. Res. 44, 690 (1965).
24. R.W. Phillips, M.L. Swartz, and B. Rhodes, J. Amer. Dental
Assoc. 81, 1353 (1970).
25. S.T. Rasmussen, A.H. Heuer, R.E. Patchin, and D.B. Scott,
J. Dent. Res., ~, 154 (1976).
26. S.T. Rasmussen, J. Dent. Res. 22, 11 (1978).
27. M.L. Swartz and R.W. Phillips, J. Amer. Dental Assoc.
50, 172 (1955).
28. H.L. Lee, M.L. Swartz, and G. Cu1p, J. Dent. Res. 48, 211 (1969).
29. G.T. Eden, R.G. Craig, and F.A. Peyton, J. Dent. Res. ~,
428 (1970).
30. E. Misrahi and D.C. Smith, Br. Dent. J. 127, 371 (1969).
31. J.S. Shell and J.P. Nielsen, J. Dent. ReS:-41 , 1424 (1962).
32. A.H. Cottrell, "The Mechanical Properties ofMatter,"
pp. 223-225, 336, John Wiley and Sons, New York, 1964.
33. A.A. Griffith, Phil. Trans. Roy. Soc. (London) A221, 163 (1920).
34. H.G. Tattersall and G. Tappin, J. Materials Sci. 1, 296 (1966).
35. G.P. Anderson, S.J. Bennett, and K.L. De Vries, "Analysis and
Testing of Adhesive Bonds," Academic Press, New York, 1977.
36. "Handbook of Adhesive Bonding", Charles V. Cagle, Editor,
Mc Graw Hill Book Co., New York, 1973.
37. R.G. Craig and J.W. Farah, Oral Sci. Rev. 10, 45 (1977).
38. J.A. Weathere11, C. Robinson, and Ha11sworth, J. Dent. Res. 22,
180 (1974).
39. W.J. O'Brien, in'benta1 Porcelain: The State of the Art-1977:'
H.N. Yamada, Editor, pp. 137-141, University of Southern
California School of Dentistry, Los Angeles, 1977.
40. C.A. Sherrill and W.J. O'Brien, J. Dent. Res. 22, 683 (1974).
41. K.J. Anusavice, P.H. De Hoff, and C.W. Fairhurst, J. Dent.
Res., Special Issue A 57, Abstract No. 869 (1978).
ADHESIVE JOINT CHARACTERIZATION BY ULTRASONIC SURFACE AND INTERFACE

WAVES

S. I. Rokhlin

Department of Materials Engineering

Ben-Gurion University of the Negev
P.O.B. 653, Beer Sheva 84105, Israel

Methods of measurement of the velocity and

attenuation of ultrasonic waves have been used for a
long time for investigating structural and mechanical
properties of materials. The main purpose of this
article is to show that a properly updated ultrasonic
method can be successfully used for investigating
surface and interface films, in particular, of
adhesively bonded structures. Theoretical analysis
of the propagation of elastic waves in a multi-layered
medium can be used for relating the velocity of surface
and interface waves with the elastic properties of the
film and substrates. If the elastic properties of the
substrates are known, then the viscoelastic properties
of films can be determined from measurements of the
velocity and attenuation of these waves. The velocity
of surface and interface waves is a function of not
only the film properties, but also of adhesion of the
film to the substrates. This makes the suggested
technique potentially promising for the evaluation of
adhesion properties of thin surface and interface
films.

307
308 S.I. ROKHLlN

INTRODUCTION

Ultrasonic methods have been used extensively for investigating

the properties of materials and for process control. Various
aspects of this problem were reviewed in the survey by Papadakis. 1
Ultrasonic waves used for testing properties of materials consist
of small-amplitude mechanical vibrations and, depending on the
vibratory mode being utilized, may produce both longitudinal and
shear stresses in the solid body. Information about the structural
properties of a substance can be obtained by measuring both the
velocity and the attenuation of the ultrasonic wave. The phase
velocity of the wave is controlled by the elastic constants of the
body. In an isotropic solid medium, which has only two independent
elastic moduli, there exist two elastic waves: the longitudinal
with velocity Vt and shear, with velocity Vt , which are obtained
from the expressions:

VR, = I(H 2]J)/p (1)

where p is the density of the material, ]J is the shear modulus and

A is the Lame parameter.

Ultrasonic velocities can be expressed in terms of other elas-

tic constants of the solid body on the basis of the coupling
equations:

2 ]J(E...;.2]J) Ev
A=K--]J A
3 3]J-E (1-2v) (1+v)

E = 9]JK E
]J = 2 (1+v) (2)
3K+]..I

where K is the bulk modulus, E is Young's modulus and v is Poisson's

ratio. For liquids ]J = 0 and the velocity of the 1ingitudina1 wave
is controlled only by the bulk modulus, i.e., VR, = KIp. For rub-
ber-like materials K » ]J, and E ~ 3]J.

Measurement of Vt and Vt is one of the most exact methods for

measuring elastic constants. Three kinds of bulk elastic waves may
propagate in an anisotropic solid body: a quasi-longitudinal, and
two quasi-transverse waves, differing in polarization and velocity.
To determine the matrix of elastic constants one must measure the
phase velocity in several different directions relative to the
crystallographic axes.

The attenuation of an ultrasonic wave is associated with ab-

sorption of elastic waves (ane1astic effect) and the scattering of
elastic waves by structural inhomogeneities. Scattering may be
decisive in po1ycrysta11ine, composite and ceramic materials. In
the course of scattering, elastic energy is lost by the ultrasonic
CHARACTERIZATION BY ULTRASONIC WAVES 309

beam in the form of a stochastically scattered field, which is

gradually absorbed. The latter is associated with conversion of
elastic into thermal energy, as a result of various anelastic ef-
fects termed internal friction.

The velocity and attenuation of elastic waves are affected by

the same microstructural factors of the material as the strength.
Using this fact, Vary2 is currently developing an application of
the ultrasonic method for evaluation of the mechanical properties
of materials.

The ultrasonic method is extensively used for evaluation of

adhesively bonded structures. A detailed survey of the latest 3
achievements in this field was given recently by Segal and Rose.
The majority of techniques developed until recently (these shall be
referred to as standard techniques) are an upgrading of methods
used for evaluation of bulk specimens. At the same time, testing
of adhesively bonded structures and coatings has a number of dis-
tinguishing features:

1) It is necessary to estimate the properties of a very thin

layer of material, the thickness of which is much smaller than the
thickness of the adherends or substrates. The standard method which
employs a normally incident ultrasonic beam cannot provide accurate
information, since the length of the acoustic path is very small.
The spectral method yields a lower accuracy in measuring the wave
velocity and attenuation than conventional ultrasonic methods.

2) The mechanical properties of the coating-substrate or ad-

hesive-adherends system is controlled not only by the "bulk" proper-
ties of the film, but also to a significant degree, by the proper-
ties of the film-adherend interface. The majority of standard
methods 3 cannot, in principle, solve this problem.

The purpose of this paper is to discuss the possibility of

application of ultrasonic surface and interface waves for charac-
terizing thin coatings and interface films. It appears that these
new techniques have great potential possibilities. These waves,
termed guided waves, propagate along the surface or interface, and
hence the length of the acoustic path becomes the same as in bulk
measurements. For specially selected testing conditions these
waves produce shearing strains in the bond-line region, and thus
provide information about the adhesion properties of the adhesive-
adherend interface.

Surface acoustic waves have only recently come into use for
evaluation of thin films. As of now the principal results using
this method were obtained for metal films. There are only few re-
sults on evaluating the properties of polymer coatings. Conversely,
the applications of interface waves were primarily developed for
310 5.1. ROKHLlN

evaluation of adhesively bonded structures. Below we consider cer-

tain methodological aspects of using these methods and describe
analytical and experimental results, pertaining primarily to the
use of interface waves.

SURFACE ACOUSTIC WAVES (SAW)

As shown by Rayleigh, an elastic surface wave may propagate

along the surface of a solid body. The elastic energy of surface
waves is localized in a subsurface layer with thickness of the order
of wavelength A. Figure 1 shows 4 the energy flux distribution of a
surface acoustic wave with depth X, normalized to the wavelength.
At a depth of the order of A there is virtually no transport of
elastic energy. Figure 2 depicts schematically the deformations in
the subsurface layer induced by a surface wave. It is seen that
both longitudinal and shear elastic displacements occur in the wave.
The magnitude of surface displacement u in the figure is intention-
ally magnified, while actually it is in the order of tens of A and
is several orders of magnitude smaller than the wavelength A
(U«A).

Velocity Vr of the surface wave in an isotropic solid body is

always lower than the velocity of the shear wave and is found from
numerical solution of the equation: 5 ,6

-<
"
~
I
tLw
Cl
z
Q
(f)
z
~ 1.2
(5
z
oz
RELATIVE POWER FLOW

Figure 1. Relative energy flux of surface acoustic waves as func-

tion of depth (Data from Farnel14 ).
CHARACTERIZATION BY ULTRASONIC WAVES 311

Unper lurbed
surface
u

Figure 2. Deformation in the subsurface layer on propagation of a

Rayleigh surface wave.

b.
r
o (3)

where
2 _ ( 2 t"2)1/2
W = 2a -1 Y - a-"
a = Vt/V r is the nondimensional wave number of the surface wave,
~ = Vt/V~ is a quantity which is a function only of the Poisson
ratio. With an accuracy better than 0.5% the velocity of a surface
wave can be calculated from the approximate equation: 5 ,6

0.87+l.l2v
V Vt l+v (4)
r

It is seen from Equations (3) and (4) that the velocity of a surface
wave is a function of both elastic constants of the isotropic
material.

The presence of a thin film on the substrate surface changes

the surface stiffness and also lOeds the surface. Figure 3 shows
schema-cically that a thin film, whose thickness is much less than
a wavelength, is subjected to a compression-tension strain in the
course of propagation of a surface wave. The strain arising in the
film is similar to strains in a thin plate upon propagation of a
longitudinal plate mode. In this case the velocity of the surface
wave depends not only on the substrate properties, but on the stiff-
ness Eo/(l-v~) and the density Po of the film (Eo is Young's modu-
lus, Vo is the film Poisson ratio). Achenbach and Keshawa 7 , and
Tiersten 8 used an approximate theory of thin plates for finding the
velocity of a surface wave in a system consisting of a substrate
312 S.1. ROKHLlN

-- -
( b)
(0 )

Figure 3. a) Deformation of a surface of a solid body with

coating; b) Longitudinal strain of a surface film on propagation
of a Rayleigh wave.

and a single coating. The corresponding equation for the surface

wave velocity is best written in the form

a 2E h
- 2 0
(S+y)hS - a l (he) 0 (5)
r + + 2 (alhS-S)
A
(I-v )
0

where

a = K/K • K = w/V • 2
S = p /p. 11 = K h· a = a -Sy
o ' t ' t' s' 1

~s is the velocity of the surface wave in the coated structure, h

~s the coating thickness, ~ is the shear modulus and p is the den-
sity of the substrate material. This equation can be extended 9
to the case of a multilayer coating.

Measurements of the surface wave velocity can yield infor-

mation about the coating properties and about the condition of the
bond between the coating and the substrate. Two methods of measure-
ment are possible:

1) direct measurement of the velocity and attenuation of the

surface wave;

2) the Rayleigh angle method.

CHARACTERIZATION BY ULTRASONIC WAVES 313

DIRECT MEASUREMENT OF THE VELOCITY AND ATTENUATION OF SURFACE

ACOUSTIC WAVES. APPLICATION TO EVALUATION OF COATING PROPERTIES

In spite of the wide application of thin films in surface

acoustic wave (SAW) electronic devices, there is very little known
about the use of surface waves for thin film property measurements.
The measurement of SAW attenuation lO - 13 was used for the study of
properties of thin films as a function of magnetic field or tempera-
ture. Measurements of the velocity and attenuation of surface
acoustic waSes during deposition of thin films were also des-
cribed. 14 ,1

A number of different methods 6 can be used for excitation of

surface acoustic waves. For relatively low ultrasonic frequencies
(to 10 MHz) one can use wedge transducers, whereas at high frequen-
cies interdigital transducers are used. As an illustration Figure
4 shows a specimen with two pairs of interdigital transducers. The
test film is placed between one pair of transducers, whereas the
second pair of transducers without a film between them is used as
the reference channel. The velocity of the surface wave for the
film region can be determined from the difference in phase velo-
cities of surface waves in both channels. The film properties can
then be evaluated from Equation (5).

Frequently surface films are multilayered or have nonuniform

elastic properties through the thickness. (Such an inhomogeneous
film can be treated as a multilayered film.) The pertinent exten-
sion of Equation (5) to the case of a multilayered surface film can
be obtained by the matrix method. 9 Using the matrix approach, we
obtained a series expansion of the exact solution with respect to
the thickness of the successive layers as small parameters. We

WAVE
..
ULTRASONIC
REFERENCE
CHANNEL

-'Wv-
FROM
FILM MAIN
~
CHANNEL
GENERATOR ULTRASONIC
WAVE

Figure 4. Schematic drawing of the experimental specimen with two

pairs of interdigital transducers.
314 5.1. ROKHLlN

compared the analytic solution thus calculated for a single surface

layer with the Tiersten equation. He found that Tiersten' s equation
corresponds to our expansion only up to the linear term. It is
reasonable to expect that this approximation should yield the same
accuracy as in the multilayered case, provided that the total thick-
ness Lh. of the multilayered system is not greater than the allowed
film thIckness h in the Tiersten equation. By carrying out the ap-
propriate calculations for the characteristic equation of surface
waves, we obtain:

2 - 2 2 E.H.
~ ~
~
r
+ (S+y)LH.e. - (a -Sy)[LH.e.]
~ ~ ~ ~
+a L 2
].l(l-v.)
~

2 -
[(a - Sy)LH. e. -13]
~ ~
o (6)

where

~
r
= W2 2 2222 222
-4a Sy; w = a +13; 13 = a -1; y = a -~ ;

K = w/V s
Vs is the velocity of the surface wave in the multilayered structure,
Ei and vi are the Young modulus and Poisson ratio, respectively for
the ith layer, ].l is the shear modulus and p the density of the sub-
strate material and Pi the density of the ith layer. K~ is the wave
number for the longitudinal waves in the substrate.

Equation (6) is commutative with respect to the order of the

successive layers. Physically this is due to the fact that as
Lhi « As (where As is the wavelength of the surface wave), the
longitudinal strain in the layers is close to being homogeneous.

As an illustration of the application of this method, let us

consider the results of Rokh1in, Ronen and Darie1,9 who employed
surface acoustic waves for analyzing the elastic properties of an
Au-Al thin-film diffusion couple. A thin film of the AU2Al inter-
metallic compound appears in the course of the diffusion reaction
at the interface between the Au-Al films. Hhen diffusion termi-
nates, the Au-A1 sandwich becomes an AU2Al film. This structural
modification induces a change in the surface wave velocity, which
can be measured during the experiment.

An example of the results of the conductivity and ultrasonic

SAH measurements as a function of time of diffusion is shown in
Figure 5. The results indicate that the conductivity and, conse-
quently, the thickness of the growing phase (Au2Al), depend linearly
on It after a certain incubation period. After this incubation time,
the phase shift of the SAH also becomes a linear function of It.
CHARACTERIZATION BY ULTRASONIC WAVES 315

AuzAI FILM THICKNESS (Al

0 iOOO 2000 3000 4000

..
100 ........ 0015

>- "
'U
I-
:> ... I
i= 075 .''II- ~
(f)
V

,_
0010 rrI
:::J
0
Z , (f)
0
V
Cl
w
050
\.
_
..... ...
I
TI
-I
N • CONDUCT,VITY 0005 "t:
:i 025 x PHASE SHIFT
<II
~
<1 n
::2
a:: _CALCUL ATIONS
0
z
0 '. 10 15 20 25 30 350

TIME '-frM,

Figure 5. SAW phase shift and electrical conductivity changes as a

function of the annealing time in the Au-Al diffusion couple. When
the growing phase AU2Al reaches the surface of the substrate the
previous monotonic behavior of the SAW velocity changes suddenly.
This phenomenon is explained by a break of the adhesion bonds on
the film substrate interface (from Rokhlin. Ronen and Darie1 9 ).

When the growing phase AU2Al reaches the surface of the under-
lying Cr layer, the previous monotonic behavior of the experimental
curves changes suddenly. The curve of the phase shift of the SAW
passes through a maximum and then drops by approximately 30% of its
maximum value. The electrical conductivity curve passes through a
minimum and then increases slowly un til s tabili ty is .attained.
This effect is in agreement with the results of Weaver and Parkin-
son16 who measured the electrical conductivity and adhesion to a
substrate of an Au-Al couple. The shape of the phase shift curve.
following its maximum is similar to Weaver and Parkinson's curve
representing their adhesion measurements. A suggested mechanism
for these changes is related to the diffusion-induced strains pre-
sent at the various interfaces.

Glass and Cr flow at significantly higher stresses than Al and

Au. As the annealing treatment goes on and the compound layer
reaches the Cr surface, the elastic stresses at this interface are
unable to relax by plastic flow. At this stage, we may expect a
break of the adhesion bonds at either the AU2Al-Cr or the Cr-glass
316 . S.1. ROKHLlN

interface. The simultaneous increase of the conductivity is attri-

buted to the relaxation of stresses associated with the loss of ad-
hesion.

As a result of the reduced adhesion to the substrate the effect

of the presence of AU2A1 on the velocity of the surface wave de-
creases. This results in an increasing velocity of the surface wave
since the velocity of the SAW at a substrate with an overlying AU2A1
layer is lower than the velocity of the SAW at a free substrate.

Using the velocity data and Equation (6) the effective elastic
modulus E/(1-v 2 ) of the Au2A1 intermeta11ic compound was determined.

LEAKY SURFACE WAVES

It was noted above that, for the isotropic case, the velocity
of the surface wave is lower than the velocity of bulk waves in the
half space, and hence the wave energy is localized near the surface.
The situation changes when the elastic half space interfaces with a
liquid. The velocity of the surface wave is higher than the velocity
of the acoustic wave in a liquid (for example, Vr in steel is about
3 km/sec, whereas the velocity of sound Vo in water is about 1.5
·km/sec). As shown in Figure 6 this results in radiation (leakage)
of elastic energy of the surface wave to the liquid. For this
reason such a wave is termed a leaky surface wave; the radiation
angle of this wave is obtained from Snell's law.

Bs =orcsin Y2.
Vs
~
B Va LEAKY
LIQUID s WAVE
.... Vs
SOLID - SURFACE
WAVE

Figure 6. Leakage of elastic energy of the surface wave to the

liquid (leaky surface wave). The radiation angle is obtained from
Snell's law. Vo - is the ultrasonic velocity in the liquid. Vs -
is the velocity of the surface acoustic wave.
CHARACTERIZATION BY ULTRASONIC WAVES 317

An elastic surface wave also may become transformed into a

leaky wave if the substrate is coated by a very stiff coating, and
also in a free anisotropic half space (when there is no liquid half
space). In the anisotropic case the surface wave velocity is a func-
tion of the propagation angle 17 and over a certain range of angles,
its velocity may exceed the velocity of the slow shear wave. This
will then cause leakage of energy from the surface to the substrate.

RAYLEIGH ANGLE METHOD

When the ultrasonic wave is incident from the liquid onto a
liquid-solid interface longitudinal and transverse refracted waves
are produced in the solid, as shown in Figure 7. Upon increasing
the incidence angle a the refraction angles S~ and St increase, so
that at a = a~ = arc sin Vo/V~ we have S~ = 90°. This angle is
termed the longitudinal critical angle or the first critical angle.
At 6 = 6t = arc sin Vo/Vt, which is termed the second critical
angle, we have St = 90° and complete internal reflection occurs.
At a > at the absolute value of the reflection coefficient should,
according to the classical theory, be equal to unity. For illus-
tration Fi§ure 8 shows by a solid line the energy reflection co-
efficient l for the case of a water-steel interface. The calcula-
tions were performed for the reflection of a plane incident wave
and absence of attenuation in the solid. The points on this figure
represent Neubauer's 18 experimental results. Near the angle a r =
arcsin Vo/Vr (Vr is the velocity of the surface wave) a steep drop
is observed in the experimental reflection coefficient. The above
effect is due to differences between the experimental and theoreti-
cal conditions: 1) presence of attenuation in the specimen; 2) in-
cidence of a bonded ultrasonic beam, which is due to the finite
dimensions of the radiator.

INCIDENT REFLECTED
WAVE WAVE

LIQUID

SOLID LONGITUDINAL
WAVE
SHEAR
WAVE

Figure 7. Reflection and refraction of an acoustic plane wave at

the liquid-solid interface.
318 5.1. ROKHLlN

8t 8r
I- 10 I
z
w
U
u::
u.
08
~
u
z 06
. E xperlmenlal
0 - Calcula ted
f=
u 04
w
---'
u.
W
0::
02
>-
<.!)
a::
w
z
w 0 40
ANGLE (degrees)

Figure 8. The energy reflection coefficient for incident plane

waves in water on a water-steel interface (solid line). The points
represent the experimental data (from Neubauer18 ).

Bertoni and Tamir19 gave a theoretical explanation of the ex-

perimentally observed effects: 1) lateral beam displacement ~ of
the reflected ultrasonic beam; 2) nonhomogeneous intensity distri-
bution in the reflected beam; 3) change in the amplitude and phase
of the reflected beam near the Rayleigh angle. They showed that if
the width of the incident beam is more than four times greater than
the value of the lateral displacement ~, the distortion in the pro-
file of the reflected beam vanishes and the value of ~ becomes equal
to the displacement width predicted by Schoch. 20 The reflected
field is produced by interference of two wave contributions: 19 ,21
1) reflected beam according to geometric acoustics, and 2) contri-
bution associated with the leaky surface wave, which is excited by
the incident beam and in the course of propagation radiates energy
to the liquid (Figures 9a,b).

Becker and Richardson 22 showed that the velocity and attenua-

tion of a surface wave can be determined with sufficient accuracy
from the measured amplitude and phase of the reflected beam.
CHARACTERIZATION BY ULTRASONIC WAVES 319

Figure 9. a) Specular reflection of the bounded ultrasonic beam

from the liquid-solid interface. b) Nonspecular reflection of the
bounded ultrasonic beam at the Rayleigh angle.

The measuring principle is indicated in Figure 9. A bounded

ultrasonic beam with width b much greater than its displacement ~
is incident at an angle 8 onto the test specimen. Angle 8 can be
changed with a high accuracy. The receiving transducer receives
the reflected wave at the same angle. This transducer is shifted
by an amount ~ to compensate for the effect of the lateral displace-
ment of the ultrasonic beam. The value of angle 8 r which gives an
amplitude minimum and phase reversal in the reflected signal is
measured. From angle 8 r is it possible to calculate the velocity
of the leaky surface wave Vr = Vo /sin8 r • whereas the attenuation
can be found from the amplitude of the reflected signal. 22

All the above remains valid for the case of a coated substrate.
320 S.1. ROKHLlN

A B

Figure 10. a) Beam reflection at angle e s in the case of a thin

coating (As »h). b) Beam reflection at angle 8 in the case of a
thick coating (As < h). s

From the measured angle 8s which gives a minimum of the reflection

coefficient the velocity of the surface wave Vs = Vo/sine s is cal-
culated. For a thin coating with thickness h « As parameter
Eo/(1-v 2 ) can be determined using Equation (5). In the case of a
thin mu~tilayered coating one can use Equation (6). Figure lOa re-
presents schematically the reflection of the beam near angle 8 s •
The curve E~ illustrates the penetration of elastic energy into the
solid body ~corresponding to the depth of penetration of energy in
the surface wave, Figure 1). Since the film thickness is small,
the principal part of the elastic energy is localized in the sub-
strate.

The situation is different for a thick coating, h » As'

Figure lOb. In this case almost all elastic energy is concentrated
in the coating and the effect of the substrate is insignificant.
The value of Vs = Vo/sine s in this case approaches the velocity of
Rayleigh waves in the coating material.

The Rayleigh angle method was employed by Jonath 23 for deter-

mining the elastic properties of an adhesive film on a substrate.
The details of the experimental technique and the experimental ap-
paratus have been given by Knollman and Hartog. 24 It was pointed
out that for materials such as adhesives, the amplitude minimum was
CHARACTERIZATION BY ULTRASONIC WAVES 321

impossible to measure at frequencies above 10 MHz. Thus, the phase

reversal phenomenon was used for determining the Rayleigh angle. By
measuring layers of different thicknesses,23,24 the reduction of
the shear modulus was found close to the adhesive-adherend inter-
face. More recently Knollman and Hartog showed 25 that adherend
surface treatments do not affect the gradient in the adhesive inter-
facial zone, whereas cure time (and/or temperature of cure) may
play an important role. Samples cured more than twice display25
a near constant shear modulus throughout the interfacial region.
In determining the surface wave velocity the effect of the substrate
was neglected, i.e., it was implied that As « h.

The Rayleigh angle method was used by Adler and McCathern 26

for measuring elastic properties of coatings with thickness of the
order of or less than the wavelength. A comprehensive theoretical
analysis was carried out for this case by Chiment, Nayfeh and
Buter. 27

A new technique based on the effect of nons~ecular reflection

at the Rayleigh angle was developed by Weglein 28 , 9 using the acous-
tic microscope.

ELASTIC INTERFACE WAVES

The term elastic interface waves is used to denote waves pro-

pagating along the interface between two media and localizing energy
in a band with a width of the order of several wavelengths.

An interface wave of a special type, termed the Stoneley wave,

may propagate along the interface between two elastic half spaces
in rigid contact with one another (Figure lla). The velocity of
the Stoneley wave should be lower than the velocities of the shear
waves in the half spaces. This condition imposes rigid limitations
on the possible pair of half spaces (on the ratio of their elastic
moduli and densities 6 ,30). In the general case the Stoneley wave
is a leaky wave and its energy is radiated in the form of a shear
wave into the half space with lower shear wave velocity (slow half
space), analogously to what is shown in Figure 6 for surface leaky
waves.

Lee and Corbly31 observed experimentally the gradual transfor-

mation of a Rayleigh wave into a Stoneley wave due to the strong
pressing together of two solids. Similar effects were measured by
Claus and Palmer 32 by an optical method with direct recording of
the Stoneley wave. Rigid contact at the interface was obtained by
moderate compression of optically polished surfaces. Recently
these results were theoretically analyzed by ~umar and Murty33 with-
in the framework of a phenomological model 34 of a loosely bonded
interface.
322 5.1. ROKHLlN

(a) (b)

RIGID CONTACT SLIP CONTACT

PI }J-I AI PI }J-I AI
STONELEY
WAVE Vs
- _ _~
L ~
y
X'
_ -INTERFACE
WAVE Vp

BOUNDARY CONDITIONS
(J Iyy =(J2yy; (J~x =(J~x (Jlyy=(J2yy; Uly= U2y
I
Uy =Uy;
2 I 2
UX=UX (J yx = 0 ;a2yx::.o

Figure 11. Different boundary situations at the solid-solid inter-

face. a) Rigid contact (continuous stresses and displacements).
b) Slip contact (shear stresses vanishing at the interface).
c) Viscoelastic layer at the interface between two solids.
CHARACTERIZATION BY ULTRASONIC WAVES 323

GUICEO
INTERFACE WAVE V?

0.6

04 .AL-Bi
LEAKY INTERFACE
0.2 WAVE

O~--~--~~~~~
I 2 4 6 810 20

Figure 12. ~xistence conditions for the interface wave on a solid-

solid interface with slip boundary conditions. (Based on data from
Lardat, Minot, Tournois 30 .)

The second limiting case of an interface wave is that of slip

contact at the interface (Figure llb) (frictionless contact). This
type of boundary condition is characterized by absence of shear
stresses at the interface, and corresponds to the presence of an in-
finitesimally thin layer of an ideal liquid at the interface. The
case of a liquid laver at the interface of solids was discussed by
Staecker and Wang. 3$ It was found that the conditions for existence
of this kind of interface waves are much less rigorous than the con-
ditions for the existence of Stoneley waves. Figure 12 shows, in
coordinates Vt /V t vs. Pl/P2' the domain of existence of guided
interface wave~, o~tained on the basis of results of Lardat, Minot
and Tournois. 36 The term guided points to localization of the elas-
tic energy near the interface. Outside of this domain the interface
wave is a leaky wave.

A more general case of the existence of interface waves was

recently studied by Rokhlin, Hefets and Rosen. 37 They analyzed the
case of propagation of interface waves in a system of two half
spaces separated by a viscoelastic layer (Figure llc). Shear modulus
324 S.1. ROKHLlN

DIFFERENT ADHERENDS IDENTICAL ADHERENDS

Vi >Vt2
T
I INTERFACE
WAVE LEAKY Vi<Vt
INTERFACE
Vi> Vt2
INTERFACE
I Vi< Vt2 WAVE GUIDED

WAVE LEAKY + INTERFACE

WAVE GUIDED
Vt2 T-~~~2----
INTERFACE
WAVE GUIDED

Figure 13. The conditions for the existence of an interface wave

for the case of two half spaces separated by a viscoelastic layer.

~o of the layer was assumed to be smaller than the shear modulus of

the half spaces. Such a model satisfactorily approximates an ad-
hesively bonded specimen, provided that the adhesive thickness is
much smaller than that of the adherends. The conditions for the
existence of an interface wave for this case are clarified in
Figure 13. The interface wave exists as a guided wave, when its
velocity is lower than the shear velocities in the half spaces. If
the shear modulus of the adhesive is lower than the shear modulus
of the half spaces, then in the case of identical half spaces there
always exists a guided interface wave.

The deformation of a thin solid interface film by the interface

wave is illustrated schematically in Figure 14. Figure l4a shows
the deformation produced by the surface wave in the subsurface layer.
It shows tension and compression regions separated from one another
by a half wavelength. Figure l4b shows the deformation of an inter-
face layer, the thickness of which is much smaller than the wave-
length. Due to antisymmetry of motion, the interface layer is sub-
jected to shear strain (Figure 14c).

A possible method for obtaining interLv2e waves is shown sche-

matically in Figure 15. The surface wave, excited in the lower sub-
strate, is transformed in the interface region into an interface wave.
CHARACTERIZATION BY ULTRASONIC WAVES 325

DILATATION
~
~ ::C)
SUBSTRATE-2

~\\\::'1/ Zlf SUBSTRATE

(b)
-,

- - ~OHESIVE
(c)

Figure 14. a) Deformation of the surface of a solid body upon pro-

pagation of a Rayleigh wave. b) Deformation of an interface film
upon propagation of an interface wave. The thickness of the inter-
face film is much smaller than the interface wavelength. c) Shear
strain of the interface film. The figure corresponds, on a magni-
fied scale, to element C singled out in Figure 14b.

SAW SAW
TRANSMTTER RECEIVER

FILM
t2
SUBSTRATE - ~

Figure 15. Illustration of the interface wave method.

326 5.1. ROKHLlN

Due to the closeness of velocities of the surface and interface

waves, only a small part of the energy is transformed into bulk
waves. The interface wave leaving the interface zone is retrans-
formed into a surface wave, and is sensed by a receiver.

One can estimate the viscoelastic properties of the interface

layer by measuring the velocity and attenuation of the interface
wave. We showed 37 that the complex shear modulus of an interface
wave ~o is related by a simple expression to the interface wave ve-
locity V.•
1

(7)

where a.. = Vt/Vi is the normalized wave number for thE: 2 interface
wave; al = a. 2 -SY; a2 = 2Sy-W; W = a. 2+S 2 ; S = (0. 2 _1)1/ ; y =
(a. 2 _vE/v2)1/2; ~r = (W2-4a. 2 SY); ~r is the characteristic function
for the ~ayleigh wave, Vt and V2 are the shear and longitudinal
wave velocities in the substrate, Po is the density of the inter-
face film material, p and ~ are the density and the shear modulus,
respectively, of the substrate, f is the frequency and h = 2rrfh/V t
is the nondimensional film thickness, 2h is the thickness of the
film.

Equation (7) is valid at h «A~, where AO is the length of

shear waves in the material of ~he interface film. When this con-
dition is not satisfied, shear modulus ~o' determined from Ejuation
(7) can be refined on the basis of the more exact equations. 7 To
determine the complex shear modulus ~o' one should substitute into
Equation (7) the complex wave number a., which is calculated on the
basis of measured velocity and loss factor of interface waves.

Numerical analysis shows that the sensitivity of this method

is high in the range of film thicknesses which are small as com-
pared with the length of interface waves. Hence at film thicknesses
from 100 to 1 ~m the measurements should be performed at relatively
low ultrasonic frequencies from 0.5 to 5.0 MHz.

It is possible to single out the principal features of the

method, which indicate that the use of the interface waves for the
evaluation of thin interface layers (in particular adhesively bonded
structures) is promising.

1) The interface wave produces shear stresses at the interface;

these stresses are most sensitive to variations of adhesion quality.

2) The interface wave propagates along the interface and hence

is sensitive to small changes in the properties of the adhesive and
CHARACTERIZATION BY ULTRASONIC WAVES 327

ALUMINlJ.I1-WATER GAP
/

10-2

FREQUENCY· THICKNESS (MHz-mm)

Figure 16. The effect on the transmission of normally incident

longitudinal waves of a thin water gap in metal.

of the bond between the adhesive and the adherends.

3) The interface wave can be used for evaluation of very thin

layers, when 2h/A t < 0.01, where 2h is the thickness of the inter-
face layer and At is the length of the shear wave in the substrate.
Conversely, bulk longitudinal waves, usually employed for evaluation
of adhesively bonded structures, are insensitive to the existence at
the interface of a thin liquid layer which exhibits no shear resis-
tance. This is illustrated by Figure 16, which shows the effect of
a thin water gap in a metal on the transmission of normally incident
longitudinal waves. When the gap widths is too small, the trans-
mission is complete, i.e., the bulk longitudinal wave is insensitive
to the presence of a thin film of a nonviscous fluid on the inter-
face. At the same time such an interface exhibits no shear resis-
tance. As was previously noted, the existence of such a break in
material continuity is reliably detected by interface waves.

4) As shown below, interface waves can be used for evaluation

of multilayered interface films, which simulate the adhesion
328 5.1. ROKHLlN

properties at the adhesive-adherend interface.

5) In the case when the thickness of the adhesive layer is

much greater than the wavelength, a leaky interface wave can be ex-
cited at the adhesive-adherend interface. This wave can, in prin-
ciple, be used for evaluating the properties of the adhesive-adherend
interface, provided that its attenuation, associated with the leakage
of energy, is not too high.

Figure 17. Hydrodynamic boundary layer in viscous liquid near the

oscillating plate.

APPLICATION OF INTERFACE WAVES FOR INVESTIGATING THE CURING AND

VISCOELASTIC PROPERTIES OF ADHESIVES

In the course of polymerization of adhesive films the latter

undergo a transition from the 1iquid,to the solid state. Let us
first consider qualitatively the behavior of an interface wave in
the course of polymerization of the interface film.

Let us assume that two elastic substrates are separated by a

thin layer of a liquid adhesive. The thickness 2h of this layer
will be taken to be much smaller than the wavelength A of the inter-
face wave. The fluid in the interface gap can be regarded as ideal,
CHARACTERIZATION BY ULTRASONIC WAVES 329

provided that the gap thickness is greater than the thickness of the
hydrodyqamic boundary layer,37 0 = (vo/w)1/2 or 0 = IIm(~o) 11/2/
[w(po)1/2], where Vo = IIm(~o) I/(pow) is the kinematic viscosity of
the fluid, ~o = ~~ - i~~ is the shear modulus, w is the angular fre-
quency, and Po is the density of the adhesive film. So we have
frictionless or slip contact at the interface between the two solids
(Figure lIb). In the course of interface-film polymerization the
viscosity of the adhesive rises, which increases the thickness of
the hydrodynamic boundary layer. When the boundary-layer thickness
is comparable to the thickness of the film, the shear stresses start
being transmitted through the film from one substrate to the other.
This effect is illustrated in Figure 17, which shows how the viscous
fluid is set into oscillation with frequency w due to vibrations of
the lower plate. Displacement u falls across the liquid layer in an
exponential manner. As viscosity rises (increasing 0) or the film

3100 _____________ v:'"

SleeI-AL INTERFACE
FILM THCKNESS 251'
MlXlNG RATO - 5 I

u
Q)

'"
"'-
E
>-
t: 2950
U
0
..J
W
>
2900 ______________ v:1.

100 200 300

TIME (min)

Figure 18. Changes in the phase velocity Vi of the interface wave in

the course of curing of the epoxy film at the steel-aluminum interface.
330 S.I. ROKHLlN

thickness 2h falls, the upper plate starts "feeling" the vibrations

of the lower plate. This results in a rise in the phase velocity of
the interface wave due to the rising resistance of the interface to
shear strain.

This means that solidification of the interface film is accom-

panied by an increase in the phase velocity of the interface wave
which increases from the velocity V~ to some velocity V., controlled
by the interface-film parameters. 1

These considerations are illustrated by the experimental re-

sults shown in Figure 18, which shows the variation in the phase
velocity Vi of the interface wave in the course of polymerization of
a thin (25 ~) epoxy film at the interface between a steel and an
aluminum slab. In its initial state the epoxy film viscosity is
low and the velocity of the interface wave is Vi ~ V9, where vi is
the velocity of the interface wave at the interface Eetween these
half spaces at frictionless contact. This velocity lies between
the ve.1ocities of Rayleigh waves for the half spaces (VAl < v9 <
v~tee1). The velocity of the interface wave rises in t~e coijrse of
the film's polymerization and becomes greater than v~tee1. However,
its value still does not attain the shear rate of the "slower" of
the half spaces V~ and, hence, in this case, the interface wave is

~ 3.0
<:>
I - --- mlxinll rolio 6 "} Sleel-Al
>13>~ _ _ mixinll rotio 5 :1 2h'25)J 14
• T. L. ........ AV%
~

,., .
2..5
j~
III
Ii) 12 oog
ILl
(!) II)

~ 2.0 10 ILl
II)
:I: II)
l)
0
~ 8 ...J
I-
0 Z
0 0
...J 6 II)
ILl II)
> 1.0 :i
0 II)
ILl 4 Z
N
:J
« 0
:
I-
:::I: AVR "!oWeel) 2
a:
~
0
50 100 150 200 250 300
TIME (min)

Figure 19. The change in the relative velocity ~V (0) and the trans-
mission losses (x) of the interface wave as a function of time of
curing. Stee1~aluminum interface; two mixing ratios of epoxy resins.
CHARACTERIZATION BY ULTRASONIC WAVES 331

....
:$?
0 4.0
~
I
>IV 3.5
"
G 3.0
(f)
UJ
<.:>
z
J! 2.5
u
>-
f- 2.0
g
.J
UJ
> 1.5
0
W
N
::J 1.0
<l
~
~ 05
z

50 100 150 200 250 300 350

(a) TIME (mi n )

2.5

E
u
......
CD
B
(f)
UJ
(f)
(f)
0
.J
Z
Q
(f)
(f)

~
(f)
Z
<l
a::
f-
50 100 150 200 250 300 350
(b) TI ME (min)

Figure 20. a) The change in the relative velocity, and b) the

transmission losses of the interface wave as ~ function of time of
curing for various temperatures. Both adherends are steel. Vr is
the Rayleigh wave velocity. (From Rokhlin, Hefets, and Rosen 3/ .)
332 S.I. ROKHLlN

always a guided wave. If the shear modulus ~o of the film is in-

creased, then the velocity of the interface wave may intersect the
line of Vi = V~ and the interface wave becomes a leaky wave.

The measured velocity and attenuation of an interface wave can

be used for estimating changes in the complex shear modulus in the
course of curing of a thin interface film.

Figures 19 and 20 show the measured relative changes in phase

velocity and transmission losses of the interface wave as a function
of the polymerization time of the interface layer at a frequency of
1.2 MHz. The adhesive used was EPOFIX, produced by Scientific In-
struments, Denmark; this adhesive polymerizes in several hours after
addition of a solidification agent. The static viscosity of the
starting mixture was about 550 cp at 25°C.

The data in Figure 19 were obtained at 25°C for two different

mixing ratios of the adhesive components. The film thickness was
25 ~, the lower adherend was steel, the upper - aluminum. The data
in Figure 20 were obtained for several temperatures at a mixing
ratio of 5:1; the film thickness was l2~. Both the upper and lower
adherends were steel.

The experimental data can be smoothed and translated 37 to the

complex wave number of the interface wave, and then to the complex
shear modulus of the film. Thus, the calculated complex shear modu-
lus of the film normalized to the steel shear modulus is plotted
against the time of polymerization (Figure 21).

As a first approximation the shear modulus was calculated from

Equation (7). These values were used as the initial guesses for the
numerical solution 37 of the exact equations. The moduli obtained
from the exact equation and from Equation (7) are identical to with-
in the resolution of the plot. The dashed horizontal line in Figure
21 represents the normalized dynamic film shear modulus after com-
plete polymerization. It was obtained by direct measurement of the
bulk shear velocity, and is the asymptote of the curves of Re(~ /~)
o
at the measured temperatures.

The velocity and attenuation of the interface wave depend on

the elastic and dissipative properties of the film and half spaces.
In particular, the maximum on the transmission loss curve and the
point of inflection on the velocity curve of the interface wave
characterize the entire system (film and substrate) as a whole.
Hence the maximum on the transmission loss curve is shifted relative
to the maximum on the curve of the imaginary part of the shear modu-
lus in the adhesive material. The viscoelastic properties of the
interface film are determined by calculations from experimental data.
The measurements are performed at ultrasonic frequencies in the 1
MHz range, which requires consideration of the correspondence between
CHARACTERIZATION BY ULTRASONIC WAVES 333

0.015 OJ r -----
fLO / 300C
(/)
;j
JT""
--1
;j
0
0
:::i!
0.01
a::
<!
w
3i
o
w
!:::! 0.00
--1
<l
:::i!
a::
oz
aoOI~~~~~~~--~--~~~~~
o 50 00 150 200 250 300 350
TIME (min)

Figure 21. The normalized complex shear modulus ~o/~ calculated

from smoothed measured values of the interface wave velocity and
the transmission losses (Figure 20). The dashed horizontal line
represents direct measurements of the normalized film shear modulus
~~/~ after full polymerization. The steel shear modulus was taken
as ~ = 0.809'10 12 dyn/cm2 • (From Rokhlin, Refets and Rosen 37 .)

the static and dynamic viscoelastic quantities in the course of

curing. In particular, one must consider the kind of correspondence
between the change in the interface wave velocity and increasing
strength in the course of curing of the given adhesive.

Such a comparison was performed for the FM-73 structural ad-

hesive. 38 The measurements were performed for a standard cure cycle
at a frequency of 0.5 MHz. The velocity of the ultrasonic interface
wave was measured during the entire cure cycle as a function of time.
The bond strength was measured over different time intervals from
the start of the cure cycle. In order to immediately stop the cross-
linking reaction, the shear strength of the specimen was measured
when the latter was rapidly cooled to room temperature.

Figure 22 compares the shear strength data obtained upon rapid

cooling with results obtained by ultrasonic measurements. The open
rectangles in this graph were obtained by parallel translation of
334 S.I. ROKHLlN

130
• ~S( SHIT

O!i

00
• TUftRULft
/1
,/ I
"I
.0
[J 8[J [J
00
• . .0
a"

_/ I
•
•
f"
03 I 0

• 1 Slll.1ICl1I (~)
Q2

QI
0
•
0
... 0
00·
I,
I •
.- ~1
••
RoIPIO CCCI.ftC.
MIA FOt
stOW cro.tHG 0
(mEltSllIrIl

I
0
10 20 30 40 50 60 70 so
T IME(min)

Figure 22. Comparison between the change in the interface wave

velocity and the shear strength of the bon(j in the course of curing
FM-73 structural adhesive.

the corresponding point for the shear strength, obtained for slow
cooling. It is seen that a rise in the velocity of the interface
wave (rise in the shear modulus of the adhesive) corresponds pre-
cisely to the time interval of the bond-strength growth.

The effectiveness of the ultrasonic method lies in the fact that

it is a nondestructive technique, which shows the growth of the ad-
hesive strength as a function of time for a single cure cycle. Ob-
taining analogous data by mechanical testing requires a large number
of experiments with different specimens, corresponding to different
times in the cure cycle. Mechanical testing is particularly time
consuming when one wishes to determine the effect of different para-
meters, such as temperature, pressure and polymerization time on
bond strength. The use of the ultrasonic interface wave technique
as a supplement to mechanical testing can significantly reduce the
volume and time of the experimental program.

These experimental results were obtained under nonisothermal

conditions. Temperature changes modify the elastic moduli of the
adhesive film and substrates and, by virtue of thermal expansion,
CHARACTERIZATION BY ULTRASONIC WAVES 335

DIRECTION OF
TENSION

Figure 23. Schematic ot the specimen for differential ultrasonic

measurement and strength testing.

change the density. This results in a change in the velocity of

the interface wave. In addition, the thermal expansion of the sub-
strate modifies the acoustical path length.

To measure the properties of the film under nonisothermal con-

ditions, it is necessary to eliminate the effect of temperature
variations in the substrates and transducers. For this purpose 39
a differential measuring arrangement, in the form of an acoustic
bridge, was developed (Figure 23). The middle transducer (trans-
mitter) radiates an ultrasonic surface wave symmetrically into two
arms of the bridge. In the left arm, the surface wave incident on
the boundary between the free surface of the lower substrate and
the interface region, is transformed into an interface wave. After
leaving the interface region, the interface wave is transformed into
a surface wave, and is sensed by a receiver. A surface wave propa-
gates along the entire right arm. An offset block which is identi-
cal to substrate 2 is placed without bonding on substrate 1, in order
to equalize the thermal masses of both arms (Figure 23). Due to
surface roughness there is no acoustic contact between the substrate
and the offset block and it does not affect the propagation of the
surface wave.
336 S.1. ROKHLlN

1.0
u : .~ . I .
'"'" , -HEATING . . . .. .

~
'-COOLING '. "

~
..,J0.5
w
...l
0
w
~
r- ."
..
0
25 50 75 100 125 150 175 200
TEMPERATURE (OC)

Figure 24. Example of the measurement of thermomechanical properties

of an FM-73 adhesive interface film. Phase-delay changes of t .he
interface wave due to the temperature cycling of an adhesive-bonded
specimen.

If the dimensions of both arms of the bridge and of the trans-

ducers are the same, then the use of the differential arrangement
eliminates the effect of thermal changes of the acoustical path.
It is clear that the difference in the delay time between the two
channels will depend slightly on temperature changes in the sub-
strates, since the equations expressing the velocities of the sur-
face and the interface waves as a function of the properties of the
substrates are not entirely identical. Thus, there arises the ques-
tion of the accuracy of temperature compensation in the bridge. The
desired estimate cannot be obtained experimentally, since the film
properties are also a function of the temperature (particularly in
the case of polymer interface films). The estimate was obtained
analytically in Reference 39, where the experimental arrangement
was also described in detail.

The method is illustrated by the determination of the thermo-

elastic properties of the FM-73 structural adhesive as shown in
Figure 24. The relative phase change of the interface wave at 0.5
MHz during the temperature cycle of the cemented aluminum specimen
is shown. The bond line thickness was 90 ~m. An increase in the
delay time corresponds to a reduction in the shear modulus of the
adhesive. These changes correspond to the ordinary thermomechanic
CHARACTERIZATION BY ULTRASONIC WAVES 337

behavior of the polymer, the mechanical properties of which are

strongly temperature dependent. At high temperature the polymer
exhibits a low-modulus, rubber~type behavior, whereas at low tem-
peratures it behaves as a high-modulus glass-type material.

THE CONCEPT OF THE EFFECTIVE SHEAR MODULUS - STRENGTH PREDICTION

An adhesive bond may fail in two ways: 1) Failure inside the

adhesive, which is termed cohesive failure. The cohesive strength
of a given adhesive correlates with its elastic modulus. 2) Failure
along or close to the adhesive-adherend interface, which is termed
interfacial or adhesion failure.

The cohesive strength of a given type of adhe~ive is governed

by its chemical structure and by the conditions of the curing pro-
cess (in particular, by the percentage of crosslinking), and by the
presence of various kinds of microdefects (e.g., microcracks or
voids). These quantities also determine the shear modulus and the
ultrasonic loss factor of the adhesive. By virtue of this fact,
for a given type of adhesive, a change in the elastic modulus cor-
relates with a change in its strength.

There is no rigorous physical theory of interfacial failure.

From the phenomenological point of view,40,4l interfacial failure
can be identified with cohesive failure of a weak boundary layer
(WBL). The WBL can be represented by any low-strength region at the
adhesive-adherend interface, and if the thickness of the WBL tends
to zero, then the interface proper can be regarded as the WBL.
.
I twas s h own exper~menta 11y 23-25,42 t h at t h e v~scoe
. 1ast~c
. pro-
perties of the adhesive are not uniform over its thickness, and that
a reduction in the elastic modulus may occur at the adhesive-ad-
herend interface. It is hence convenient to consider a multilayered
model of a bond line, consisting of an adhesive and WBL (Figure 25).
It will be assumed that WBL characterizes the adhesion properties
between the adhesive and the adherend.

According to this model, the shear modulus of the adhesive, cal-

culated on the basis of the measured velocitv and attenuation of in-
terface waves is the effective shear modulus Peff. It character-
izes the effective elastic properties of the multilavered adhesive-
WBL system. The concept of the effective dynamic shear modulus of
the bond line was introduced by Rokhlin, Hefets and Rosen. 43

The qualitative description of the properties of the interface

wave in the presence of a-WBL can be given on the basis of the matrix
method. 44,45 We obtained a series expansion of the exact solution
with respect to the thickness of interface films and compared this
solution with the solution for the case of a single film. 37
338 S.1. ROKHLlN

p, fL,>"

po, fLo,ho )

p, fL,>"

Figure 25. A multilayered model of a bond line, consisting of an

adhesive and weak boundary layers.

According to this comparison, we obtained Equation (7) for a single

interface film when only linear terms are retained in the matrix
expansion. A similar approximation for the multilayered system
(substrate/WBL/film/WBL/substrate) makes it possible to write the
characteristic equation for the velocity of the interface wave in
the form

- -Peff -2
II
eff
III = hi'. r I(S-h
- - a -2ha. a )
P 1 2
(8)

where

h h
o
+ hw

and

(9)
CHARACTERIZATION BY ULTRASONIC WAVES 339

2ho - is the thickness of the adhesive layer. hw' Pw' llw - are the
thicknesses, density and shear modulus of the weak boundary layer.
The other terms have the same meanings as in Equation (7).

When contact between an adhesive and substrates is not ideal

the interface wave velocity decreases. Hence, the effective shear
modulus lleff found from Equation (8) on the basis of the experimen-
tally measured velocity will be smaller than the actual modulus of
the film.

For example, in the case of absence of shear bonding between

the adhesive and at least one of the substrates (slip contact) the
velocity of the interface wave will be close to the Rayleigh wave
velocity and the calculated lleff will be equal to zero. It is thus
seen that the interface-wave velocity will characterize not only
the elastic properties of the adhesive and its cohesive strength,
but also the bond between the adhesive and the substrates, i.e., the
interfacial (adhesion) strength. This means that the ratio lleff/llo
can be used as a criterion of the bond strength. If the effective
shear modulus lleff is measured for a given bond, and the shear
modulus II of the adhesive is measured on a reference specimen, then
Equation ~9) can be used for estimating45 the properties of WBL.

For the study of the correlation between the strength of the

bond and the effective shear modulus,43 the metal-adhesive interface
was weakened by modification of the surface of the substrate by a
thin lubricant film. The velocity and attenuation of the interface
wave were Deasured in the course of polymerization of the adhesive.
After completion of the polymerization process the specimen was
tested in order to determine the shear strength of the bonu. Figure
26 shows a normalized plot of the shear strength of the bond vs. the
effective shear modulus measured by the interface-wave technique.
The general transmission loss factor D

shown in Figure 27 serves as another parameter correlating with the

shear strength. Here ~ is the amplitude of the transmitted signal,
read in all the experiments at the time corresponding to the loss
maximum in reference bonds, Ao is the amplitude of the signal at
the initial stage of the polymerization, and AF is the amplitude of
the signal at the final stage of polymerization. Quantity AF makes
allowance for scattering of the interface by micro-inhomogeneities
of the adhesive layer and of adhesive-adherend interfaces. It was
shown by Claus and Kline 46 and Claus and Palmer47 that the attenu-
ation of interface waves is significantly affected by the quality
of the surface topography of the adherends .48 The correlation between
D and strength shows that information on the final adhesion strength
340 S.1. ROKHLlN

et::
<t 0.8
w
J:
(f) i= 0..6
ot!)

1::l~ 0.4
...Jet:: +
~ tii 0.2 "
a:
~ 0. 0.2 0.4 0.6 0.8 t
NDRMALIZED EFFECTIVE
MODULUS

Figure 26. Relation between the normalized effective shear modulus

and the normalized shear strength (from Rokh1in, Hefets, and
Rosen 43 ).

J:
l-
t!)
Z
W

~ 0.8
0::
<t
W 0.6
i7j
o 0.4
w
N
:J 0.2
<t
~
§ 0 5 10 15 20 25 3D
GENERALIZED TRANSMISSION LOSS ~CmR{dB)

Figure 27. Relation between the generalized transmission loss fac-

tor and the normalized shear strength (from Rokh1in, Hefets, and
Rosen 43 .)
CHARACTERIZATION BY ULTRASONIC WAVES 341

of the bond is present at the intermediate stages of the cure (re-

lation between the strength and A ).
m

Let us clarify the physical basis of the above correlation be-

tween the parameters of the interface wave and the bond strength.
At each of the points along the wave path the velocity and the at-
tenuation of the interface wave are characterized by the adhesion
properties of the metal-adhesive interface and the elastic properties
of the adhesive proper. The measurements were performed along an
interface segment of length L. Hence the velocity and attenuation
are integral characteristics of the effective elastic properties of
the interface not only over the thickness of the adhesive but also
along the total acoustic path. The velocity represents resUlts ob-
tained by averaging the velocities at two different lengths: 1) at
segments with good adhesion, the velocity which is equal to the
velocity of the interface wave in the reference specimen; 2) at seg-
ments with inferior adhesion, the velocity which is lower, and lies
between the velocity at the reference specimen and the velocity of
the interface for slip contact. Hence the total velocity decreases
as compared with that at the reference specimen.

:
l-
t!)
Z
.... .
a::
I-
(I) 0.8 ,.~
0
z
0
CD
0.6 Ll.....
....
0
N
0.4 "" ~.~
o CON-SILICONE
:::::;
c
=e Q2 GREASE .
a::
0
z
0 100
% INTERFACIAL FAILURE
Figure 28. The normalized shear strength of the bond as a function
of the percentage of interfacial (adhesion) failures (data from
Smith and Smith 49 .)
342 S.I. ROKHLlN

The mechanical shear-lap joint test is performed perpendicular

to the direction of propagation of the interface wave. Since the
interface regions with inferior adhesion have a lower shea~ stiff-
ness the effective width of the lap joint decreases. Hence the
critical value of the failure stress at good parts of the interface
is attained at lower test forces as compared with the reference
specimen. The transmission of shear forces from one adherend to
another through the adhesive film is controlled by the effective
shear modulus. It is therefore natural to expect a correlation be-
tween the interface wave velocity and the bond strength.

The above is confirmed by the results of Smith .<lnd Smith. 49

Figure 28 shows the relation obtained by them between the shear
strength of the bond and the percentage of the overall surface of a
specimen with an interfacial (adhesion) mode of fracture. As in
our experiments, the ~dhesion strength was modified by introducing
specified amounts ot impurity to the adherend surfaces prior to
bonding. It appears to us that the correlations shown in Figures
28 and 26 are well supported and complement one another.

CONCLUSION

The surface wave method has only recently been introduced for
investigating the properties of thin films. The surface wave pro-
duces longitudinal strains in the film and its velocity is related
to Young's modulus of the film. Although most of the results until
now were obtained for metal films, it is clear that this method
should be useful for investigating photoresist materials and the
curing of polymeric coatings. The Rayleigh-angle method appears
promising for investigating solid polymer coatings.

Unlike surface waves, interface waves were immediately utilized

for the study of adhesively bonded structures. The interface wave
produces shear strains in the interface layer and it is hence sensi-
tive to the state of the adhesion contact between the film and the
substrates. Interface waves can play an important role in in situ
study of the curing process and of viscoelastic properties of ad-
hesives and for strength prediction of adhesive bonds.

The difference in the form of elastic deformations produced by

surface and interface waves result in a number of basic differences
in their applications:

1) Surface waves can be used for measurjng Young's modulus,

whereas interface waves can be utilized for obtaining the shear
modulus.

2) The sensitivity of the surface wave method to film proper-

CHARACTERIZA liON BY ULTRASONIC WAVES 343

ties decreases with reduction in film thickness. In the interface

wave method the sensitivity can be high particularly for small
thicknesses cf interface films.

3) The effective stiffness for a multilayered surface film is

obtained as a successive combination of the stiffnesses of the in-
dividual films, whereas the effective modulus in a multilayered
interface film is equivalent to a parallel combination.

4) Interface waves are more sensitive to the state of adhesion

at the film-substrate interface, since the weak coupling "shunts"
the elasticity of the system as a whole. At the same it was shown
that reduction of the adhesion between the film and the substrate
can also be observed by the surface wave method.

REFERENCES

1. E.P. Papadakis in "Physical Acoustics," W.P. Mason and R.N.

Thurston, Editors, Vol. 12, p. 277, Academic Press, New York,
1976.
2. A. Vary in "Mechanics of Nondestructive Testing," W.W. Stinch-
comb, Editor, p. 123, Plenum Press, New York, 1980.
3. E. Segal and J.1. Rose in "Research Techniques in Nondestruc·-
tive Testing," R.S. Sharpe, Editor, Vol. 4, p. 275, Academic
Press, London, 1980.
4. G.W. Farnell in "Surface Wave Filters," H. Matthews, Editor,
p. 20, J. Wiley & Sons, New York, 1977.
5. I.A. Viktorov, "Rayleigh and Lamb Waves," Plenum Press, New
York, 1967. r

6. B.A. Auld, "Acoustic Fields and Waves in Solids," Vol. II,

J. Wiley & Sons, New York, 1973.
7. J.D. Achenbach and S.P. Keshava, J. Appl. Mech., 34,397 (1967).
8. H.F. Tiers ten , J. Appl. Phys., 40, 770 (1969). --
9. S.!. Rokhlin, Z. Ronen, and M.p-.-Dariel, Thin Solid Films, 89,
109 (1982).
10. M. Toguchi and F. Akao, Ultrasonics Symposium Proceedings,
IEEE Cat. 79CMl482-9SU, 439 (1979).
11. M. Levy, H. Salvo, Jr., D.A. Robinson, M. Maki and M. Tachiki,
Ultrasonics Symposium Proceedings, IEEE Cat. 76CHl120-SSU, 633
(1976) •
12. D.R. Snider, H.P. Fredricksen and S.C. Schneider, J. Appl. Phys.,
52, 3215 (1981).
13. 1: Feng, M. Tachiki, C. Krischer and M. Levy, J. Appl. Phys.,
53, 177 (1982).
14. E. Harnik and E. Sadar, J. Appl. Phys., 52, 3705 (1981).
15. G. Gorodetskv and I. Lachterman, Rev. Sci. Instrum.,~, 1386
(1981).
16. C. Weaver and D.T. Parkinson, Philos. Mag., ~, 377 (1970).
344 S.1. ROKHLlN

17. G.W. Farnell in "Physical Acoustics," W.P. Mason and R.N.

Thurston, Editors, Vol. 6, p. 109, Academic Press, New York,
1970.
18. W.G.N. Neubauer in "Physical Acoustics," W.P. Mason and R.N.
Thurston, Editors, Vol; 10, p. 61, Academic Press, New York,
1973.
19. H.L. Bertoni and T. Tamir, J. Appl. Phys., ~, 157 (1973).
20. A. Schoch, Acustica, 2, 17 (1952).
21. L. Adler, Revue du Cethedec NS80-2, 116 (1980).
22. F.L. Becker, R.L. Richardson in "Research Techniques in Nondes-
tructive Testing," R.S. Sharpe, Editor, Vol. 1, p. 91, Academic
Press, London, 1970.
23. A.D. Jonath in "Adhesion and Adsorption of Polymers," L.H. Lee,
Editor, Vol. A, p. 175, Plenum Press, New York, 1980.
24. G.C. Knollman and J.J. Hartog, J. AppL Phys., 53,1516 (1982).
25. G.C. Knollman and J.J. Hartog, J. AppL Phys., 53, 5514 (1982).
26. L. Adler and D.A. McCathern, J. AppLPhys., 49-,-2576 (1978).
27. D.E. Chiment, A.F. Nayfeh, D.L. Buter, J. App~ Phys., 53, 170
(1982). -
28. R.D. Weglein, Appl. Phys. Lett., 34, 179 (1979).
29. R.D. Weg1ein, AppL Phys. Lett., 35, 215 (1979).
30. R.N. Thurston, J. Acoust. Soc. Am.~64, 1 (1978).
31. D.A. Lee and D.M. Corbly, IEEE Trans. Son. Ultrason., SU~24,
206 (1977).
32. R.O. Claus and C.H. Palmer, Appl. Phys. Lett., 31,547 (1977).
33. V. Kumar and G.S. Murty, IEEE Trans. Son. Ultrason., SU~29,
138 (1982). --
34. G.S. Murty, Phys. Earth and Planet. Interiors, 11,65 (1975).
35. P.W. Staecker and W.C. Wang, J. Acoust. Soc. Am~53, 65 (1973).
36. C. Lardat, J.P. Minot and P. Tournois, IEEE Trans.:Son. Ultra-
son., SU~22, 16 (1975).
37. S.I. Rokh1in, M. Hefets and M. Rosen, J. Appl. Phys., 51, 3579
(1980) •
38. S.I. Rokh1in, J. Composite Mater., in press.
39. S.I. Rokhlin, J. Acoust. Soc. Am., in press.
40. R.J. Good in "Adhesion Measurement of Thin Films, Thick Films,
and Bulk Coatings," K.L. Mitta1, Editor, p. 18, American
Society for Testing Materials, Philadelphia, 1978.
41. . J. J. Bikerman in "Adhesion Measurement of Thin Films, Thick
Films, and Bulk Coatings," K.L. Mittal, Editor, p. 30, American
Society for Testing Materials, Philadelphia, 1978.
42. Y.S. Lipatov in "Adhesion and Adsorption of Polymers," L.H.
Lee, Editor, Vol. B, p. 601, Plenum Press, New York, 1980.
43. S.I. Rokhlin, M. Hefets and M. Rosen, J. Appl. Phys., 52, 2847
(1981). -
44. S.l. Rokhlin, (1981), unpublished.
45. S.l. Rokhlin and M. Rosen, Thin Solid Films, 89,143 (1982).
46. R.O. Claus and R.A. Kline, J. Appl. Phys., 50-,-8066 (1979).
47. R.O. Claus and C.H. Palmer, IEEE Trans. Son-.-Ultrason., SU~27,
97 (1980).
CHARACTERIZATION BY ULTRASONIC WAVES 345

48. R. O. Claus. and R.T. Rogt:~s in "Physicochemical Aspects of

Polymer Surfaces," K.L~ Mittal, Editor, Vol. 2, p. 1101,
Plenum Press, New York, 1983.
49. T. Smith and P. Smith in "Adhesion and Adsorption of Polymers,"
L.H. Lee, Editor, Vol. A, p. 123, Plenum Press, New York, 1980.
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS

T. C. Ward, Margaret Sheridan, and D. L. Kotzev*

Chemistry Department and Polymer Materials and

Interfaces Laboratory, Virginia Polytechnic
Institute and State University, Blacksburg,
Virginia 24061

Free films and lap-shear bonded metal joints

were compared with respect to their linear dynamic
viscoelastic behavior. Thermosetting and thermo-
plastic adhesives were investigated. Damping
phenomena and relative shear moduli indicated a
broader mechanical relaxation occurring at higher
temperatures in the bonded assemblies. Triblock
copolymer model adhesives were used to illustrate
the importance of sample preparation in the
evaluation of bonded materials.

INTRODUCTION

In terms of a total systems analysis approach to adhesive

bonding l there exists, among others, one clear role with which
polymer scientists might identify, namely, that of evaluating
proper material response functions and their connection to
joint failure events. This paper was motivated, to some
extent, by a desire to explore a particular response function,
the linear viscoelastic quantity tan 6, which is widely used
by polymer researchers to characterize primary and secondary
relaxations in bulk polymers, most notably the glass transi-
tion, Tg .

* Permanent Address: Scientific-Industrial Center for

Special Polymers, Zheljo Voivoda 4A, 1156, Sofia, Bulgaria

347
348 T.e. WARD ET AL.

It is important to recognize that the molecular relaxations

associated with maxima in the tan 0 function for glass
containing composites are highly time (or frequency) dependent
in any geometry of testing. These maxima indicate changes in
the response of the entire assembly undergoing testing; hence~
they may originate from a number of sources, but not from a
transition in the viscoelastic response of the polymer. Only
the kinetic events associated with the relaxations are
modified by a new time-temperature set of variables. At more
rapid strain rates, polymeric materials respond as if the
temperature had been lowered. Thus, it is well accepted that
quantities such as high speed impact strength of bulk polymers
are quite sensitive to test rate and temperature. In
particular, for amorphous glasses, such as epoxies, functions
based on the reduced temperature, T-T g , best fit the evidence
concerning the variation of impact properties on heating of
the polymer once test rates are specified. Hence, any
mechanically perceived alteration of the glassy relaxation is
of paramount importance in bonded joint performance.

One of the fundamental questions in adhesion science is

whether or not the results from a bulk property test will
reliably represent the response behavior of the same polymer
in a thin bond line (typical of the type encountered in a
structural adhesive application) and of the entire bonded
structure. Over the past few years investigations have been
underway in our labs which suggest that some caution is
appropriate when transferring test results obtained on free
polymer samples into situations where the polymer may be
(geometrically) under more severe constraints. Other workers
have cited data to indicate that adhesive joints often do not
perform as predicted from properties of the bulk polymer
adhesive. 2 However, some researchers 3 have reported good
correlation between in situ and bulk properties, on FM-73 in
the cited case. In ~r present paper, observations on
different bonded and free polymer adhesives are presented to
help clarify this situation, particularly .in the case of rapid
strain rates (110 Hz) being applied at low amplitude strains.

Overall, two major objectives were defined for this

investigation. First, nondestructive evaluation using very
small strain deformations of both the free and the bonded
adhesives was of interest. Although fracture and failure must
be accurately modeled in real systems by nonlinear mechanics,
it is generally appreciated that linear viscoelastic behavior
is indicative not only of the kinetics of molecular motions in
a Boltzmann superposition sense but may also be related to
many of the experimentally observed ultimate properties.
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 349

Furthermore, most of the design of polymer molecular

architectures is based on knowledge gained from examining
unsupported polymers. In other words, our ideas about
structure/property correlations in polymeric adhesives
originate many times from research on samples which, when
tested, differ physically from the configurations encountered
in actual applications.

A second objective of this work was to focus on bond

thickness and bond thermal and stress history vis-a-vis the
nondestructive type of test. There are many citations in the
literature which deal with adhesive bond line thickness 4 ,5, of
course, but usually with regard to ultimate properties; GlC'S
and time's-to-failure are good examples. Work on epoxy bonded
systems at the Naval Research Labs 6 ,7 over the years has
emphasized that, depending on testing rate and temperature,
mode I fracture energies may be optimized by selection of bond
thicknesses in the range of 0.02-0.07 em. The viscoelastic
natures of these adhesives was often cited in the models which
were formulated by these workers to explain the observed
performance of the bonds. Indeed, various master curves were
developed relating qualities such as tan 0 to test frequency
and temperature. Shifts of a few degrees in the Tg for CTBN
modified epoxies were found to vary the reduced frequency
(time) scale's of the unbonded adhesive by as much as a
decade. 7 From the point of view of the impact strength of an
adhesively bonded system, such shifts would be of major
significance. The occurence of such shifts, or changes of the
breadth and shape of the tan 0 vs temperature curves
themselves by virtue of the bonding process was also an item
to be explored in our work. Particularly when long term
environmental effects on adhesively bonded joints are of
interest, it is important to know how the composite joint
behaves viscoelastically relative to the bulk adhesive in
order that accelerated testing procedures might be developed;
the shifts. in relaxations would playa role in the analysis.

There is conflicting evidence on the influence of a rigid

substrate or filler on the small strain viscoelastic
properties of a thin bond line. Recent modeling has predicted
a shift in the relaxation maxima of tan 0 curves which was
attributed to chan~es in boundary layer properties of the
polymer molecules. Such modeling involves the application of
a three-phase system, effectively a composite, for predicting
conditions for the shift and resolution of relaxation maxima
of tan 0 curves. The three phases are the boundary-layer
polymer, the bulk polymer, and the filler. Calculation of the
glass transition temperature of filled polymers follows from
two assumptions. First, the boundary layer, having properties
350 T.C. WARD ET AL.

differing from the bulk polymer due to the action of the

filler surface, and the bulk polymer have different glass
transition temperatures. Second, the filler, whose
concentration determines the concentration ratio of boundary
layer polymer to bulk polymer, also affects the shape of tan 0
curves because of its high modulus. 8 Experimental work by
Lipatov et. al. 9 showed the addition of glass beads into epoxy
resin resulted in a shift in the glass transition temperature
to higher temperatures and a decrease in the maxima values of
tan O. Similar results were observed for epoxy resins filled
with quartz powder. For poly(butyl methacrylate) filled with
glass beads there was a considerable decrease and some
broadening of the relaxation maxima. However, there was no
shift in the glass transition temperature for this latter
system. Their results were explained in terms of a decreased
molecular mobility of the boundary-layer polymer due to the
effect of the filler.

The idea of an "interphase" of polymer with modified

properties has frequently been referred to in reports on
filler or substrate modification of polymer response near
Tg's.lO However, there is .comment in the literature which
seriously questions the ability of a high modulus surface to
influence macromolecules at any significant distance from the
interface. II Indeed, it is clear that because of the high
possibility of residual stress fields due to thermal or curing
operations in the preparation of filled or bonded polymer
systems it may be quite difficult to identify meChanistically
the origin of any anamolous behavior. Nevertheless, math-
ematical modeling of the interphase to allow for increases or
decreases in the Tg of a filled composite has now appeared. 12
Theocaris and Spath is predict that strong bonding between a
filler and its matrix material results in a higher composite
Tg , weak bonding leading in the opposite direction. 12

In summary of the above review, it is apparent that one

may not easily anticipate the outcome of a linear viscoelastic
test scheme on a bonded joint assembly, even if the polymeric
adhesive has well documented properties. There is one salient
feature of such comparisons, on the other hand, which must be
kept in mind. As was clearly pointed out recently concerning
the mechanical response of a nonhomogeneous viscoelastic
glassy composite undergoing heating, the relaxations (or
dispersions) of the entire assembly may mistakenly be solely
identified with macromolecular motions; whereas, in fact, an
interjection of instrumental factors into the observed
behavior may have occurred. 13 ,14 These papers on the torsion
braid experimental method emphasize that a fiber glass braid/
amorphous polymer composite exhibits quite different tan 0
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 351

functions of temperature at approximately constant frequency

than does the neat resin itself. l3 ,l4 In order to avoid this
kind of misinterpretation it must be emphasized that in the
present research we were investigating from precisely the
opposite point of view: namely, while concerned about
molecular dispersions in the resin, the response of the joint
as a whole was of major concern. Clearly, this is an
important question from a design criterion, one seldom
addressed in the past.

EXPERIMENTAL

Since adhesive joints are most commonly designed, tested

and used in lap shear mode, we chose a test specimen which in
some cases was formed from two degreased stainless-steel
plates adhering at the overlap. The thickness of adhesive
layer (h) was usually 0.02 cm, but could be in the range of
0.01 to 0.06 cm. In the initial phase of the study the bond
thickness was not controlled, but simply allowed to vary about
the 0.02 cm value. Details of the bonded assembly appear in
Figure 1. Tensile forces were applied to the non-lapped ends
of the joint by a Rheovibron.

In the later work, only on the triblock styrene/isoprene/

styrene adhesive described below, a second design for the
joint which is shown in Figure 2 was developed. Shims were
used to control bond thickness at exactly 0.3, 0.5 or 0.7 mm.
Each end of the joint then was fitted into new clamps on the
Rheovibron having slots offset as indicated in the figure.
The entire assembly was rigidly screwed together. Titanium
6,4 alloy (phosphate fluoride etch) was used for the bonded
substrates in this case.

The Rheovibron viscoelastometer, introduced by

Takayanagi,15 has found broad application for dynamic tensile
testing of polymeric films and fibers. Recently it has been
used by Murayama 16 ,17 for measurement of the dynamic shear and
dynamic compression mechanical properties of materials. For
this purpose special shear and compression grips were
developed. Two other reports of use of the Rheovibron in a
mode where a viscoelastic material was sheared between
parallel plates have been cited; however, in both cases the
temperatures were such that only rubbery and liquid-like
materials were examined. 18 ,19 In this present paper, the
tensile motion of the Rheovibron instrument clamps was
transformed into strains in the adhesive that closely
approximate those of simple shear, with a calculated magnitude
of approximately 10- 3 . A frequency of 110 Hz and a heating
rate of l°/min were selected for all testing except on the
352 T.e. WARD ET AL.

I .J
~
a= 3-5 mm

adhesive
aj E1'lll.l2222lZZZ?ZZZ!~~ ~i
b=10-30 mm

I. 50
~I h=0.1-0.6mm

Figure 1. Lap shear bond assembly for mounting in Rheovibron.

All dimensions in mm's.

!;;;;~:::I
IL-_ _rl_rL_ltr._l3....1 --i

Figure 2 Modified lap shear assembly with controlled bond

thickness for mounting in Rheovibron. Dimensions are similar
to those in Figure 1. Middle and bottom views are of new
clamps for Rheovibron.
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 353

triblock copolymers where the rate was 1/2°/min. Inertial

effects and bending moments in the joint were not observed
during any tests. In the materials examined in this study the
Massa correction 20 was not applied.

After the adhesive material was subjected to sinusoidal

shear strains, calculations of modulus proceeded according to
theory.2l Taking into account the construction of the
Rheovibron the dynamic complex shear modulus, G* in dynes/cm 2 ,
would be 22

G* =
ADab

where A is the amplitude factor, D is the corrected value of

the dynamic force, h is the thickness of adhesive layer, a is
the width, and b the length of the adhesive layer. After
recording the tan 0 value, the storage and loss moduli were
calculated in the usual fashion. In some cases, however, due
to a discrepancy between the sample geometry and the
instrument's working range, the calculated values of G* were
observed to be too low. But, decreasing the values of "a" and
"b" led to poor result reproducibility, while the magnitude
of h was limited by the co-axiality of the clamps and by the
actual thickness of the adhesive line for practical
comparisons. We found that good qualitative results were
obtained by plotting the ratio of the observed value of the
shear modulus to its initial value at room temperature, G/Gin'
as a function of increasing temperature.

Two crosslinking adhesive systems were investigated.

First a Shell Epon 828 resin was cured with a stoichiometric
amount of bis(p-aminocyclohexyl) methane (PACM-20, DuPont) for
2.5 hr. at 150°C.' Also a recently synthesized crosslinking
allyl 2-cyanoacrylate was examined. 23 Use of this material as
opposed to the alkyl ~ster is known to produce improved
thermal resistance, created by thermally induced crosslinking
of the anionically cured 2-cyanoacrylate monomer. 24

Sev(;ral thermoplastic adhesives were melted between the

apppropriate substrates in a compression mold and cooled to
form the adhesive joint. These were: (1) poly(ethyl 2-cyano-
acrylate), (2) styrene/isoprene/styrene (s/r/s) triblock
copolymers of varying composition kindly provided by the
Phillips Petroleum Co., and (3) a novel polysulfone/polyester
block copolymer having the following structure 25 , with a 50/50
(wt.) % composition and an overall molecular weight of 25,000
g/mol. The joint preparation conditions corresponding to the
numbers above were (1) room temperature, slight pressure;
354 T.C. WARD ET AL.

(2) various pressure and temperature cycles which are

discussed below, (3) 220°C, 2000 psi, slow cooling.

The S/l/S samples were selected to have narrow molecular

weight distributions, MW/Mn = 65000/63000, and MW/Mn =
86000/83000 containing 50% and 40% by weight of styrene,
respectively. These samples were found to be highly
microphase separated in structure and to have a lamellae type
morphology.

For polymer tested in the bulk (free) form in the usual

Rheovibron tensile geometry, typical sample dimensions of 0.2
to 0.6 mm thickness, 3-5 cm length and 0.5 cm width were
chosen. Thus, the polymer part of the composite joint was
comparable in size to the free film. Calculations of tan 0
and storage modulus E' proceeded in the usual way for free
films.

RESULTS

Figures 3 and 4 show the results obtained for the common

epoxy resin tested as a film in a tensile geometry and then as
an adhesive bond between metal substrates in the shear mode.
The tan 0 maximum appeared at 187°C in the bulk neat material
and at 5°C higher than this in the bonded adhesive system.
However, there was a substantial broadening of the latter
peak, reminiscent of the changes produced by adding mineral
fillers to thermosets. The maximum damping in the bonded
joint is lower by almost a factor of ten. Less dramatic was
the larger drop in the free film's storage modulus before a
plateau appeared in the relaxation.
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 355

9.5
LO

--
OlE 9.0
u
z
<I:
I-
"-
'" 8.5
~
>.
"C 0.8
--
W
~
<.!)
0 0.4
..I

0.0

80 160 240
TEMPERATURE lo e I

Figure 3. Dynamic mechanical storage modulus and damping for

epoxy free film (at 110 Hz).

Comparisons of the polysulfone/polyester block copolymer

mechanical spectra appear in Figs. Sand 6, with the free film
results in the former. The single glass transition indicated
at . above 200°C was intermediate between those recorded for
either homopolymer also tested at 110 Hz and heated at IO/min.
As observed in the epoxy systems, the maximum in tan 0
appeared SoC higher when the lap shear geometry was employed
for testing. Furthermore, the difference in the magnitude of
the mechanical damping in the two samples was even larger than
for the epoxy. Again, the temperature range of the relaxation
was broadened as detected in the joint, while the storage
modulus drop was approximately equivalent to that for the
unconstrained film.
356 T.e. WARD ET AL.

0 .0 0
W
- 0 .2 z
<
~

-0.4
0 .08

0. 06
<.?Ji
~
~
0
....I 0 .04

80 160 240
TEMPERATURE / °c /
Figure 4. Dynamic mechanical relative shear modulus and
damping for epoxy bonded joint assembly (at 110Hz).

Dynamic mechanical testing of poly(ethyl 2-cyanoacrylate)

produced results shown in Figures 7 and 8 in the unbonded and
bonded configurations, respectively. The sharp relaxation
with a maximum at about 110·C for the free film was modified
in the bonded joint to become a quite broad temperature re-
sponse at this test frequency, with a more complex shape. The
magnitude of the higher temperature damping maximum in the
joint is approaching the absolute values described above for
the free films. In addition, the first peak found on heating
is about 20·C higher than in the unrestricted film. In
previous work we have noted that the poly(ethyl 2-cyano-
acrylate) was found to have very poor thermal stability,
showing a total loss of tensile lap shear strength on aging
for 24 hours at lSO°C.24
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 357

10
.....
(II
E
....u
fII
<II
c to
'"'
"0 Z
....
W
'"
I-

S!
0
0
...J
9

80 140 200 260

TEMPERATURE r eI
Figure 5. Dynamic mechanical storage modulus and damping for
polyester/polysulfone block copolymer film (at 110Hz)

0.0

- 0,4
~
-0.8
z
-1.2 '"
I-

'o1J
2
0
0
...J

80 140 200 260

T EMPERAT URE r ei
Figure 6. Dynamic mechanical relative shear modulus and
damping for polyester/polysulfone block copolymer adhesively
bonded joint (at 110 Hz).
358 T.e. WARD ET AL.

"'E 9.8 0.6

u
.....
CIo
;. 9.0 0.1,.,..
"U
z
'w ~
R2 0.2
<.:>
o
....J

30 60
T, 'c

Figure 7. Dynamic mechanical storage modulus and damping for

poly(ethyl 2-cyanoacrylate) film (at 110 Hz).

-c
b~e,;- -0.4
~

2
"o
..J
-0.8

eo ~eo

TEMPERATURE / ·C I

Figure 8. Dynamic mechanical relative shear modulus and

damping of poly(ethyl 2-cyanoacrylate) bonded adhesive joint
(at 110 Hz).
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 359

In contrast to the alkyl 2-cyanoacrylate adhesive, the

allyl version had quite different mechanical response as is
presented in Figure 9. The maximum in tan 6 for the bonded
system is about SoC higher in temperature than that seen for
the alkyl version, but then decreases gradually with continued
heating. More striking is the recovery of the inital shear
storage modulus to its original value on heating. Previously,
we determined that a thermally induced crosslinking was
responsible for this behavior. 24 As the vitrification due to
network formation proceeded, the glassy storage modulus was
approached, in spite of the increase in temperature at 1°C/min
(which apparently was slower than the progress of the chemical
reaction). The fully-cured allyl ester adhesive was found to
have a Tg greater than 200°C.24

0.0

-0.2 Lo
z
-0.4 <
~

-0.6
0.2
~c

c, I(i'-
Q
C)
0
..J
0.1

80 120 160 200

TEMPERATURE r ei
Figure 9. Dynamic mechanical relative shear modulus and
damping of a poly(allyl 2-cyanoacrylate) bonded adhesive joint
(at 110 Hz).
360 T.e. WARD ET AL.

The final experimental results were obtained on the

triblock (styrene/isoprene/styrene) thermoplastics, and
represent a second phase of the research in which bond
thickness was tightly controlled rather than varying around a
nominal value of 0.2 mm. The design of Figure 2 was employed
for all of this work.

Preparation of the adhesively bonded joints with the

S/l/S polymers was extremely important because of the
sensitivity of the polymer morphology to thermal history. For
example, we were able to extrude films of similar triblock
materials which showed different free film storage moduli
(almost a factor of two) in the extruded and transverse
directions. 25 Also, using Dse we sometimes noted unexpected
thermal response of these materials at about 20 0 e below the
glass transition temperature which was attributed to a "stress
release phenomenon." It was demonstrated that differences in
the compression molding operation could be associated with
either the presence or the absence of such stress release
events. 25 Thus, the S/l/S materials served as model adhesives
with which to evaluate the trends in the results for the other
polymers described above.

Results of experiments on S/l/S free films and on S/l/S

bonded joints with the thermoplastic glue line at 0.3, 0.5 and
0.7 mm appear together in Figure 10. All preparations were
made with the 50% styrene content polymer. The preparation
temperature was l50 0 e and at least 5000 psi was applied in the
molding. Pressure was maintained during cooling. The Tan o's
are enlarged in Figure 10 compared to the usual presentation
in order to emphasize sample differences. An increase in the
rate of the molecular relaxations seems to appear in the bulk
free film (B) on heating, with a very rapid increase in
damping in the vicinity of lOOoe. For this unbonded film the
Tan 0 passed out of the range of the instrument (for these
sample dimensions) at about 108°e. In contrast, for all three
bond thicknesses investigated the damping of the joint
assembly was lower than that of the free film at equivalent
temperatures, while the onset of very high damping shifted
upward, approximately 20, degrees depending on bond thickness.
It is apparent that in the very thin bond-line joints there
are indications of several temperat~re ranges where small
maxima or details in the damping appear. These results were
independent of heating 'rate and reproducible as long as the
same joint preparation conditions were observed. A major
question of concern was whether the small peaks were real or
artifacts.
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 361

J
0.3

0.2

<0
z
<!
f-
/1
0.1 .-o---~

70 90 T .C 110 130

Figure 10. Dynamic mechanical damping of 50% styrene S/l/S

bonded joint and free film at 110 Hz. Bond thicknesses shown
on curves are in cm for joints. Free film = B.

When the experiments represented by Figure 10 were

repeated with the 40 wt. % styrene S/l/S block copolymer,
similar observations were again recorded. These results
appear in Figure 11. Preparation conditions were the same as
in Figure 10. As expected, the magnitude of the damping was
higher overall due to the increased isoprene fraction in the
copolymer. However, the major features of the previous curves
were retained. These are (1) higher temperatures required for
appearance of damping maxima in joints as opposed to free
films (2) unusual behavior of the Tan 0 prior to very high
damping phenomena, which seems to depend on bond thickness in
the joints as opposed to the free film (B).

Further experiments were conducted on the 50%

styrene content S/l/S copolymer. Modifications of the molding
procedure for the S/l/S adhesives were made in an attempt to
362 T.e. WARD ET AL.

0.8

0.6

'-';)Q4
$

90 T,C 00

Figure 11. Dynamic mechanical damping of 40% styrene S/I/S

bonded joint and free film at 110 Hz. Bond thicknesses shown
on curves are in cm for joints. Free film = B.

understand the behavior of the fabricated joints as a whole.

The major changes were to provide for slow cooling of the free
films with and without any application of external pressure
and to vary sample thickness. As a consequence of this new
manufacturing process, the mechanical behavior was also
altered. It was reasoned that if pressure-induced frozen
stress changed mechanical response in bonded composites, a
very thin free film prepared under conditions identical to
those in the joints might also exhibit similar response.
Confirmation of this hypothesis appears in Figure 12, where
data on two free films are shown. The curve labeled P,O.1
represents results on a 0.1 rom thick, cooled-with-pressure
sample. In the other curve (B), a 0.6 mm thick,
cooled-with-pressure, material response is indicated. These
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 363

7 T,C 90

Figure 12. Dynamic mechanical damping of 50% styrene S/l/S

free fi lms, cooled with> 5000 psi pressure. B = 0.6 mm,
P,O.l = 0.1 rrm thick.

latter data are replicates of those 1n Figure 10. Influence

of the sample thickness is striking in this figure.

A final set of experiments on the 50% styrene S/l/S

materials is summarized in Figure 13. Both curves shown
depict the outcome of tests on free films. Now, holding the
sample dimensions in each run almost the same with a thickness
of 0.1 mm, preparation of the specimens differed only in that
cooling was accomplished after molding with either> 5000 psi
pressure applied (p) or with no pressure (NP). The difference
in Tan 0 was dramatic again and parallels that found when a
thick specimen (B) was used with pressurized cooling, Figure
12, instead of the thin one. Neither of the comparisons made
in Figures 12 or 13 causes a shift of the tan 0 curves to
higher temperatures as was found when bonded joints were
investigated.
364 T.C. WARD ET AL.

0 .4

0.2

I'

70 T,C 90 11)

Figure 13. Dynamic mechanical damping of 50% styrene S/l/S

free films of 0.1 rom thickness. P = cooled with> 5000 psi
pressure, NP = cooled without pressure.

DISCUSSION AND CONCLUSIONS

In both the thermosetting and thermoplastic adhesives

there were always different Tan 0 (at 110 Hz) functiqns of
temperature when free and bonded films were compared at equal
heating rates. Generally, the viscoelastic response of the
composite joint occurred over a wider temperature range than
that of the free film, again heating rates being equal.
Substantially different magnitudes of the monitored relaxation
events were observed in each case. Non-equil ibrium effects in
the testing procedure do not seem to be responsible for any of
these observations, non-equilibrium states in the joints may
be.

At this time no firm conclusions may be put forth

concerning the origins of the variations in dynamic mechanical
properties with respect to free film vs. bonded lap shear
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 365

joints; however, some facts are clear. In the thermoset

materials and the polyester-polysulfone single phase block
copolymer the differences in performance of bonded and free
films are not as large and may reflect normal shifting due to
the change in the monitored viscoelastic function in each case
(tensile elongation vs. shear). Since simple shear is a
constant volume process, and all other types of deformations
require change in volume (Poisson's ratio * 0.5), a correct
constitutive equation description of the polymer's response
would predict different relaxation behavior at constant
temperature, or different temperature profiles in isochronous
experiments. In the investigations of the present paper this
effect should be quite small, and difficult to calculate.

A second view of the difference in mechanical behavior of

the free and bonded films would attribute the physical
variations to frozen stresses (originating from shrinkage,
non-equilibrium morphology and crosslinking). Certainly the
results on the S/I/S block copolymers suggest that thermal
history must not be ignored in evaluating damping in a
polymerically bonded structure placed in oscillatory shear
fields. As the bond thickness was decreased the magnitude of
the changes in viscoelastic response usually increased,
depending on the polymer. Further experimentation at other
frequencies and under certain isothermal conditions where
stress release may be measured would be informative in
clarifying the mechanism in this case.

With regard to long range polymer modulation (e.g.

limitation of conformation isomers, damping of normal modes)
by a highly rigid surface, the present experiments lead us to
make no statement. The substrate surface area to volume of
polymer ratio is so small that one finds it difficult to
believe that any variation of dynamic mechanical relaxations
of the overall bonded structure would manifest themselves in
such simple experiments. Even in much more sensitive
experiments on thin films, no mention is made of
instrumentally induced limitation of molecular motion. 27

Finally, the series of investigations on the S/I/S block

copolymers reveals some interesting conclusions vis-a-vis
dynamic mechanical testing of compression molded heterophase
polymers. It appears that either free films or bonded joints
may be manufactured which will show a fine structure in the
Tan 0 vs. temperature experiment, prior to the actual maximum
damping-of the film or assembly. Cooling thin films with
applied pressures greater than 5000 psi was required to
produce these details. Using thick films, or thick bond
lines, or cooling without pressure would completely remove
366 T.C. WARD ET AL.

these smaller damping peaks. Confirming our DSC work, there

was always a "bump" in these curves in the region of 70"'C.
The presence of the substrates, furthermore, when a thin bond
line was used, shifted the large Tan 0 damping to
substantially higher temperatures, free films never showed
this behavior.

Because of their nice model adhesive character, the S/l/S

copolymers' behavior helped clarify the response observed in
the other adhesives of this study. The generally broadened
peaks of mechanical energy damping, occurring at higher
temperatures, for these other samples may be manifestations of
similar unrelaxed stress fields in the bonded joints. The
caveat which was suggested on completion of this work
reinforces the importance of sample preparation technique in
systems where glass formation occurs during joint manufacture
(such as is the case for structural adhesives).

ACKNOWLEDGEMENTS

The support of the Office of Naval Research and the Army

Research Office is gratefully recognized. Also the synthesis
of polymers by Dr. James E. McGrath and his research group
significantly aided this project.

REFERENCES

1. W. B. Jones, Jr., Organic Coatings and Applied Polymer

Science Proceedings, 47, 247.(1982).
2. D. W. Dwight, E. Sancaktar and H. F. Brinson, in
"Adhesion and Adsorption of Polymers", L. H. Lee, Editor,
Part A, 141, Plenum Press, New York, 1980.
3. G. Dolev and O. Ishai, J. Adhesion, 12, 283 (1981).
4. L.R.F. Rose, J. Adhesion, 14, 93 (1982).
5. R. W. Hylands and E. H. Sidwell, J. Adhesion, 1!, 203
(1980).
6. W. D. Bascom, R. L. Cottington and C. O. Timmons, J.
Appl. Polym. Sci., Appl. Polym. Symp., 32, 165 (1977).
7. D. L. Hunston, S. S. Wang, A. J. Kinloch; Organic
Coatings and Applied Polymer Science Proceedings, 47,
408 (1982); also see J. T. Bitner, J. L. Rushford,1W. S.
Rose, D. L. Hunston and C. K. Riew, J. Adhesion, li,
3 (1981).
8. V. F. Babich and Y. S. Lipatov, J. Appl. Polym. Sci., ~,
53 (1982).
9. Y. S. Lipatov, V. F. Rosovizky and V. V. Shifrin, J.
Appl. Polym. Sci., ~, 455 (1982).
NONDESTRUCTIVE EVALUATION OF SOME BONDED JOINTS 367

10. G. J. Howard and R. A. Shanks, J. Macromol. Sci.-Phys.,

B19(2), 167 (1981).
11. P. Peyser, Polym - Plast. Technol. Engr., 10(2), 117
(1978).
12. P. S. Theocaris and G. D. Spathis, J. Appl. Polym. Sci.,
27, 3019 (1982).
13. ~ J. Plazek, J. Polym. Sci., 20, 1533 (1982).
14. D. J. Plazek and G-F. Gu, J. Polym. Sci., 20, 1551
(1982). --
15. M. Takayangi, in "Proceedings of the Fourth International
Congress on Rheology," C. Klason and J. Kubat, Editors,
Part I, p. 161, Swedish Soc. Rheo. Publisher, Gothenburg,
1965.
16. T. Murayama, J. Appl. Polym. Sci., 19, 3221 (1975).
17. T. Murayama, in "Proceedings of the Seventh International
Congress on Rheology," E. H. Lee and A. L. Copley,
Editors, p. 402, Interscience, New York, 1977.
18. B. H. Shah and R. Darby, Polym. Eng.. ScL, 16, 46
(1976).
19. P. F. Erhardt, J. J. O'Malley and R. G. Crystal, ACS
Polym. Preprint, 10(2), 812 (1969).
20. D. J. Massa, J. Appl. Phys., 44, 2595 (1973); also A. R.
Ramos, F. S. Bates, R. E. Cohen, J. Polym. Sci., Polym.
Phys. Ed., 16, 753 (1978).
21. S. Timoshenko, "Strength of Materials," Part I, Van
Nostrand, Princeton, 1955.
22. Instruction Manual No. 68, Rheovibron Model DDV-II-C,
Toyo Bladwin Co., Ltd. Tokyo (1973).
23. D. L. Kotzev, C. Konstantinov, P. Novabov, V. Kabaivanov,
Bulg. Pat. No. 23321 (1977).
24. D. L. Kotzev, T. C. Ward, D. W. Dwight, J. Appl. Polym.
ScL, 26, 1941 (1981).
25. J. E. McGrath, Private Communication.
26. A. Wood, M. S. Thesis, Virginia Polytechnic Institute and
State University, Blacksburg, Va. (1982).
27. J. D. Ferry, "Viscoelas tic Properties·· of Polymers," 3rd
Ed. Chapter 5, Wiley, New York, 1980.
ULTRASONIC ASSESSMENT OF CURE RATE EFFECTS IN BONDED HONEYCOMB

STRUCTURES

R. A. Pike and R. S. Williams

United Technologies Research Center

Silver Lane
East Hartford, Connecticut 06108

The effect of variations in cure temperature on

fillet formation in adhesively bonded honeycombstruc-
tures has been defined in terms of fillet length, de-
gree of cure and Tg of the cured neat adhesive. The
results are correlated with predicted behavior based on
ultrasonic NDE analysis prior to testing. The ultra-
sonic test method uses advanced pattern recognition
,analysis of attenuation and velocity data extracted
from a multi-frequency Lamb wave propagated along the
face sheet of the honeycomb. Laboratory applications
of this test method are described.

369
370 R.A. PIKE AND R.S. WILLIAMS

INTRODUCTION

The increasing awareness that adhesive performance is controll-

ed not only by the condition of the adherend surface but also the
condition or "state" of the adhesive and the process parameters
used during fabrication is expected to result in improved reliabil-
ity as well as bond performance. Improved reliability will also
result from the implementation of a well defined NDE analysis
directly relatable to the intrinsic properties of the cured ad-
hesive. The increasing application of honeycomb adhesively bonded
systems as structural components wherein the adhesively bonded
surface is considerably less than in a normal metal-to-metal
system underscores the necessity of obtaining a complete defini-
tion of the factors which affect bond performance and quality.

Recognition of the fact that assurance of bond quality and

integrity involves many facets, a multidisciplinary approach that
combines NDE, adhesive technology and fracture mechanics has been
employed at United Technologies Research Center to form the basis
for a comprehensive understanding of adhesive performance. A
description of the NDE method employed to define the bond strength
of metal-to-metal bonds involving the ultrasonic measurements and
mathematical method, RESID (Recursive Structural Identification)
has been reported. l An adaptation of this NDE method for bonded
honeycomb structures is described in the following sections of
this paper.

Heat-up or cure rate during fabrication has been identified

as one of the major factors which control bond performance. This
effect has been previously alluded to by a number of workers in
terms of stress levels,2 rheological cure behavior 3 and tensile
lap shear strength. 4 In-house investigations have involved the
correlation of adhesive chemistry, strength and damping character-
istics with heat-up rate. S

The objective of the study described here was to correlate

ultrasonic wave response, NDE, with the pertinent physical and
mechanical properties of adhesively bonded honeycomb structures
using cure temperature as the primary fabrication variable. The
effect of heat-up·rate on bonded honeycomb performance and NDE
response is also being investigated and will be reported when
complete.
ULTRASONIC ASSESSMENT OF CURE RATE EFFECTS 371

EXPERIMENTAL

The reticulatable (unsupported) and scrim supported EA-9649

adhesive film used in the study was kindly supplied by the Hysol
Division of Dexter Corporation, Pittsburg, California. The adhesive,
designated as a 178°C (350°F) curing system, has been previously
described. 6

Honeycomb specimens were fabricated from 5.0mm thick, 10.16

cm 2 solid and perforated 2024 aluminum plates using 8mm hexagonal
aluminum honeycomb. All surfaces were acetone rinsed, vapor
degreased, etched (FPL) and primed prior to bonding. Supported
film bonded the solid face sheet; unsupported film, the perforated
plate.

Thermomechanical analysis (TMA) was carried out using a DuPont

1090 Thermal analyzer at a heating rate of 10°C/min.

Dielectric analysis of neat adhesive was performed as pre-

viously described 6 using a Tetrahedron, Inc. Audrey Model 203
dissipation factor bridge with a Model 1501 automatic press. Tem-
perature was automatically recorded by using a thermocouple
inserted directly into the two-ply adhesive lay-up. A heat-up
rate of 2.5°C/min was used with the 7.6cm 2 sample.

Bonding was carried out at 50psi pressure at four different

cure temperatures, namely 140°, 160°, 180° and 200°C using a 2.5°C/
min heat-up rate. The time to reach the specified temperature thus
varied from 46, 54, 62 and 70 minutes, respectively.

Adhesive flow or fillet length was determined by averaging ten

fillet lengths measured in millimeters under lOX power magnification.

The ultrasonic plate wave technique developed for honeycomb

structures is shown in Figure 1. A longitudinal wave from the trans-
mitter is propagated through the Lucite into the skin of the honey-
comb at the Rayleigh critical angle. By the process of mode
conversion, a surface or plate wave is generated that propagates
along the face sheet of the honeycomb. At the other end of the
fixture, this process is reversed and the wave detected by the
receiver. Since the wave travels over a known distance, the
attenuation per unit distance and the velocity is easily measured.
372 R.A. PIKE AND R.S. WILLIAMS

TRA SMITIER SOUND ABSORBTIVE MATERIAL

RECEIVER

a) CROSS SECTION

FACE SHEET""\ £. SURFA~~ ~AVE O~ ~~TE WAVE

"
1--'", '....
; c
' .... _,'

FlLLIl'-A::7'VE lr
,-- ,-- ,-, ; --~

71
, ""... ' .. _
,,'~ _ d
" _
,

b) ENLARGED DETAIL SHOWING INTERACTION

WITH THE ADHESIVES

Figure 1. Plate wave transducer.

The level of interaction with the sheet adhesive, the fillet, and/
or the honeycomb is determined by the wavelength (or frequency).
At frequencies greater than 15-20 MHz, a surface wave is generated
that will be affected little by the bond. At lower frequencies
(2-5 MHz), a plate wave will be generated that will interact
strongly with the adhesive adhering to the lower surface of the
face sheet. At still lower frequencies (500 kHz-2 MHz), the plate
wave will interact with the fillet and the honeycomb. A RESID
algorithm,as discussed below, is used to correlate the ultra-
sonic data obtained with the actual measured properties of the
adhesive or bonded structure.

RESULTS AND DISCUSSION

Three methods were used to assess the effect of varying cure

temperature on both supported and unsupported EA-9649 adhesive film
in terms of flow, gel time and final Tg. These were dielectric
and TMA analysis of the neat adhesive and fillet length measure-
ULTRASONIC ASSESSMENT OF CURE RATE EFFECTS 373

ments of the reticulated unsupported adhesive film in a bonded

honeycomb structure.

The dissipation factor profile curves obtained on the two

types of adhesive film are illustrated in Figures 2 and 3. The
pertinent data are listed in Table I.

Table I. Cure Temperature Effects on EA-9649 Adhesive-Dielectric

Analysis a

Cure Temp, ISAb Time to gel, Tg,

°c (%flow) minutes °Ce:

Supported Adhesive 200 97 72 246

180 70 81.5 220
160 89 86 203
140 71 101 139

Unsupported Adhesive 200 45 70 239

180 25 81 205
160 35 90 185
145 42 102.5 142

a. Analysis run at 1000 Hz, 50psi pressure, 2 ply lay-up.

b. ISA = Increase in surface area.
c. Measured by TMA on the cured sample.

For the unsupported adhesive an increase in Tg point and decrease

in time to gel with increasing cure temperature was observed as
expected. The flow, as measured by the increase in surface area
during cure, was found to be minimum at the specificatio~ 180°C
level. Higher flow was experienced above and below this tempera-
ture. As previously shown 6 ,7 changes in the major melting point
and gel times were influenced by the changes in the condition of
the adhesive as well as fabrication parameters. The supported
adhesive in general gave a similar cure response with varying cure
temperature. Minimum flow was again found to occur at 180°C. The
low flow associated with the 140°C cure i~ probably due to the
effect of the support scrim coupled with the higher viscosity of
the adhesive. Accompanying these changes in resin behavior was
the higher final dissipation factor obtained by samples cured at
temperatures other than the specification 180oC. This, as has
been previously discussed for aging 6 and moisture 7 effects on
EA-9649 adhesive, is indicative of a different final molecular
374 R.A. PIKE AND R.S. WILLIAMS

a:
o
~
li-
Z
o 010
i=
<t:
0..
U5
(/)
180
o 150
0 L-__L -__Ll~4~0~__- L__~__~__~__~__~__~__~
o 10 20 30 40 50 50 70 80 90 100 110
TIME . MINUTES

Figure 2. Dielectric analysis - effect of cure temperature,

Supported EA-9649 film adhesive.

structure in either a fully or partially cured state. DSC

analysis of the cured adhesive showed the presence of unreacted
catalyst at the 140°C and 160°C temperatures. This was
apparent from the slight endotherm (melting) which occurred at
approximately 220°C. The melting point of dicyandiamide is
2l6°C at the heat-up rate used for the analysis. Thus, at

020
UNSUPPORTED EA9549 FILM ADHESIVE
a:
0
I-
U
<t:
li-
- .......
Z
0 010
i=
<t:
0..
U5
(/)

0
0 10 40 50 60 70 80 90 100 110
TIME . MINUTES

Figure 3. Dielectric analysis - effect of cure temperature,

Unsupported EA-9649 film adhesive.
ULTRASONIC ASSESSMENT OF CURE RATE EFFECTS 375

temperatures below 1800C the adhesive appears to be only par-

tially cured. The marked effect of cure temperature on flow
related to resin viscosity, cure kinetics and temperature rise
indicates the necessity of careful monitoring of process para-
meters to achieve reproducibility in bonded structures.

In order to correlate the flow response obtained by dielectric

analysis with fillet formation in a bonded honeycomb structure,
measurements of fillet length were made on the reticulated unsupported
adhesive using specimens fabricated at the four curing temperatures
with primed and unprimed honeycomb. The results are shown graph-
ically in Figure 4. The degree of flow obtained on the primed
honeycomb with perforated plate on top was in exact agreement with
the dielectric analysis results, i.e., a minimum flow was obtained
at the 180°C cure temperature with longer fillets being formed at
cure temperatures above and below the specified cure temperature.
The flow on the unprimed honeycomb showed a slight minimum at 160°C
indicative of the different response of the spreading characteris-
tics of the adhesive to a metal or resin surface. Reversing the
honeycomb construction so that the perforated plate was on the
bottom (adhesive flows up) gave a minimum flow condition at 180°C
on the unprimed surface. Thus, varying cure temperatures produce
trends in adhesive flow and final cured state properties which
should affect not only the mechanical properties of a bonded joint
but the response of the adhesive to ultrasonic wave stimulation.

The RESID algorithm approximates a complex nonlinear relation-

ship with a network of simple binary quadratic functions. The
interconnections of these elements proceed in a selective and
intelligent manner so that,

(1) Dissimilar features are optimally combined.

(2) Higher order feature combinations are created as they
are needed.
(3) Inappropriate combinations of features are automatically
excluded.
(4) Overfitting the data set is minimized.

The RESID, Recursive Structure Identification, algorithm is

based on the polynomial theory of complex systems introduced nearly
a decade ago by the Russian academician, A.G. Ivakhnenko 8 The
theory has since been implemented by a number of investigators.
376 R.A. PIKE AND R.S. WILLIAMS

4.0,.-------------------------,
UNSUPPORTED EA9649 ADHESIVE

PERFORATED HONEYCOMB
3.5
o BOnOM UNPRIMED
• TOP UNPRIMED
(J)
II: l:; TOP PRIMED
UJ
I-
UJ 3.0
:2
:::i
-l
:2
:i
0
2.5
-l
u...

2.0

1.5~~---~----L-----~---~----~----~
140 160 180 200
PROCESS TEMPERATURE, DC

Figure 4. Adhesive fillet length(in terms of flow) vs processing

temperature.

The motivation for RESID is threefold. First, the relation-

ship between features and the quantity to be determined, or item
to be classified, is unknown and often nonlinear. Second, the
feature set may be quite large and may contain different or even
qualitative types of features. Third, it is desirable to train
the classifier using as small a data set as possible.

In general, the nonlinear relationships between the material

properties and the ultrasonic features can be represented by a
multinomial power series which in this case is impractical to
use.

The approach used for this investigation was to approximate

a complex function with an interconnection of simple functions.
A suitable elemental function is the binary quadratic function,

which contains six coefficients and is shown schematically in

Figure 5. With this strategy, it is possible to achieve fourth
power relationships with as few as eighteen coefficients.
ULTRASONIC ASSESSMENT OF CURE RATE EFFECTS 377

r-----------------~
I
I bO
I
w1~----~------~--~

w2~~--_r--~--~--~
I E X
I
I
I

! ~
L __________________ J

Figure 5. Network for predicting material property.

There are several approaches to obtaining the element coef-

ficients and interconnections. One strategy is to fix the network
structure, and depend on many of the coefficients becoming small
enough to ignore. The coefficients can ,be computed using either
a global random search or a deterministic least squares fit. A
second strategy is to grow the network layer by layer until
adequate mixing of the variables has occurred. RESID uses the
second method.

For the first layer, RESID examines all pairwise combinations

of features. Those elements having minimum square error fit to
the output variable are retained. The six coefficients in each
element are computed using deterministic matrix operations.
Subsequent layers are added by examining all pairwise combina-
tions of previous layer outputs. Those elements having errors
less than either of the errors of its parents are retained.
When this condition is not met, the layer generation terminates.

A key step in the RESID algorithm is the partitioning of

the data set into three mutually exclusive subsets: training
subset, selection subset, and evaluation subset. The training
subset is used to determine the coefficients in each of the net-
work elements. The selection subset is used to eliminate those
elements that have low correlation with the known output. The
evaluation subset is used to give an independent assessment of
overall network performance. The training and selection subsets
are applied alternately as each level of the network is grown.
The network stops growing when the algorithm starts to fit the
noise type variations in the data.
378 R.A. PIKE AND R.S. WILLIAMS

Single feature ultrasonic data (in this case plate wave attenu-
ation at 1 MHz) versus the fabrication temperatures and measured Tg
of the adhesive are shown in Figure 6. Although there is a slight
general correlation, i.e., increasing Tg with increasing cure temper-
ature, the data exhibit considerable scatter. This is expected
since no single ultrasonic feature is uniquely a function of the
adhesive intrinsic properties such as Tg.1 Additional ultrasonic
features must be used to provide multi-dimensional data for process-
ing by the RESID algorithm. Some of these data, ultrasonic plate
wave attenuation at several frequencies and power levels are shown
in Figure 7. It should be noted that maximum attenuation for two of
these features occurs at the 180°C cure temperature which corresponds
to the minimum fillet length results.

Application of pattern recognition methods (RESID) to the

multiple ultrasonic data greatly improves the correlation. Using
the RESID network, as described above, the scatter can be reduced
such that the maximum prediction error is 10 percent as shown in
Figure 8. The horizontal axis is the predicted cure temperature
and resulting Tg using the RESID network with five ultrasonic
attenuation and velocity features, the vertical axis is the actual
cure temperature and measured Tg. It should be noted that the
graph contains both training as well as evaluation specimens.
The former were used to determine the parameters required to
finalize the RESID network.

200 r-----~-----,-----.------._----.-----_.

150r-----+------r----~------+-----~----~

llJ
a:
~
....
~100r-----+------r----~------+-----~----~
u.
llJ
Cl
Z

50r-----+------r----~------t-----~----~

0
140 160 180 200
FABRICATION CURE TEMPERATURE
I I I

150 190 220 245

TG, ·c
Figure 6. Honeycomb cure property vs NDE feature.
ULTRASONIC ASSESSMENT OF CURE RATE EFFECTS 379

160 o 2.5 MHz LOW POWER

o 1 MHz HI POWER
(:; 1 MHz HI POWER
SENDI3 MHz
RECEIVE
x 1 MHz LOW POWER
z 120 SEND/3 MHz
0 RECEIVE
;:::
«
=>
z
w
t:«
w 80
>
;:::
«
....J
w
a:
0

40

0
100 140 180 220
CURE TEMP . ·C
I I I I
150 190 220 245
TG . ·C

Figure 7. Cure temperature vs attenuation.

This approach has also been successfully used to predict a

number of other metal-to-metal adhesive bond properties, including
bondline thickness, porosity, glass transition temperature, degree
of cure, and bond strength. l For each property, a different RESID
network is derived, although the ultrasonic features for each net-
work may be selected from the same set of data. For example,
network I to predict property I may use five features from a
possible sixteen; network II to predict property II may use six
entirely different features or it could share some of the features
used for network I. In all cases a network can, by design,
accommodate variations in the adhesive system; however dramatic
changes in the adhesive/adherend system will require the use of a
different set of networks.

Correlation between ultrasonic wave response and flatwise

tensile data for the bonded honeycomb specimens is in progress.
Preliminary results are very encouraging and will be reported
when complete.
380 R.A. PIKE AND R.S. WILLIAMS

21Sr-------------------------------~
~
• =TRAINING SPECIMENS
245 u.i
cr • =EVALUATIO SPECIMENS
::J
t--
«
cr
w 185
220 0..
::!
W
~ I-

0
I-
UJ
cr
::J
190 (J
z 155
0
>=
«
(J
ISO 1i'
CO
«
'"- 125~ ____ ____ ______L __ _ _ _ _ _ _ _
~ ~ ~ ~

140 152 164 176 188 200

PREDICTED CURE TEMPERATURE . ·C

ISO 190 220 245

Figure 8. Honeycomb cure property prediction.

CONCLUSIONS

A test method using experimental ultrasonic measurements and

recursive structure identification (RESID) to analyze the data has
been developed that is potentially capable of predicting adhesive
bond properties in honeycomb bonded structures. Cure temperatures
should be strictly monitored to ensure optimum adhesive performance.

ACKNOWLEDGEMENTS

The authors wish to acknowledge the technical contributions of

Dr. P. E. Zwicke in the development of the RESID algorithms and Mrs.
J. P. Pinto for the determination of adhesive properties.

REFERENCES

1. R.S. Williams and P.E. Zwicke, Materials Evaluation 40,312 (1982).

2. C.L. Mahoney, paper (No.790l5l) presented at the Soc. of Auto-
motive Engineers, Congress & Exposition, Detroit, February 26,1979.
3. J. Thuen and R. Hinrichs, in "Proc. 25th National SAHPE Symposium
& Exposition", held May 6, 1980, p. 126.
4. E. Sancaktar, H. Jozavi and R.M. Klein, paper presented at the
5th Annual Adhesion Society Meeting, Mobile, AL, February 22,1982.
5. R.A. Pike, F.P. Lamm and J.P. Pinto, to be published.
6. R.A. Pike, F.P. Lamm and J.P. Pinto, J. Adhesion ~,143 (1981).
7. R.A. Pike, F.P. Lamm and J.P. Pinto, J. Adhesion 11,229 (1982).
8. A.G. Ivakhnenko, IEEE Trans. on Systems, Man and Cybernetics,
SMC-l, No.4, October 1971, p. 364.
THE THREE-POINT BEND TEST FOR ADHESIVE JOINTS

N. T. McDevitt and W. L. Baun

l1echanics and Surface Interactions Branch

Air Force Wright Aeronautical Laboratories
vlright-Patterson Air Force Base, Ohio, 45433

Unless a large amount of degradation has occurred in

the interfacial region of a metal-to-metal adhesive bond,
the T-peel and wedge opening test usually provide only
mode I information on the bulk adhesive. The single-lap
shear test shows primarily mixed-mode failure. However,
the thickness of the adherend must be carefully
considered in order not to have a large amount of mode I
failure. When these three tests failed consistently to
find defects that were incorporated into the interfacial
region of test specimens, the three-point bend test
geometry was evaluated. This initial study, utilizing
the short beam shear test geometry, has shown that
several pieces of information may be obtained from the
experimental data. The initial portion of the test curve
appears quite linear and the slope of the curve seems
directly related to the bulk properties of the adhesive.
The portion of the curve where a definitive failure
occurs appears directly related to interfacial failure of
the bonded joint.

381
382 N.T. McDEVITT AND W.L. BAUN

INTRODUCTION

Historically the effect of stress and environment on the

overall strength of metal-to-metal adhesive bonds has been
determined primarily by three mechanical tests; (1) peel, (2) wedge
opening, or (3) lap shear. However, attempts to correlate data
from accelerated laboratory experiments to actual service life,
using tr~~e mechanical tests, have met only with marginal
success. Actual in-service failures of an adhesive bonded
structure have been reported to occur in the region of the
adhesive-oxide interfaceS A recent study using surface
instrumentation techniques shows failure sites also occurring at
the oxide-metal interface. Therefore, this interphase region of
adhesive-oxide-metal should be of concern in laboratory studies of
adhesive bond line failure. Since the joint geometry of the peel
and wedge opening tests provide only crack opening data, primarily
toughness properties of the adhesive are obtained and very little
information is generated about the subtle changes that may occur in
the interphase region. The lap shear test is capable of generating
data from the interphase region but care must be taken in the
proper choice of thickness of the test adherend in order to
eliminate most of the peel forces.

When dealing with high strength aluminum alloys the aspect of

corrosion has to be considered as a potential problem. In most
corrosion processes a thin film is involved at some stage. The
controlling factor in the growth of this film is generally the type
of environment and length of time of exposure of the test specimen.
In order to study the subtle effects of surface preparation and
corrosion films on the strength of the adhesive-oxide and/or
oxide-metal interfaces our interest was directed to a test geometry
that would be simple to fabricate and would load the bond line with
as much pure shear force as possible.

We chose the three-point bend test for our study and the
purpose of this paper is to illustrate the general aspects of this
test method on metal-to~metal adhesive bonds. This study, to our
knowledge, is the first use of this test method on metal sandwich
structures.

EXPERIMENTAL

Aluminum Alloy Specimens

Rectangular specimens were cut from a sheet of 7075-T6 bare

aluminum alloy. Each specimen was 10 x 15 x 0.35 centimeters. All
of the specimens were degreased with an acetone wipe, then alkaline
etched with O.IN NaOH for three minutes at room temperature.
Specimens were deoxidized with a solution of 5:1 HN0 3-HF for two
THE THREE-POINT BEND TEST FOR ADHESIVE JOINTS 383

minutes at room temperature, then desmutted with 50% HN0 3 for 30

sec.

Duplicate panels were then anodized in a 1. OM H3 P0 4 bath

according to the conditions described in Table 1. Each palr was
then bonded with FM 123-2 adhesive (in tape form), cured at 250°F
and 25 psi, without the aid of a primer. Specimens 3.8 x 1.3 x 0.7
centimeters were cut from the bonded panels. Duplicate specimens
of each surface preparation were subjected to the environmental
conditions described in Table II.

Table 1. Anodization Conditions for 100M H3 P0 4

Surface Time in Average Oxide

Pre)2aration Volta~e Bath Thickness

1 10 0.5 min 225A

2 10 2.0 1000
3 10 10.0 3200
4 10 16.0 3600
5 40 2.0 1200
6 40 5.0 3000

Table II. Type of Curing and Exposure for Test Specimens.

Test
A specimens cured then tested within 24 hours
B specimens cured then post cured at 220°F for 8 hours
C specimens cured then stored in desiccator under ambient
conditions for 90 days
D specimens from Test C subjected to dry heat, 100°C, for
63 hours
E specimens from Test C subjected to S02 environment at
R.T. for 2160 hours
F specimens from Test C subjected to S02 environment at
100°C for 72 hours
G specimens from Test C stressed to 1000 lb for two
minutes then subjected to 100% R.H. and 100°C for 8
hours

Mechanical Test

The data were obtained for each specimen using an Instron test
machine. The three-point bend test procedure aligns the specimen
so the resultant of the applied load is perpendicular to the
adhesive bond line, see Figure 1. The support noses were adjusted
384 N.T. McDEVITT AND W.L. BAUN

to a span four times the average specimen thickness. The test was
carried out by loading each specimen at a crosshead speed of 0.02
inch per minute until interfacial failure occurred. All tests were
stopped after 6 minutes. All tests were performed at room
temperature.

1_ _- - -SPECIMEN SUPPORTS

_---ADJUSTABLE SUPPORT BASE

--BASE

SHORT BEAM SHEAR TESTING FIXTURE

SPECIMEN

Figure 1. Illustration of short beam shear test geometry

ANALYSIS OF DATA

The three-point bend test, usually called the short beam shear
test, is one of the most widely used test methods for evaluation of
the shearing strength of composite materials. Several assumptions
are made in applying beam theory to this test, one of which is that
the shear stress S is distributed across the transverse face of the
specimen with a maximum at the center according to the following
Equation (1):

S = 0.75 (L/wt) (1)

THE THREE-POINT BEND TEST FOR ADHESIVE JOINTS 385

where L is the load applied at the center of the beam, w is the

width of the beam, and t is the thickness. The fracture load (L)
is obtained from the test record, and the value of S calculated
from the equation is assumed to represent the shearing strength of
the composite.

This same test has been used to study the sandwich structure
of meta1-to-meta1 adhesive bonds. This is apparently the first
attempt to apply this geometry to metal adhesive joints. No
attempt has been made to mathematically analyze our specimen from
beam theory. In the present study, the test was used only to
differentiate between the interfacial properties of various surface
preparations that had been subjected to changing temperatures
and/or environments.

Figure 2 represents the test record of a completely nonbonded

specimen. This load-displacement curve was obtained from a
specimen that was prepared by physically laying a film of FM 123-2
adhesive between two adherends of 7075-T6 aluminum. The zero
displacement point is obtained by dropping a tangent along the
linear portion of the curve. The displacement of the specimen is
plotted as a function of time. A load value, M, is obtained from
the linear portion of the curve where it intercepts with the one
minute mark of the abscissa. After two minutes the stiffness slope
changes and the load values represent slight metal deformation
mixed with the energy required to overcome the friction of the
adhesive tape opposing the slide of the two opposite faces of the
adherends. The load value, M , from this curve represents our
baseline for a completely nonbo8ded structure.

Figure 3 represents the type of curve obtained from a specimen

where the adhesive has been cured and has not been subjected to any
environmental tests. This is a baseline curve for a strongly
bonded specimen. The load value, Mb , (or slope) is obtained from
the linear portion of the curve after one minute. After two
minutes the slope changes and the load take up is more gradual. In
this region of the bonded specimen the load values represent a
slight metal deformation mixed with the energy required to overcome
the adhesion of the adhesi ve-oxide-meta1 interfaces. The energy
required to overcome these interfacial forces is distinguished by a
definitive break in the test record (L). The load value at this
point is taken as the yield strength of the bonded interface (YSI).
After the break in the curve a definite falloff in load value is
recorded, a second curve will show another load value being
recorded. This second portion of the record represents the load
value of the deformation of the adherends plus the friction
generated by the type and degree of failure of the interface. The
type of failure is distinguished by cohesive and/or adhesive. The
larger the degree of adhesive failure in the interfacial region the
closer the second value will parallel the curve obtained from the
nonbonded specimen.
386 N.T. McDEVITT AND W.L. BAUN

2590

2009

..0 1590
"'"
.-I
'-'
Q
~
1090
0
...:I
50e

TIME (min)

Figure 2. Load-displacement curve for non-bonded specimen.

45ee
L,,\

4900

3500

3009

g 2500
0
« 2099
a
...J

1500

100e

500

6
TI ME (mi n)

Figure 3. Load-displacement curve for bonded and cured specimen.

THE THREE-POINT BEND TEST FOR ADHESIVE JOINTS 387

We can evaluate bond line stiffness (BLS) from the linear

portion of these curves by using Equation (1) and relating the data
from bonded specimens to the nonbonded specimens, Sb/Sn' as in
Equation (2):

(2)
BLS

RESULTS AND DISCUSSION

The following data represent the preliminary step in the

development of a testing method that will detect subtle changes in
the chemistry of surfaces and interfaces when they are exposed to
harsh environments. Of particular interest was the determination
of changes occurring at the adhesive-oxide-metal interfaces.
Figures 2 and 3 show typical test curves of a nonbonded and bonded
test specimen. The bonded specimen was post cured and was not
exposed to any environmental tests. It represents a good bonded
joint. Figure 4 shows a typical curve from a bonded specimen that
has been subjected to environmental tests. This test record is
characteristic of a degraded bond line. The YSI break in the curve
is still present but at a very low value. The last type of curve
recorded in this study is shown in Figure 5. This represents a
bonded specimen that has totally failed inter facially due to the
environment test. Although the test record is similar to the
nonbonded specimen curve (Figure 2), Figure 5 does show a larger
area under the curve as would be expected.

Two types of information can be obtained from the test records

generated by this study. One type deals with the definitive break
in the curve and is denoted the yield strength of the interface
(YSI). The second utilizes the initial linear portion of the
curve. When the load value, obtained after one minute of
displacement, is referenced to the nonbonded specimen, a
bondline-stiffness (BLS) ratio value for the cured specimen is
obtained.

The fact that the stress generated by the three-point bend

test is driven to the interfacial region can be easily observed by
a visual inspection. The determination of failure at a specific
interface, adhesive-oxide~r oxide-metal, can only be determined by
surface instrumentation. Figure 6 shows the surfaces of
specimens that were completely opened after Instron testing. These
three surfaces are responsible for the test records shown in
Figures 3,4, and 5. Figure 6a is the surface of the failed
388 N.T. McDEVITT AND W.L. BAUN

specimen considered to be a good bond. This good bond still shows

a reasonable amount of adhesive failure (oxide metal surface
exposed). Since a lap shear and a wedge opening test on a
similarly prepared bond line showed only cohesive failure,
apparently more stress is generated in the interfacial region in
the short beam shear test. Figure 6b shows the surface of the low
strength bond line with considerably more adhesive failure. Figure
6c shows the surface of the totally failed interface. A harsh
sulfur dioxide atmosphere has generated corrosion products that are
visible on these surfaces.

g 2509
c
<C
0 2999
...J

1509

1999

509

TIME (min)

Figure 4. Load-displacement curve for test specimen with low

strength interface.

g 2590
c
<C
0 2099
...J

1509

1999

590

TIME (min)

Figure 5. Load-displacement curve for test specimen with

interfacial failure.
THE THREE-POINT BEND TEST FOR ADHESIVE JOINTS 389

Figure 6. Low magnification photographs of test specimens (a)

specimen from Figure 3, (b) specimen from Figure 4, (c) specimen
from Figure 5.
390 N.T. McDEVITT AND W.L. BAUN

The test data were generated from two groups of specimens.

Table III reports- the data from the group that was cured under
various conditions but were not subjected to environmental tests.
Table IV reports the data from the specimens that were subject to
various temperatures and atmospheric conditions.

The data spread in the YSI values for Groups A and C is very
small; therefore, the average value for these data is taken as the
control value. These data including the data spread are shown in
Figure 7 plotted against surface preparations. The data population
is small for these groups, but there is a definite indication that
the 10 volt, 10 minute and 10 volt, 16 minute anodizations should
provide the more reliable surface preparation.

32 313

Jl e-0

@
c 3000
<t
0
...J

2g e~

28 1313

27ee
1 2 3 5 6
SURFACE PREPARATION

Figure 7. Spread of load values for control specimens.

Data from the specimens subjected to the environmental tests

are shown in Table IV. Data from the more severe tests (E, F, and
G) were averaged and plotted against surface preparation (Figure
8). The data spread is larger for the specimens exposed to
environmental test, but this would be expected. The data
population again is small; however, a definite trend toward lower
Table III. Yield Strength of the Interface (YSI) and Bondline Stiffness (BLS) Data -I
:::r
from Control Specimens. m
-I
:::r
Surface Controls ::n
m
Prep m
.;,
A B C o
Z
-I
BLS YSI (lbs) BLS YSI (lbs) BLS YSI (lbs) OJ
m
Z
1 1.53 2825 2.06 3050 1.84 2758 o
-I
2 1.59 3012 2.15 3552 1.80 2793 m
Cf)
3 1.55 3200 2.12 4120 1.86 3222 -I
4 1.58 3175 2.12 4202 1.92 3202 "T1
o::n
5 1.53 3025 2.01 3725 1.84 2900
6 1.59 2875 1.98 3650 1.85 2940 »
o
:::r
m
Cf)

Table IV. Yield Strength of the Interface (YSI) and Bondline Stiffness (BLS) Data from <:
m
Specimens Exposed to Environmental Tests. L
o
Z
Surface Environment -I
Cf)
Prep
D E F G

BLS YSI(lb) BLS YSI(Lb) BLS YSI(lb) BLS YSI(lb)

1 1.53 3100 1. 76 2520 NIF* 1.62 2455

2 1.62 2960 1.82 2760 1. 76 2475 1. 70 2590
3 1.60 2780 1. 81 2945 1. 70 2820 1. 73 3045
4 1.65 3210 1.82 3145 1. 70 2900 1.53 2500
5 1.59 3000 1.88 2910 1. 70 2515 1.56 2640
6 NIF* 1.82 2400 1.69 2690 1.65 2847
w
CD
*No Interface Failure (See Figure 5)
392 N.T. McDEVITT AND W.L. BAUN

values is consistent and follows the curve for the control

specimens, that is the lOV, 10 minute and 10V, 16 minute surfaces
give overall better results.

300101
g
Q
c:(
0
....I
28fl (1

SURFACE PREPARATION
Figure 8. Comparison of the spread of load values for control and
test specimens.

Before any data were obtained from the bonded specimens it was
presumed that the ratio of the slopes,

BLS
~
M
n

would also give some indication as to changes in the surface

preparation. Examination of the BLS values in Tables III and IV
does not show any trend with respect to changes in the Surface.
Instead the BLS values remain reasonably constant for each
individual group of specimens. Therefore, the bond line stiffness
value represents a figure of- merit for the bulk adhesive and its
response to each particular environment.

The fact that the YSI values give information on the

interfacial region of the bond line and the BLS values relate
primarily to bulk adhesive properties can be shown from the
following analogy. Data were obtained previously from a wedge
opening test study using an SO environment. These specimens
were exposed to this environment ~r 168 hours at room temperature.
THE THREE-POINT BEND TEST FOR ADHESIVE JOINTS 393

Zero crack growth was recorded and no interfacial failures

observed. Since we know the wedge opening test generates , data
primarily from the bulk ad.hesive, a zero crack growth would
indicate the S02 environment is not very detrimental to the
adhesi ve strengtn properties. At the same time we know the S02
environment is quite harsh for a phosphor~ acid anodized aluminum
alloy surface as can be seen in Figure 9. More than likely the
S02 environment has some effect on the oxide-metal interface of
these specimens but the stress generated at this interface by the
wedge opening test was not sufficient to cause failure.

Figure 9. Phosphoric acid anodized surface of 7075-T6 bare

aluminum alloy subjected to S02 environment.

In the present study a number of specimens were subjected to

this same environment (Test E and F) and then evaluated using the
short beam shear test. Since the S02 environment is essentially
394 N.T. McDEVITT AND W.L. BAUN

selective, having very little effect on the bulk adhesive

properties but detrimental to the oxide-metal interface, we may
take the BLS data as indicative of the bulk adhesive properties and
the YSI data as indicative of the interfacial properties. If we
use the average value of the BLS data from Test C as our control we
can see there is a 1.7% change in the average BLS value of Test E.
On the other hand the average value for YSI (Test E) changes by
6.4% when compared to the same value for Test C. Comparing F (the
more harsh conditions) to C we can see a change of 7.5% in BLS
while the YSI value changes 9.7%. It is apparent from these data
that the S02 environment is more detrimental to the oxide-metal
interface tIlan 7 to the bulk adhesive as was indicated by the
previous study. However, it is also obvious that as the effect
of the environment becomes greater on the bond line (Test F) the
two pieces of data obtained from this test cannot remain
independent from each other.

CONCLUSION

The data obtained from this preliminary study indicate the

short beam shear test can be a valuable mechanical test for
adhesively bonded structures. This test appears capable of
following subtle changes that may occur in the interphase "between
the adhesive-oxide and oxide-metal regions of a bonded structure.
Analysis of detailed stress-strain data appears to give information
on each component of the adhesive bond, rather that just on the
resultant bonded structure.

REFERENCES

1. N. L. Rogers, National SAMPE Technical Conference, Vol. 2, p.

571 (1970).
2. N. L. Rogers, in "Proc. Symposium on Durability of Adhesive
Bonded Structures," p. 57, Picatinny Arsenal, Dover, NJ,
October 1976.
3. V. S. Srinath, in "Proc. Conference on Structural Adhesives
and Bonding", p. 24, arranged by Technology Conferences
Associates, El Segundo, CA, March 1979.
4. W. D. Bascom, Adhesives Age, p. 28, April 1979.
5. W. L. Baun, Air Force Materials Laboratory, TR-79-4126,
Wright-Patterson AFB, OR, September 1979.
6. W. L. Baun, these proceedings.
7. N. T. McDevitt, W. L. Baun, and J. S. Solomon, Air Force
Materials Laboratory, TR-77-184, Wright-Patterson AFB, OR,
October 1977.
ACCELERATED AGING PROCEDURES FOR GLUE-WOOD BONDS

AND THEIR USE FOR IN-PLANT QUALITY CONTROL

R.B. Jathar

Elmendorf Board Corporation

RFD 112, Box 303
Claremont, N.H. 03743

Bond degradation in a particulate material like

waferboards or oriented strand boards bonded with syn-
thetic adhesives like phenol formaldehyde or urea for-
maldehyde can be explained as a failure or the breaking
of the three dimensional mechanical branching of the
cured adhesive due to the stresses generated by shrink-
ing and swelling of wood as a result of moisture changes,
as well as due to the chemical degradation of the adhe-
sive which is hydrolytic in nature. In a service en-
vironment, the principal factors causing glue-wood
degradation are (1) heat, (2) moisture, (3) stresses
(generated during shrinkage and swelling), (4) micro-
organisms, and (5) chemicals.

Various bond degradation procedures -- more com-

monly known as "accelerated aging procedures" have been
developed over the years to assess the extent of deg-
radation of the adhesive joint caused by factors that
influence degradation in service. This paper discus-
ses the approaches used in devising the various accel-
erated aging test procedures, their practicality in
routine quality control and their limitations. The
experimental part deals with the comparison of four
different aging methods and has suggested one of them
for use in in-plant quality control.

395
396 R.B. JATHAR

INTRODUCTION

Ever since the introduction of exterior particleboard, there

has been a lack of precise information about several factors with
respect to its use. Areas in which the information was needed were
the strength properties, dimensional stability and durability
(the ability of the boards to retain their strength properties at
an acceptable level after extended periods of exposure. to the
weather). To seek this information it was necessary to establish
reasonable levels of initial strength as well as acceptable levels
of these properties after exposure.

Because of the relatively recent introduction of exterior

quality composite boards into structural applications, there was
not time enough to gather information resulting from long-term ex-
posure tests for those boards put under severe conditions. Thus,
products have sometimes been used on the basis of an intuitive feel-
ing regarding their long-term durability in particular use sit-
uations.

The studies involving the problem of predicting the long-term

durability have followed two broad approaahes in the glued wood pro-
ducts industry. The first approach is to identify the factors
causing degradation and attempt to determine the effect of the range
of intensities of each of these factors occurring in service environ-
ments. The principal factors causing glue wood bond degradation in
products using Urea Formaldehyde, Melamine Formaldehyde or Phenol
Formaldehyde are: a) Heat; b) Moisture; c) Stresses generated in
the glue-wood bonds during shrinkage and swelling (which are a
result of moisture changes in wood); dio Micro-orfsnisms; and e)
Chemicals. S Several studies (Gillespie ,Hergink ,Bryan and
Schniewi~d , Northcott, ~eibich and Currier ,Freeman and
Kreibich , and Troughton ) have established the above factors to
be the causes of degradation. An adhesive bonded wood joint in a
product like plywood or in a reconstituted panel like Oriented
Strand Board or Waferboard, when in service may degrade at a rate
controlled primarily by only one of the above factors or more likely
due to a £8mbination or a synergistic interaction of these factors;
Gillespie •

In light of the above, the logic behind the design of laboratory

aging procedures (provided that all the materials in use are affected
by the degrading factors in broadly the same way and the interactions
between these factors are not markedly different in natural and ar-
tifical exposure) is that it should be possible to devise a labora-
tory test incorporating all the salient features of and producing
the effects of several years of exposure to weather in a small frac-
tion of this time. This is usually attempted by increasing the
amount and frequency of exposure to these degrading influences. For
example, increasing exposure to extreme temperatures or ultra-violet
ACCELERATED AGING PROCEDURES 397

radiation or cyclic humidity conditions, without increasing their

intensities beyond those occurring in natural exposure. Thi, has
also been called the performance evaluation approach, Carrol .

The second and purely empirical approach has been to select

severe laboratory conditions which produce changes in properties
which can be correlated reliably with those which occur in practice.
Anyone, or more than one, degrading factors chosen in such a lab-
oratory test may, if necessary, exceed in its intensity beyond what
is found in natural exposure, to bring about a change which is
equivalent to and correlatable to the effect of prolonged natural
exposure. Thus, the laboratory conditions chosen to accelerate the
bond degradation may be totally unrealistic in nature. The present
ASTM D-1037 accelerated aging test is a typical example of this ap-
proach (termed as the performance simulation approach). There is
no theoretical basis for subjecting the particleboard specimens to
steaming at 950 C in order to determine the probability that the bond
will withstand long-term exposure to weather. Other examples of
such test procedures are the German standard DIN, which is a two
hour boil test, and the<-standard for exterior plywood- PS 1-74, con-
sisting of a four hour boil, dry and four hour boil procedures. Test
procedures like these prompted remarks from manufacturers like, "We
don't boil our houses," which is clearly the result of a feBl~ng
that such a test for plywood is too severe to be realistic. '

There are other problems associated with the design of a stan-

dard. One of them is the conflict between the need for a quick test
to detect mismanufacture and the need for more prolonged tests with
exclusion ability, i.e. one which can distinguish and exclude adhe-
sive types known to be unsuitable in service. The second problem
is the necessity of setting up glue line performance requirements
or the bonding quality control tests to detect mismanufacture by
simulating acttual use conditions. But this procedure becomes im-
practical from the point of a routine quality control aspect. Thus,
bonding quality requirements can be set up by using a performance
simulation test procedure which can be performed within a very short
period of time.

Thus, it is clear that there are several requirements that are

to be satisfied by accelerated test procedures and a clear distinc-
tion between the purposes that an accelerated aging test is designed
to serve is necessary. The purposes that a standard procedure can
serve are:
1) A simple, low-cost procedure that provides rapid feedback
information acd detects mismanufacture of the sort that
is known to impair the performance of the product.

2) A test procedure that distinguishes and excludes pro-

ducts made with adhesives known to be unsuitable in
applications for which the product will be used.
398 R.B. JATHAR

3) A performance simulation test procedure that can be

used to evaluate the effect of known or unknown man-
ufacturing variables on the long-term durability of
the product made with a known adhesive.

4) Similar to 3), but for products made with an unknown

or new adhesive.

It is very clear that all the test procedures which would be

used in a routine, in-plant quality control would fall in the
first category. The objective of the experimental part of this
study was, in fact, to design and select a test procedure which
will satisfy the requirements of such a test for quality control.

Before going into the experimental details, a review of the

various accelerated aging (A-A) treatments studied by several re-
searchers is presented in the following section. This review also
explains the reasons why I selected the four accelerated aging
procedures for my experiments and comparisons.

Review of Accelerated Aging Procedures

16
Lehman collected data from outdoor exposure tests and at-
tempted to correlate these data with results of standard and other
laboratory exposure tests. Standard exterior-grade particle-boards
from four manufacturers were used for this study. Standard tests
for strength properties were made in exposed and unexposed speci-
mens and on specimens exposed to laboratory aging. The exposed
specimens were tested after one-half, one, and two years. The
laboratory accelerated aging procedures consisted of:

1) Vacuum Pressure Soak and Dry (VPSD) Test; five cycles.

o
a) Dry at 220 F for 22 hours.
b) Soak in water at 700 F with 25-30-inch vacuum for
30 minutes.
c) Submerge in water at 70 0 F with 75 psi for 60 minutes.
2) ASTM accelerated aging; six cycles (AA).
3) West Coast Adhesive Manufacturess Association (WCAMA)
Test; consisting of six cycles.
a) Submerge in water at 70 0 F with 27-inch vacuum for 30
minutes.
b) Boil in water at 2l0 0 -2l2 0 F for 3 hours.
o
c) Dry at 220 F for 20 hours.

All the specimens were reconditioned at 70 0 F and 65% relative

humidity before testing for strength properties.

It was found that the WCAMA Test was the most severe, followed
ACCElERATED AGING PROCEDURES 399

in decreasing order of severity by ASTM AA and VPSD. Regression

analysis was done on the data and the correlation coefficients in
all cases were found to be very low except in case of VPSD vs. two
year exposure. The study concluded that VPSD is the best predictor
of these properties after exposure. Thus, this study proposed a
new test which serves purpose Number 3, listed earlier, but is still
not very much different from the standard ASTM procedure in terms of
its capability to be a quick quality control test procedure which
can satisfy Purpose 1.

Gressel ll studied the behavior of phenolic bonded particle-

boards subjected to two-hour boil, four-hour b~~ll7six-hour boil,
and 1, 2, and 3 cycles of the French CTB Test.' The boiled
specimens were tested wet, as well as after reconditioning at 20 0 C
and 65% relative humidity. Matched specimens were also exposed for
seven months, one and two years.

The CTB procedure is obviously not suitable as a routine quality

control procedure, due to the time required to execute it (3 weeks).
The boil tests showed that the strength properties did not deterior-
ate further after a certain period of boiling which, in case of
phenolic bonded particleboards, was found to be about two hours. He
also concluded that the particleboards bonded with a boil-proof
adhesive which reach a plateau level of their strength properties
after a certain boiling period can, in all probability, be classi-
fied to be "as weather resistant as a phenolic bonded particleboard."
The reduction in strength properties after two hour boil and wet
tests were found to be consistent with the reduction obtained approx-
imately after five years of exposure. The two hour boil test was
also found to be a very suitable procedure for quality control,
providing rapid feedback of reliable information about the durability
of the boards and was proposed for adoption as a standard test. It
is now the German standard DIN 68761.

Beech 6 subjected exterior-quality particleboards from five man-

ufacturers to outdoor exposure for two years, as well as the follow-
ing laboratory exposures.

1. Weatherdrum: Panels of chipboards fixed on a ten-sided

rotating drum (3 rph) were exposed to ultraviolet light and rows of
water jets, which operated intermittently for 16 out of each 24
hours. The total exposure was about 3,000 hours. o
2. Cyclic Test: Exposure to three days of water soak at 20 C,
one day freeze at _12°C and three days drying at 70°C. After three
cycles, recondition at 25 0 C and 65% relative humidity.
3. Two hour boil and wet test, and also test after recondition-
ing.
o
4. Hot water soak (70 C for five hours), both wet and dry
testing.
400 R.B. JATHAR

The cyclic test results correlated very well with two years
exposure. None of the other tests were as successful. Although
the boiling and hot water tests were not very effective in predic-
tion, internal bonds (IB) on wet samples gave highly significant
correlations with data obtained from outdoor exposure. This sug-
gested that two hour boil could be a simple and rapid test for pre-
dicting the suitability of exterior particleboards.

Lehman 17 designed and carried out a study to provide data on

the relationships between long-term durability and accelerated aging
tests for a variety of structural panels including waferboards.
Outdoor exposure consisted of 13 inch by 22 inch specimens mounted
vertically facing south. The accelerated aging treatments used
were:
1. ASTM D-1037 accelerated aging, 1-24 cycles.
2. VPSD, 1-24 cycles.
3. Spray-dry ASTM D-2829 (SD), 1-24 cycles. This consisted
of four hours spraying with water at 70 0 F (21 o C), four hours drying
under heat lamps at 1500 F (66°C), repeat spray dry and eight hours
rest period.
4. Two hour boil (2HB), 1 cycle.

After completion of the last wet cycle, prior to testing in all

accelerated aging tests with the exception of wet boil test, all
specimen were dried for three days at l20 0 F (49 0 C) to a moisture con-
tent (M.C.) of three or four percent and then tested.

All the specimens from the first and second cycle of spray-dry
(SD) and AA test were found to be almost equal to six or 12 months
of outdoor exposure, in respect to Modulus of Elasticity (MOE). At
the end of the 24 cycles, the tests ranked in the order of severity
were AA, VPSD, 2HB (1 cycle), and SD, with MOE retentions of 71, 80,
81 and 82 percent, respectively. In case of Modulus of Rupture (MaR),
all the accelerated tests were found to be more severe than one year
exposure, even in the first cycle. The ranking in severity after 24
cycles was found to be the same as in the case of MOE, except that
2HB was almost equivalent to AA. Internal Bonds (IB) also showed
the same ranking. No correlations were found between natural and
accelerated weathering tests or between the various accelerated aging
tests. Lehman concluded after the study that in most cases, one or
two cycles of accelerated aging were more severe than six or 12
months of natural exposure, which is the maximum period for which
the boards are exposed unprotected during building construction
(houses, etc.). Thus, considering the minimal losses of strength
properties in this period, little concern should be placed on the
durability of these panels during the construction period, provided
minimal exposure to liquid water occurs after construction. Hence,
he suggested that a larger emphasis be placed on a test of assurance
of bond quality, which based on this study could be 2HB or VPSD, or
ACCELERATED AGING PROCEDURES 401

a combination such as vacuum pressure soak -- 30 minutes under 70 0 F

(2l o C) water and 28 in. (711 mm) Hg vacuum, 60 minutes at 60 psi
o 0
(0.41 mPa) , two hours boil and wet test or dry at 120 F (49 C), and
t'est.

He proposed that, for durability testing, some cyclic test

with conditions closely approaching the natural exposure be adopted.

Based on the conclusions of the above discussed studies, it

was decided that 2HB be used as one of the accelerated aging tests
for waferboards for the purpose of this study.

An extensive amount of work has been done on the behavior of

the plywood glue-line durability, its behavior in short-term accel-
erated aging tests, as well as in actual exposures. Northcott,
Kreibich and Currier 19 compiled a review of 94 such studies dealing
mostly with plywood and some dealing with laminated timber beams.
This work was prompted by the Steering Committee for Accelerated
Testing of Adhesives (SCATA) which was formed specifically with an
objective of developing a test procedure, or several of them, by
which the service life of any glued product in any specified serv-
ice environment can be accurately predicted from short (a few weeks
or months) laboratory tests.

In an effort to attain this goal, Northcott, Kreibich and

Currier 18 studied the efficiency of eleven bond-degrade accelerating
(BOA) systems, in order to select the most potential estimator of
bond service life. BOA systems were methods (such as repeated cy-
cles of wetting and drying, or continuous soaking in water) for
accelerating the degradation of the bonds. The criterion for ac-
ceptability of a BOA system was that it must rank glues in the order
of their durabilities in exterior service. Hence, the study used a
range of highly durable to non-durable glues. The material used to
evaluate the BOA systems was 3/8 inch thick plywood shear specimens.
The criterion chosen for comparison of BOA systems was the number
of tre tment days required to reduce the breaking load of plywood
shear specimen to 100 pounds. The system which required the least
time would be most desirable. The study found that BOA #6, consis-
ting of cycles of 10-minutes boil in water, followed by immersion in
ice-cold water for 3.75 minutes, followed by 60 minutes of drying in
a forced-air oven at 225 0 F, was the most promising of the eleven
systems studied. BOA #6 ranked the glues closely to the predicted
order of exterior durability and attained strength reduction in the
least amount of time.

Based on the above study, a mUltiple boil-dry cycle test was

developed and was adopted as a standard test (ASTM 03434). This
test, however, is intended for comparing potential long-term dura-
bility of exterior adhesives rather than for use as a quality con-
trol test.
402 R.B. JATHAR

It was on the above basis that one cycle of BDA was included
as one of the accelerated aging treatments for this waferboard study.
From an exploratory test using commercially produced 7/16 inch thick
waferboards. it was found that one cycle of BDA and then immediate
measurement ofMOR reduced the MOR by about 50%. which is more than
the strength reduction obtained after the ASTM D-1037 standard
treatment.
20
Shen and Wrangham observed that the German standard (DIN)
two-hour boil test for evaluating the internal bond retentions was
not applicable to particleboards made with thick flakes. e.g .•
waferboards. because of the uneven swelling produced by boiling.
The same conclusion was reached from studies at the Ottawa Forest
Products Laboratory. done to examine the applicability of the German
Standard to waferboards. Non-uniform swelling resulted in non-uni-
form loading of the IB specimens giving lower test values. Based
on this study, the Canadian Standard applies the two-hour boil test
to a bending specimen. Shen and Wrangham. however. proposed a tor-
sion shear technique as suggested by Gertjejansen and Haygreen. 13
A good correlation between the failing torque and internal bond of
the particleboard specimen was found.

The Canadian Standard for waferboards CSA 0188(75). consisting

of two-hour boil. cool. and static bending test. when wet took about
three hours and apparently the waferboard manufacturers were not en-
tirely satisfied with this procedure. Hence. upon the request of
the waferboard manufacturers. the Eastern Forest Products Laboratory
undertook a study to develop an even quicker. yet reliable. acceler-
ated aging test that would fulfill the requirements of an in-plant
quality control test.

In an attempt to achieve the above, Shen 21 adopted the torsion

shear method in combination with two-hour test treatment as a stan-
dard laboratory procedure for evaluating the bonding of exterior
grade phenolic waferboard. The experimental work confirmed the ex-
istence of a correlation between torsion shear and Modulus of Rup-
ture only when both are wet-tested. In experiments done to assess
the influence of boiling period on torsion shear, it was found that
the first 20 minutes of boiling of the one-inch square torsion shear
specimen swelled and weakened the specimen to a great extent and
further boiling did not further deteriorate the strength properties
considerably. Thus, the study concluded that the two-hour boiling
can be shortened to 20 minutes and the entire test completed in less
than 40 minutes.
14
In a similar study by Gaudert and Szabo ,to assess the effect
of boiling period on other properties of waferboard, a striking re-
sult was observed. Data obtained from specimens taken from 10 com-
~ercia1 waferboard panels of three thicknesses showed that a plateau
ACCELERATED AGING PROCEDURES 403

condition was reached in bending strength in the first five minutes

of boiling. This pattern was found to be quite consistent within
each panel, although in individual panels the bending strength re-
tention varied from a low of 39% to a high of 58%, probably as a
result of density variations between the boards.

So, based on the above observations, it was decided to use a

20-minute boil, cool and wet test procedure in the present study.

EXPERIMENTAL DESIGN

The number of observations per treatment to be used in the

present study, or in other words the sample s~8e, was determined by
using the method outlined by Cook and Larntz. For the purpose of
sample size determination it was assumed that the mean of two treat-
ments will be considered unequal if a difference of more than 75%
of the standard deviation per observation is detected for any prop-
erty.

The sample size indicated by this method was 40 observations

per treatment. The four treatments compared were ASTM aging (AA),
BDA, 2 hour boil (2HB) and 20 minute boil (20MB); the controls
served as a reference. Considering the easy availability and the
small expense involved, it was decided to use a sample size double
the one determined statistically.

As the objective of the experiment was to study the behavior

of waferboards under the four accelerated aging treatments, it was
decided that a cross section of commercially produced waferboards
be taken for the study.

Waferboards manufactured by two different manufacturers were

collected for this study. The waferboards were taken from different
press loads manufactured on different days and times of the day,
and the test specimen cut in such a way that they came from differ-
ent locations in the press.

The following four physical properties were evaluated after

subjecting the samples to the four accelerated aging treatments.
The physical properties of the control (untreated) samples were
also measured.

1. MOR (Modulus of Rupture) was determined by the static bend-

ing test procedure as described in ASTM D-l037.
2. MOE (Modulus of Elasticity) was determined from the load-
deflection curves obtained in static bending.
3. IB (Internal Bond), or the tensile strength perpendicular
to the surface was determined from a 2" x 2" specimen per ASTM D-l037.
404 R.B. JATHAR

It was observed in the case of waferboards that the bending

specimens fail visibly due to shear and that the shear spreads across
the entire length of the specimen. This is especially true in the
cases of those which have been subjected to accelerated aging treat-
ments. For this reason, the test specimens for internal bond were
taken from separate samples instead of cutting them from the bending
specimen, as is done conventionally.
4. The Internal Torsion Shear Strength of I-inch square sam-
ples was also determined. The testing apparatus was fitted to the 13
Instron testing machine, as directed by Gertjejansen and Haygreen.

For treatments involving boiling, a special apparatus was

fabricated. Perforated steam pipes were arranged in a tray in such
a way that the pipes circulated across the length and width of the
tray. This ensured that there was no drop in temperature after the
specimens were immersed in boiling water and also maintained the
water at boiling point.

As some difficulty was encountered in testing the wet internal

bond specimens, the 2-inch square specimens were face-sanded and
the faces allowed to dry out to facilitate their gluing to the
testing blocks. All the wet IB's were tested at moisture contents
above the fiber saturation point.

The manufacturing data for the boards from the two manufactur-
ers are given below.

Manufacturer A
Board Thickness: 7/16 inch.
Avg. Density: 42 pounds per cubic foot (pcf).
Resin Content: 2.7% (solid) based on oven dry (O.D.)
weight of wafers.
Wax: 1.15%, based on O.D. weight of wafers.
Wafer Dimensions: 0.03 in. thickness (0.75 mm.), l-l~
inch square.

Manufacturer B
Board Thickness: 7/16 inch.
Avg. Density: 42.5 pcf.
Resin Content: 2.5% to 3%.
Wax: 2%.
Wafer Dimensions: 0.025 in. thickness (0.63 mm.), l-l~
inch square.
ACCELERATED AGING PROCEDURES 405

STATISTICAL ANALYSIS

Analysis of Variance (ANOVA)

As the specimens for this study were taken from different

boards (15 from Manufacturer A and 3 from Manufacturer B) and also
from different positions within a board, a pilot analysis was done
to evaluate the board effect, the position effect, and the possible
interactions of treatment by board, treatment by position, board
by position and treatment by board by position. These analyses
showed the board effect present in the case of IBIs only. However,
for the overall analysis it was decided to neglect the board effect
and analysis based on a split plot design, as laid out in Figure 1,
was adopted.

Whole Plots
~nufacturer Treatment 1 2 3 4 ....... 30

AA

A BDA
2HB
20 MB

AA

B BDA
2HB
20 MB

Figure 1. Split Plot Design for Analysis of Variance.

In the analysis of variance, the main plot analysis is that of

the two manufacturers. The sub plot analysis contains the sum of
squares for the four accelerated aging treatments, the sum of
squares for the manufacturer by treatment interactions and the
sub plot error.

Th
Koehler
Z5ANOVA was performed
) using the model:
using the program IVAN (Weisberg and

Y(IJK) M(I) + El(IJ) + T(K) 4 MT(IK) + E2(IJK)

406 R.B. JATHAR

In the above model M equals Manufacturer, T equals accelerated

aging treatment, and MT equals the Manufacturer by treatment inter-
action. EI and E2 are the errors associated with the whole plot and
sub-plots respectively. The whole plot error sum of squares is used
to test the hypothesis that the boards from the two manufacturers
had equal strength properties averaged over treatments. The sub-
plot error sum of squares was used to test the hypothesis of equality
of the treatment effects and also to test the significance of the
manufacturer by treatment interaction.

In the above analysis, a significant manufacturer by treatment

interaction was seen to be present in the case of all properties ex-
cept MOE. As the presence of a significant interaction complicates
interpretation of the main effects, a separate analysis of variance
was also done for each of the manufacturers to evaluate the treat-
ment effects more accurately. The IVAN model used for this analy-
sis was: Y(IJ) T(J) ~ E(IJ).

Regression Analysis

Linear regression analysis was done to find out if a simple

linear relationship exis'ts between the values after the ASTM accel-
erated aging and the three short accelerated aging methods investi-
gated in this study. The regression analysis was done separately
for the samples from the two manufacturers, as well as on both the
manufacturers together. The following pairs of treatments were an-
alyzed to find out if a linear relation exists.

1. ASTM Acc Aging (Y) vs. BDA (X)

2. " vs. 2HB (X)
3. " vs. 20MB (X)

The analysis was done separately for each of the four proper-
ties evaluated in the study, viz.: MOR, MOE, IB, and IB by the
torsion shear method.

RESULTS AND DISCUSSION

Table I lists the means and coefficients of variation for the

thickness, density and percent thickness swelling for the specimens
from the two manufacturers.

It was observed that the specimens from Manufacturer B exhib-

ited greater variability in the original thickness and density, but
a much lower variability in the percent thickness swelling after
the four accelerated aging treatments.
ACCELERATED AGING PROCEDURES 407

Table I. Means and Coefficients of Variation (C.O.V.) for Thick-

ness, Density, and Percentage Thickness Swelling For
Specimens from Manufacturers A and B.

Thickness (in. ) Densit;y (lbs./in 3 ) % Thickness Swell

Treatment Mean C.O. V. Mean C.O.V. Mean C.O.V.

Control A 0.423 4.25 43.5 4.18

B 0.439 11.40 43.0 5.99
Accelera- A 0.425 4.50 43.2 4.80 34.9 49.7 (M)
ted Aging B 0.451 11.5 42.5 5.4 26.7 21.6
Bond
Degrade A 0.422 4.7 43.4 3.7 33.4 20.0
(BDA)
Accelera- B 0.433 10.2 42.7 5.8 29.4 11.0
ted Aging
Two Hour A 0.422 4.7 42.9 4.5 50.S 20.3
(2HB)
Boil B 0.446 11.2 42.9 5.2 45.0 11.4
20 Minute A 0.422 4.7 43.2 3.5 37.5 30.5
Boil 0.445 11.2 10.0 34.8 (20MB)
B 42.1 16.4
408 R.B. JATHAR

The thickness changes shown in Table I illustrate that the two

hour boil (2HB) produced the maximum thickness swelling (averaging
over both Manufacturers) followed in decreasing order of magnitude:
20 MB, BDA and AA. For all treatments, the average thickness
swelling of the specimens from Manufacturer B was observed to be
smaller than that for Manufacturer A. Higher percentage of wax
used by Manufacturer B probably explains this phenomenon.

Table II contains the means, coefficients of variation and the

percent strength retentions after the four treatments for the Manu-
facturers A and B.

The results of the analysis of variance on the values of the

four strength properties obtained after the four accelerated aging
treatments are discussed in the following sections.

Modulus of Rupture: (Table III)

It was observed that the specimens from the two manufacturers
did not differ significantly with respect to their response averaged
over the four accelerated aging treatments. Comparison of the four
treatment means averaged over manufacturers by linear contrasts show-
ed that the ASTM accelerated aging treatment (AA) was significantly
different and least severe compared to the average of BDA, 2HB and
20 MB. The comparison between BDA and the average of 2HB and 20 MB
did not show any significant difference. The third comparison was
between 2HB and 20 MB and the difference was found to be insignifi-
cant.

Comparison between the four treatment means (averaged over the

two manufacturers) by Bonferroni method (Bonferroni Significant Dif-
ference or BSD) indicated that only AA was significantly different
(least severe) from all the other treatments. There was no signifi-
cant difference between BDA, 2HB and 20 MB.

The analysis of variance done separately for each of the manu-

facturers showed that for both the manufacturers AA was the least
severe treatment and significantly different from the others. BDA,
2HB and 20 MB did not differ significantly.

The same conclusions were reached from the data which were
transformed into logarithms. Transformation of the data was neces-
sary, in the case of MOR, because a pattern was observed in the
plot of the treatment means and the standard deviation within a
treatment.

Regression analysis showed that there was no statistically sig-

nificant relationship between the results of the four treatments.
 l>
(')
(')
m
r
m
:II
~
m
Table II. Table of Strength Properties of Controls and Accelerated Aged Specimens 0
l>
G')
Z
Treatment MOR (Esi) LB. (Esi) LB. bl T.S. (Esi) MOE (Esi) G')

Mean (C.O.V.) %Reten- Mean (C.O.V.) %Reten- Mean (C.O.V.) %Reten- Mean (C. o. V• ) %Reten- ."
:II
tion tion tion tion 0
(')
m
0
C
:II
Control A 3467 (19.6) 80.14 (26.9) 94.62 (25.0) (15.6) m
643,400 en
B 3231 (28.4) 66.31 (27.9) 85.56 19.3) 624,200 (25.0)

AA A 2216 (27.3) 63.9 17.44 (66.5) 21.8 17.45 (63.2) 18.4 484,600 (21.8) 75.3
B 2652 (26.8) 82.1 28.82 (34.0) 43.5 27.93 (43.0) 32.6 460,300 (34.9) 73.7

BDA A 1870 (18.4) 53.9 27.30 (51.3) 34.0 28.33 (55.1) 29.9 397,500 (17.1) 61.8
B 1620 (22.0) 50.1 22.08 (32.2) 33.3 22.15 (37.1) 25.9 326,000 (24.9) 52.2

2HB A 1683 (21. 6) 48.5 15.26 (64.4) 19.0 16.18 (49.6) 17.1 425,000 (15.7) 66.1
B 1756 (28.1) 54.3 20.46 (31. 3) 30.9 19.15 (43.2) 22.4 346,700 (22.4) 55.5

20MB A 1742 (16.8) 50.2 23.54 (53.9) 29.4 20.67 (41.3) 21.8 407,600 (14.6) 63.4
B 1600 (19.8) 49.5 24.94 (29.3) 37.6 21.24 (33.6) 24.8 317,100 (19.0) 50.8

~
o
co
410 R.B. JATHAR

Table III. Modulus of Rupture.

i.) Analysis of Variance

Source d.£. S.S.. M.S. F

Manufacturer 1 54753 54753 .1891

Error-1 58 .16790E~08 • 28949E+06
Treatment 3 • 24618E+08 .82061E~07 39.97 *
Manu. X Tr. 3 .41638E~07 • 13879E+07 6.76 *
Error-2 174 .35723E . 1-D8 .20531E-+06
Total 239 .81350E~08

*significant at 5%

ii.) Comparison Between Treatment Means for MOR using Bonferroni

Significant Difference (BSD).

a) Averaged Over Manufacturers A and B

20 ME BDA 2HB AA
1676 1734 1757 2458

b) Means for Manufacturer A

20 ME BDA 2HB AA
1764 1825 1703 2245

c) Means for Manufacturer B

20 ME BDA 2HB AA
1587 1615 1811 2672

(Underlined means are not significantly different at 5% level).

iii.) R2s of Regression for Modulus of Rupture

Manufacturer AA - BDA AA - 2HB AA - 20 ME

A 0.0001 0.0366 0.0279

B 0.0408 0.0158 0.0705
Overall 0.0000 0.0005 0.0160
In the analysis of variance: d.£. : Degrees of Freedom; 8.S. : Sum
of squares of deviations; M.S. : Mean Square; F: Variance Ratio.
ACCELERATED AGING PROCEDURES 411

Internal Bond: (Table IV)

It was found from the analysis that the two manufacturers did
not differ with respect to their average response over the four
treatments, i.e., the mean of IBs averaged over the four treatments
for Manufacturer A was not significantly different from that of B.

Using the BSD criterion, it was found that 2HB differs signi-
ficantly, i.e., it is the most severe test in the case of lB. The
BDA treatment was found to be least severe. However, there was not
significant difference between AA, BDA and 20 MB.

From the analysis of variance done individually for A and B,

it was found that 2HB was the most severe treatment on IB in the
case of both manufacturers. However, for Manufacturer A, 2HB did
not differ significantly from AA and 20 MB and for Manufacturer B,
2HB did not differ significantly from BDA and 20 }ffi. No correlation
was found between any of the pairs of treatments in the case of IBIs.

IB Computed by Torsion Shear Method: (Table V)

Results obtained from the combined analysis (Manufacturers A

and B combined) followed exactly the same order as in the case of
internal bond strengths measured by the standard method.

Results from the analysis done separately on the two manufact-

urers also followed the same pattern as in the case of conventional
IBIs. Regression analysis also gave the same results as in the case
of conventional IBIs.

Modulus of Elasticity: (Table VI)

In the case of MOE, the specimens from the two manufacturers

were found to be significantly different. The average MOE for A
averaged over the four treatments was significantly higher than the
mean for B. There was no interaction between manufacturers and
treatments. The mean MOE after each of the treatments was consis-
tently higher in the case of A.

The comparison between the four treatment means averaged over

the two manufacturers, using the BSD criterion, indicated that AA
was significantly different (least severe) from BDA, 2HB and 20 MB.
BDA was the most severe of the four treatments, but not significant-
ly different from 2HB or 20 HB.

Because no treatment by manufacturer interaction was present,

separate analyses were not performed for MOE.
412 R.B. JATHAR

Table IV. Internal Bond.

i.) Analysis of Variance

Source d. f. S.S. M.S. F

Manufacturer 1 203.18 203.18 .8497

Error-1 56 13391 239.12
Treatments 3 1682.8 560.93 9.498 *
Manu. X Tr. 3 1544.0 514.67 8.715 *
Error-2 168 9921. 7 59.057
Total 231 26742

*significant at 5%

ii.) Comparison Between Treatment Means for IB using BSD.

a) Averaged over Manufacturers A and B

2HB 20 MB AA BDA
18.11 23.22 23.85 25.23

b) Means for Manufacturer A

2HB AA 20 MB BDA
16.25 19.15 22.89 28.01

c) Means for Manufacturer B

2HB BDA 20 MB AA
19.97 22.44 23.55 23.19

(Underlined means not significantly different at 5%)

iii.) R2 s of Regression for IB

Manufacturer AA - BDA AA - 2HB AA - 20 MB

A 0.0640 0.3913 0.2107

B 0.0455 0.0242 0.0130
Overall 0.0067 0.2743 0.1129
ACCELERATED AGING PROCEDURES 413

Table V. Internal Bond by Torsion Shear.

i.) Analysis of Variance.

Source d.f. S. S. M.S. F

Manufacturer 1 . 78478E+0l .78478E+01 .3848Et03

Error-1 44 8973.0 203.93
Treatments 3 1466.7 488.91 7.362 *
Manu. X Tr. 3 1580.8 526.94 7.935 *
Error-2 132 8765.8 66.408
Total 183 20786

* significant at 5%

ii.) Comparison Between Treatment Means for IB by TS using BSD.

a) Averaged over Manufacturers A and B

2HB 20 ME AA BDA
18.8 22.0 22.8 26.73

b) Means for Manufacturer A

2HB AA 20 ME BDA
18.32 18.71 22.51 30.87

c) Means for Manufacturer B

2HB 20 ME BDA AA
19.28 21.48 22.59 26.89

<Underlined means are not significantly different at 5%)

iii.) R2 s of Regression for LB. by Torsion Shear.

Manufacturer AA - BDA AA - 2HB AA - 20 ME

A 0.4849 0.1838 0.0104

B 0.0252 0.0006 0.2604
Overall 0.0754 0.0537 0.2091
414 R.B. JATHAR

Table VI. Modulus of Elasticity

i.) Analysis of Variance

Source d.L S.S. M.S. F

Manufacturer 1 • 27881E ... 12 . 27881E+12 21.08 *

Error-1 42 • 55545E·H2 • 13225E+ll
Treatments 3 • 42987E+12 .14329EH2 14.51 *
Manu. X Tr. 3 .37502E+ll •12501E+ll 1.266
Error-2 126 • 12439E . H3 • 98720E+10
Total 175 • 25455E+13

* indicates significance at 5%

ii.) Comparison Between Treatment Means for MOE using BSD.

Averaged Over Manufacturers A and B

BDA 20 MB 2HB AA
350800 354100 376200 472300

(Underlined means not significantly different at 5%)

2
iii.) R s of Regression for MOE.

Manufacturer AA - BDA AA - 2HB AA - 20 MB

A 0.0073 0.0414 0.0387

B 0.0003 0.1443 0.0258
Overall 0.0011 0.0386 0.0123
ACCELERATED AGING PROCEDURES 415

No correlation was found between any of the pairs of treatments

in the case of MOE's.

CONCLUSION

The study has shown that for the properties MOR and MOE, the
three alternative accelerated aging tests, viz., 2HB, 20 MB and BDA
are significantly more degradative in effect than the present stan-
dard ASTM D-I037-72a accelerated aging treatment. This behavior was
found to be quite consistent irrespective of the manufacturer from
which the specimens were taken, the boards from which the specimens
were cut and the position of the specimens in the press.

The three alternative treatments, viz., 2HB, 20 MB and BDA were

almost equally degradative in effect.

It can be concluded that all of the alternative treatments in-

vestigated in this study could be used as substitutes for ASTM ac-
celerated aging test with respect to the bond degrading effect on
MOE and MOR. If values of MOE and MOR obtained after ASTM aging
indicate mismanufacture in the test specimens, the alternative aging
treatments investigated in this study will also yield values of MOR
and MOE which will lead to the same conclusion.

In the case of internal bond strength (measured conventionally

and by the torsion shear method), the 2HB (two hour boil, cool and
test wet) treatment was found to be significantly more degradative
in effect when the values for the manufacturers were averaged to-
gether. The IB values, however, exhibited a variation depending on
the boards from which the specimens came and also to some extent
the position of the specimens in the press. This highly variable
behavior of waferboards with respect to internal bond strength is
not unusual and ~6 been observed by others. For this reason,
Shen and Wrangham suggested that the two hour boil test should
not be applied to IB specimens.

Thus, if a test procedure were to be selected based on the

findings of this study, anyone of the three alternative treatments
will satisfy the requirements of a quick in-plant quality control
test providing rapid feedback regarding bond quality. Also, these
treatments are inexpensive to perform as compared to ASTM aging
treatment. These tests could be coupled with static bending pro-
viding a measure of MOE and MOR. The study also shows that 20 MB
can accomplish the quality control objectives as well as the 2HB
treatments and faster. Thus, 2HB could be eliminated in any choice
to be made between the three alternative tests.
416 R.B. JATHAR

Between 20 MB and BDA, from a practical standpoint, BDA has an

advantage over 20 MB, because the one hour drying in the BDA treat-
ment removes the difficulty encountered in wet testing of IB speci-
mens (if necessary). Also, the BDA treatment (10 minutes boil, 4
minutes in ice cold water and 1 hour drying at 2l2 o F) does not take
much more time than the 20 MB (20 minutes boil, 30 minutes cooling
with water and wet test).

ACKNOWLEDGMENTS

I wish to express my sincere appreciation to Dr. John G.

Haygreen, Department of Forest Products, University of Minnesota,
for the guidance provided for the above study.

I would also like to gratefully acknowledge the encouragement

and support provided by Paul J. Heenan, President, Elmendorf Board
Corporation, Claremont, New Hampshire, to participate in the ACS
symposium in Kansas City.

REFERENCES

1. American Society for Testing and Materials, "Standard Methods

of Evaluating the Properties of Wood Base, Fiber and Particle
Panel Materials," ASTM Designation D 1027-72a, 1972.
2. Deutsche Normen, "Flachpresplaten fur das Bauwesen," DIN 68763,
Beuth Vertrieb GmbH, Berlin 30, 1971.
3. Anonymous, "Mat Formed Wood Particleboard," Commercial Standard
CS-236-66, U.S. Department of Commerce, 1966.
4. Anonymous, Forest Products J., 16(6), 19 (1966).
5. E. L. Bryan and A. P. Schniewind, Forest Products J., 15(4),
143 (1965).
6. J. C. Beech, "An Accelerated Test for Predicting the Perform-
ance of Chipboards for Exterior Use", Research Paper IS,
Princess Risborough Laboratory, Buckinghamshire, U. K., 1973.
7. M. N. Carrol, Forest Products J., 28(5), 23 (1980).
8. M. N. Carrol, "We Still Don't Boil Houses: Part II. Test
Procedures for Particleboard used in General Building Con-
struction", Paper presented at Washington State University
Particleboard Symposium, 1980.
9. H. G. Freeman and R. E. Kreibich, Forest Products J., 18(7),
39 (1968). --
10. R. H. Gillespie, Forest Products J., 15(9), 369 (1965).
11. P. Gresel, "Utersuchungen an Freiberwitterten Spanplatten.
Holz als Roh-und-Werkstoff", 27 (10), 142 (1968).
12. P. Gresel, "Utersuchungen an Freiberwitterten Spanplatten.
Holz als Roh-und-Werkstoff", 27 (10), 366 (1969).
ACCELERATED AGING PROCEDURES 417

13. R. GertjeJansen and J. G. Haygreen, Forest Products J., ~

59(1971) •
14. P. Gaudert and T. Szabo, "Technical Report", Forintek Canada
Corp., Ottawa, 1980.
15. B. G. He eb ink , Forest Products J., 17(1), 59 (1967).
16. W. F. Lehman, "Durability of Exterior Particleboard", Wash-
ington State University Particleboard Symposium, No.2, 275-
305, 1968.
17. W. F.Lehman, "Durability of Composition Board Products", Pro-
ceedings of Particleboard Symposium, No.11, 351-368, 1977.
18. P. L.Northcott, R. E. Kreibich and R. A. Currier, Forest
Products J., 18(5), 58 (1968).
19. P. L. Northcott, R. E. Kreibich and R. A. Currier, "Supplemental
Bibliography and Literature Review Report No.1", Task Group
on Accelerated Treatment Design, SCATA, Unpublished Report,
1964.
20. K. C. Shen and D. Wrangham, Forest Products, J., 21(5), 230
(1971). --
21. K. C. Shen, "A Proposed Rapid Accelerated Aging Test for Ex-
terior Waferboards", Eastern Forest Products Lab, Ontario,
Information Report OP-X-160E, 1977.
22. G. W. Snedocor and W. G. Cochran, "Statistical Methods, Sixth
Ed.," Iowa State University Press, Ames, Iowa, 1967.
23. G. E. Troughton, J. Wood Sci., 1(3), 172 (1967).
24. G. E. Troughton and S. Z. Chow,~Wood Sci., 1l. 29 (1968).
25. S. Weisberg and K. J. Koehler, "IVAN Users Manual", University
of Minnesota, School of Statistics, 1976.
26. Centre Technique du Bois, "Marque de Qualitie des Panneau de
Particles", CTB-H, Agglos 454, C.T.B., Paris, 1975.
27. Association Francaise de Normalization, "Epreuve de Vieilliss-
ment Accelere Par La Method Dit V313", NF 51263, AFNOR, Paris,
1972.
28. R. D. Cook and K. Larntz, "Sample Size Determination for 'Fixed
Effects' ANOVA Models", Technical Reprot #212, University of
Minnesota, 1973.
PEEL-STRENGTH AND ENERGY DISSIPATION

T. Igarashi

Department of Technology
Gunma University
Kiryu, Gunma 376, Japan

The effect of energy dissipation on peel-strength

is discussed on the basis of mechanics. In the case of
T-pee1ing of specimens consisting of two flexible tapes
bonded by elastic adhesive, steady peel-strength is sim-
ply given as 2f = r, where f is the peeling force per
unit width of the tape and r is twice the surface energy
of the fracture surface. However, in the cases where the
energy is dissipated in the adhesive layer or in the ad-
herends, another term w' representing the energy dissi-
pation is to be added to the peeling force, i.e. 2f =
r + 2w'. The energy w' associated with the deformation
of the adhesive layer is given by the work done at the
tip of the crack per unit length of the adherend during
a cycle of the elongation to fracture and subsequent recov-
ery of the adhesive. This energy w' can be estimated from
an area u' enclosed by the stress-strain hysteresis curve
obtained from extention test of the adhesive and the
thickness h of the adhesive layer, as 2w' = hu'. The values
of u' obtained from experimental studies on the dependence
of peeling force on the thickness of the adhesive layer
agree with these obtained from extent ion test of the
adhesives. On the other hand, the energy w' associated
with the energy dissipated in the adherends is given by
the area enclosed by moment-curvature hysteresis curve
obtained from a bending test of the adherends. Especial-
ly in the case of elastic-plastic adherends such as
mild steel, w' is approximately given as w' = 3EIe e ,
where E is the Young's modulus, I is moment of ineft~a
of the cross section, e is the bending curvature at
yielding, and e is theecurvature of the adherends dur-
p
419
420 T.IGARASHI

ing the peeling. Experiments show that the difference

between the measured peeling force for the specimens
consisting of the mild steel adherends and for the
specimens consisting of flexible canvas adherends,
which corresponds to the energy dissipation in the
mild steel adherends, agree with the values calcuH.ted
using the relation w' = 3EIe e •
e p

J;,IST OF SYMBOLS

E Young's modulus of adherend pla.te

f peeling force per unit width
free energy of a system
free energy of elastic deformation of a system
surface free energy of a system
thickness of adhesive layer
I moment of inertia of cross section of adherend plate
M bending moment of adherend plate
M bending moment of adherend plate at which yielding of the plate
e
just takes place
Mp bending moment of adherend plate during steady peeling
R radius of curvature of adherend
t thickness of adherend plate
u elastic strain energy per unit volume of adhesive to be
recovered during the recovery of the deformation
u' energy dissipated per unit volume of adhesive during deforma-
tion and the subsequent recovery
U' energy dissipated in a system during deformation and the sub-
sequent recovery
w' energy dissipated per unit width and unit length of adhesive
layer or adherend plate during deformation and subsequent re-
covery i.e. U' for peeling specimen with unit width
x distance measured along direction of propagation of peeling
y distance measured from neutral line of bending plate
Ye y at elastic-plastic boundary
r twice surface energy per unit area of fracture surface
E normal strain, i.e., elongation
t dE/dt
fracture elongation
yield elongation
maximum elongation corresponding to maximun curvature of ad-
herend during steady peelil~g
e curvature of adherend
ee curvature of adherend plate at which yielding of the plate
just takes place
ep maximum curvature of adherend plate during steady peeling
er maximum residual curvature of adherend plate after peeling
PEEL-STRENGTH AND ENERGY DISSIPATION 421

El peel angle
stress
fracture stress
yield stress
maximum stress corresponding to maximum curvature of adherend
during steady peeling

INTRODUCTION

The original study of peeling of flexible tapes on the basis

of mechanical point of view was carried out by Hata. 1 For the peel-
ing of a flexible, inextensible tape bonded to a rigid surface, he
proposed a very simple and elegant relation,

f = f/(l - cosEl) (1)

where f is the peeling force per unit width of the tape and 0 is
the peel angle. f is named as the work of adhesion, which is equiv-
alent to twice the surface energy of fracture surface (Figure 1).

,f

W/7////##~/ j
Figure 1. Peeling of flexible tape
from a rigid surface.

This important relation is the basis of later studies of peeling.

Further studies on peeling were made by Bikerman,2 Spies,3
Kaelble,4 and Yurenca. 5 In their treatments, the deformation be-
havior of the adherends was assumed to obey an elastic equation
for bending plates. Gardon 6 pointed out that steady peel strength
depends on the energy of fracture, and derived a relation between
the glue line thickness, the diameter of the tear tip, and the
tear energy. Orowan 7 has mentioned that the strength of adhesive
joints is derived from the viscoelastic deformation of adhesives.
He also pointed out the importance of energy dissipation. Gent
and Petrich 8 have discussed the role of the energy dissipation
during the deformation of viscoelastic materials.
In the present paper, we will consider the effect of energy
dissipation in both of the following cases: case one when the en-
ergy is dissipated in the adhesive layer 9 ,and in the other it is
dissipated in the adherends. lO
422 T.IGARASHI

BASIC EQUATION OF PEELING

In the case of T-peeling of specimens consisting of two flex-

ible tapes, such as canvas, bonded by an elastic adhesive (Figure
2), the steady peeling force, i.e.,force required to continue the
peeling, is derived from the following thermodynamic condition

of = of + of s = 0 (2)
e
where F is the free energy of the system, Fe is the free energy of
the elastic deformation and Fs is the surface free energy.
As the work done by the external force f is stored as the po-
tential energy i.e. Fe' the following e~ression is obtained for
the. specimen with unit width
(£b)
-F = 2 ) fdx(application) (3)
e 0
where the integration with respect to the displacement x of the
point of the application is taken to the point at which the elon-
gation of the adhesive layer reaches the breaking point. The var-
iation of the surface free energy accompanied by the progression
of the crack dx is given by

OFs = rox(crack) (4)

where r is twice the surface free energy per unit area. Frnm equa-
tion (3)

-205 fdx(application) (5)

-zJI- aX(!!aCk) dx(application)ox(crack) (6)

Because the adherends are assumed to be inextensible,

dx(application) = dx(crack) (7)

Equation (6) is integrated with respect to df to the peeling force

f(e:b)'

-oFe = -Zf(£b)ox(crack) (8)

From Equations (Z),(4) and (8), omitting the parentheses for F and
x, one can obtain

2f =r (9)

For peeling accompanied by energy dissipation one may adopt

the following expression instead of Equation (2)
PEEL-STRENGTH AND ENERGY DISSIPATION 423

f
t
'": d X """",,,Uon

Adherend (Canvas)

ActIesive l ayer

dx c.ock

~..: dx owicotion
~
f

Figure 2. T-peeling specimen with flexible, inextensible

adherends and rubbery adhesive layer.

of = oFe + oFs + oU' = 0 (10)

where U' is dissipation energy. If the energy is dissipated in the

adhesive layer, U' becomes for the specimen with unit width

oU' huox(crack) (11)

where h is the thickness of the adhesive layer and u' is the ener-
gy dissipated per unit volume of the adhesive during the deforma-
tion to fracture and subsequent recovery, i.e. the area enclosed
by stress-strain hysteresis loop obtained from extent ion test of
the adhesive(Figure 3).

~ ------------------ ~

Figure 3. Stress-strain curve of adhesives. Materials

consisting of adhesive layer are prepared for
simple extension test. Specimen is first ex p
424 T.IGARASHI

tended to the elongation just before its break-

ing point and is then returned immediately. In
actual peeling process, energy corresponding
to u' is dissipated in the adhesive layer
and that to u is stored as elastic strain
energy of residual adhesive layer.

With the same consideration as mentioned above, Equations (10) and

(11) lead to the following relation instead of Equation (9)

2f r + hut (12)

On the other hand, let us consider peeling of specimens con-

sisting of inelastic adherends bonded with elastic adhesive. In these
cases energy may be dissipated not in the adhesive layer but in the
adherends during the propagation of peeling. With propagation of
peeling, any part of the adherend is subjected to one cycle of de-
formation of bending and the subsequent recovery(Figure 4). The

f f
t t

Tm///T//T///
Figure 4. Propagation of peeling. Let us fix our eyes
on any point A fixed on the adherend. With
the start (A), is bent (Al), and then it is
placed from A to Al and A2. Accordingly the
adherend at this point, which is straight at
the start(A), is bent (AI), and then it is
straightened again (A2).
PEEL-STRENGTH AND ENERGY DISSIPATION 425

energy dissipation may be associated with this deformation. Then

same treatment as above, peeling force becomes

2f = r + 2w' (13)

where w' is the energy dissipation per unit width and unit length
of the adherend during the bertding cycle.

DETERMINATION OF DISSIPATION FUNCTION FOR BENDING

The energy dissipation function w' can be obtained from the

relation between the bending moment M and bending curvature e of
the adherend as

w' = ~ Mde (14)

where the integration should be carried out for a cycle of the mo-
ment- curvature hysteresis loop(Figure 5). Let us estimate this

M~
A
- - - - - - -.:;,.-""
- ------..,
B

e., e

f--------- ------ - ----: 0'

E

Figure 5. Bending moment and curvature of adherend. Corre-

sponding to elastic deformation of adherend,bend-
ing moment of adherend plate increases linearly
with increase in its curvature(OA). After begin-
ning of yielding, increase in the bending moment
levels off to Mp corresponding to steady peeling
(AB). With propagation of peeling, curvature
of adherend decreases. Curve BC corresponds
to "spring back" of the adherend plate.
Curve CE corresponds to "stretching" process and
shape of the curve resembles bending process. r e
is the residual curvature of adherend of the spec-
imen after it has been removed from the cramp.
426 T.IGARASHI

value for mild steel adherend assuming it behaves as an elastic-

plastic body. In Figure 6 a schematic stress-strain curve for mild
steel is shown with a solid line. To reduce the difficulties in
mathematical analysis it may be convenient to approximate the
stress-strain curve for mild steel to the broken line representing
elastic-plastic property.

d
d. - ._ -.-....
····r-······-----~

Figure 6. Stress-strain curve of mild steel. Actual stress-

strain curve is approximated to broken line in
calculation. Ee means effective strain and Oeis
the corresponding stress.E p and 0 are strain and
stress respectively correspondingPto maximum
curvature of adherend du~ing steady peeling.

-1-

Figure 7. Side view and cross section of the adherend plate.

As unit length of the adherend is shown in
this side view, e means the curvature and
R is the radius of curvature of the adherend.
The specimen has unit thickness.
Yield areas are shown by hatch in the
cross section. Ye is the coordinate of
elastic-plastic boundary.
PEEL-STRENGTH AND ENERGY DISSIPATION 427

The bending moment of a plate of the adherend is generally

given by
M'" (t/2 a(y)dy (15)
)-t/2
where t is thickness of the adherend p1ate.y is the distance measured
from the neutral line. and a(y) is the stress directed along the
length of the plate (Figure 7).

Within the linear proportional part of the stress-strain

curve in Figure 6, the relation

a(y) = e:(y)E (16)

is valid, where e(y) is the elongation and E is Young's modulus 6f
theadherend. From Equations (15),(16) and the following geometri-
cal relation from Figure 7:

e(y) = ye , (17)

we obtain the relation,

M = EIe (18)

where the moment of inertia of cross section I of the adherend is

given by

(19)

Equation (18) corresponds to the linear par.t OA of the M-e~curve in

Figure 5.

After yielding of the adherend has taken place, the integration

of M should be separated into two terms; one is for the elastic part
and the other is for the plastic part, i.e •• the yielded part.In the
latter, the a-e relation is assumed to be

a(x) = const. = e:eE (20)

where ee is yield elongation. Then.

M _ Ee ~Ye y2dy + 2<eE ) t/2 ydy (21)

-Ye Ye

(22)

where Ye is y at the elastic-plastic boundary. From Equation (17),

Ye is expressed as
428 T. IGARASHI

Ye = Ee/ S (23)
On the other hand. by representing the curvature at the initiation
of yielding by 8e • also from Equation (23). and substituting t/2
for Ye one obtains

Hence. from Equations (19).(23) and (24). M becomes

1 Se 2
M = --2-- ElSe { 3- (--S--) }. (25)

Equation (25) gives the curve AB in Figure 5. where Sp is the max-

imum curvature of the adherend during the peeling.

Here. let us summarize the properties of the mild steel plate

used as adherends in our experiments. The mechanical properties
are shown in Table I. The relation between stress and strain is

Table I. Mechanical Properties of Mild Steel.

2
Tensile Strength 35kg/mm
Yield Strength 20 u

Yield Elongation 0.08--0.1 %

Young's Modulus 2l.000kg/mm2

just like tha~ shown in Figure 6. At an elongation of 0.15 - 0.2%.

about 30kg/mm of stress is observed. and then the stress gradually
increases with increase in strain. se in Figure 6 is thou ht to be 3
the effective yield strain and is estimated to be 1.5xlO- (0.15%).
The value of curvature at the initiation of yielding 8e is estimated
to be 0.1(cm- 1) for an adherend with thickness 0.3mm from Equation
(24). The value of El is estimated to be 4.7kgcm from Equation (19)
and the value of Young's modulus shown in Table I.

Equation (25) indicates that value of the moment will increase

with increase in 8 after the initiation of yielding and will level
off for 8 > 28 e • However. the value of the maximum curvature 9p is
estimated to be about 2 (cm- 1). Hence the value of the moment after
initiation of yielding can be regarded as constant. which is given
from Equation (25) and is denoted with ~:
3
Mp = --2-- ElSe (26)

Further. taking account of the experimental fact that

(27)
PEEL-STRENGTH AND ENERGY DISSIPATION 429

one can finally approximate the M-9 hysteresis loop to a rectilin-

ear curve O+A'+B+D'+E in Figure 5. Thus the dissipation function
w' of Equation (14) is simply given as

w' (28)

(29)

EXPERIMENTAL AND RESULTS

Specimens for the peel tests are prepared from hot-pressed

Viny10n-Ny10n canvas (adherend No.1-No.4 in Table II) and mild
steel plates (adherend No.5 in Table II) bonded with a rubbery ad-
hesive layer (Table III). The canvas is treated with "Resorcino1-
forma1ine-latex" prior to press. On the other hand, the surface of
mild steel plates are treated with "Chem10ck"(Hughson) to promote
affinity for rubbery adhesive layer. The rubber vu1canizates used
as the adhesive layer are tested by simple extension to fracture.
The specimens are prepared with 2mm width and cut with an ASTM C
dumbe11 die. The average values of fracture elongation and stress
(Eb and 0b) are first measured. Then, other specimens are elongated
to these average values, and if they did not fracture they are im-
mediately unloaded at the same speed to obtain stress-strain curve
and to evaluate energy dissipation. Results of the tensile tests

Table II. Preparation and Properties of Specimens.

Adhesive Steady Peeling

Specimen No. Adherend.
No. h(mm) Force (kg/cm)

1
1 Canvas G1 j dep. on thick. II
0.5
2 Canvas G2 ..j.. dep. on thick. I!
3
0.5 indep. of thick. II
3 Canvas G3 ..j..
4.0
3
indep. of thick.
4 Canvas G4 1
3.6 ~ 5.3

5 Mild steel G4 1 10.9 + 13.4

Cross-head speed:10Omm/min for No.1-No.3, 50mm/min for No.4&No.5

II: for the dependence of peeling force on thickness, see Figure 8.
430 T.IGARASHI

Table III. Composition of Rubber Adhesives.

PHR for
Gl G2 G3 G4

BR - 40 40 -
SBR 100 60 60 100
HAF Carbon 45 - - -
FEF Carbon - 1 - -
MT Carbon - - 20 -
Diatomaceous Earth 30 20 - -
CaC03 (surface-treated) - 40 - -
Chalk - 50 - -
Process Oil 30 15 - -
ZnO 5 5 5 5
Stearic Acid 1.5 1.5 1.5 1.5
S 2 2 2 2
TMTD 0.1 0.1 3 0.2
CZ 1 1 - 1

l50°C,15min Press Cure.

Table IV. Results of Tensile and Peeling Tests.

Tensile Peel
Specimen No.
Cfb(kg/cm) Eb u' (kg/cm 2 ) r(kg/cm) u' (kg/cm 2)
1 155 7 413 9.0 220
2 25.6 7.3 29 8.5 27
3 20 1.0 0 4.0 0

Strain rate(E)=5Omin- 1 •

for these rubber vulcanizates are shown in Table IV. Results of the
peeling tests with the specimens No.l- No.3 are shown in Figure 8
and Table IV. The dependence of peeling forces on the thickness of
adhesive layers lead to the values of u' using Equation (12). The
values of r are obtained using extrapolation (h+O) in Figure 8.
In the cases of the peeling tests with specimen No.4 and 5,
difference in the peeling forces resulting from the difference of
their adherends become the point of consideration (Table V).
To calculate the values of 2w' using Equation (29), the maximum
values of curvature of adherends during peeling ep are observed
photographically.
PEEL-STRENGTH AND ENERGY DISSIPATION 431

~ 100
E Specimen
u
No.1
.,."-
CI

N 50

2
h (mm)

Figure 8. Dependence of peeling force on thickness of

the adhesive layer. Adhesive layer with con-
siderable energy dissipation:No.l, with medi-
um energy dissipation:No.2, without energy
dissipat ion: No.3. All adherends of the spec-
imens in this figure are made of canvas.

Table V. Effect of Difference in Adherends.

2f(kg/cm) 2w'
(obs.)
v /I ep Calculated
Specimen No. using obs. 1111
(mm/min) (obs.) Eq. (29)
4 5
(Canvas) (Steel)
1000 2.1 5.3 13.4 5.9 8.1
50 1.8 5.3 13.2 5.1 7.9
5 2.2 3.6 10.9 6.1 7.3
# v:cross-head separation speed.
#/1 Observed values of 2w' are obtained from the difference between
observed 2f for (Steel) and that for (Canvas).

DISCUSSION

The relation between peeling force and thickness of the adhe-

sive is found experimentally to be linear as expected from Equa-
tion (12) in the case of specimens with canvas adherends. In
432 T.IGARASHI

particular with the adhesive with no energy dissipation, the

strength is independent of the thickness of the adhesive as expect-
ed from Equation (12) by assuming u'=O. These results signify that
in the case of fracture of materials with energy dissipation, ex-
ternal force is required to do the additional work associated with
the deformation to break the materials even in the steady peeling.
Agreement between the values of u' obtained from stress-strain
curve of simple extension tests and those obtained from the depen-
dence of peeling force on the thickness using Equation (12) is fair-
ly good (Table IV).
In the same manner, energy dissipated in the adherends will
also make a contribution to the peel strength. Here the difference
in the peeling force between specimen No.4 and No.5 is thought to
result from the energy dissipated in the elastic-plastic adherends
of mild steel (Table V). Agreement of 2w' calculated using Equation
(29) with those obtained from experiment is rather satisfactory.
The "tough" materials will become "tough" composites when they are
bonded with the same adhesive.

Thus the significant role of energy dissipation in fracture

strength is apparent from this study.

Finally i t should be noted that the problems discussed here

are exclusively concerning the propagation of peeling. The force
required to initiate the peeling is generally not equal to the
steady peeling force 11~12.

REFERENCES

1. T.Hata, Kobunshi Kagaku (J.Polym.Chem.Japan) i, 67 (1947)

2. J.J.Bikernu>n, J.App1.Phys. 28,1484 (1957)
3. G.J.Spies, Aircraft Eng. 25-,-64 (1953)
4. D.H.Kaelb1e, Trans. Soc. Rheo1. 2, 161 (1959); ibid. i, 45 (19
60)
5. S.Yurenca, J.App1.Po1ym.Sci. ~, 136 (1962)
6. J. L. Gardon, in "Treatise on Adhesion and Adhesives", Vol. 1,
R. L. Patrick, Ed., Marcel Dekker, Inc., New York,1967.
7. E.Orowan, J.Franklin Inst.,290, 493 (1970)
8. A.N.Gent and R.P.Petrich, Proc. Royal Soc. (London)Ser.A 310,
433 (1969)
9. T.lgarashi, J.Polym.Sci.Polym. Phys.Ed.13, 2129 (1975)
10. T.lgarashi, Nihon Setchaku Kyokai Shi (J.Adhesion Society
Japan) 9, 1 (1973)
11. T.lgarashi, J.Polym.Sci.Polym.Phys.Ed. 16, 407 (1978)
12. T.lgarashi, in "Adhesion and Adsorption of Polymers",
L.-H.Lee, Editor, p.42l, Plenum Press, New York, 1979.
THE ESTIMATION OF ADHESION IN FILLED POLYMER SYSTEMS

Yu. S. Lipatov, T. T. Todosiychuk, P. K. Tsarev and

L. M. Sergeeva

Institute of Macromolecular Chemistry

The Ukrainian SSR Academy of Sciences
252160 Kiev, USSR

Some new methods of estimating the adhesion

strength between filler particles and polymer matrix
are proposed. All the methods are based on the
determination of stresses at which the continuity of
samples is disturbed.

433
434 Yu.S. LlPATOV ET AL.

INTRODUCTION

The problem of adhesion strength between the filler part-

icles and the polymer matrix is one of the most important con-
siderations for development of composite materials. The thermo-
dynamic approach to it~_~olution might be more rigorous but its
application is limited • Though the thermodynamic estimation
describes the adhesion characteristics rather well, it does not
provide the real value of adhesion strength which depends on
numerous factors, in addition to thermodynamic considerations.
Adhesion strength is not an equilibrium characteristic but a
kinetic one depending, e.g., on the rate of peeling. When
adhesive joints are destroyed, their strength is usually deter-
mined under nonequilibrium conditions causing the discrepancy
between the thermodynamic work of ~dgesion and the adhesion
strength determined experimentally , • Nevertheless, adhesion
strength between filler particles and matrix is a very important
characteristic for filled polymers and its determination is of
great importance. There are available some results relative to
quantitative measurements of6a1hesion characteristic in composite
materials during deformation' • For coatings and films, a method
of adhesi~ngestimation under quasi-equilibrium conditions was
developed ' • However, this method cannot be applied to composite
materials containing fillers.

The current concepts regarding mechanism of polymer re- ,

inforcement allow us to assume that the usual viscoelastic
deformation of a polymer matrix, accompanied by polymer separa-
tion from the filler surface, takes place during deformation of
the composite. The failure of adhesion bonds in these cases may
have a character of interfacial or cohesive break in a weak layer
near the interface. Practically, both types of adhesion failures
have been observed. However, it is very difficult to determine
precisely the type of adhesioYoEl~lure, especially in polymers
filled with dispersed fillers • But the adhesion strength in
such systems may be experimentally determined without knowing the
exact mechanism of failure.

We have made some attempts to perform a quantitative

estimation of adhesion strength in filled polymers using both
model and industrial samples. Our methods of adhesion strength
determination are based on the following hypothesis: if a sample
with high filler content undergoes stretching from zero stress to
destruction of materials then such stress where the continuity of
sample structure is disturbed co~responds to a quantitative
measure of adhesion strength 13 ,14. Thus, our task is to find the
stress at which continuity disturbance ensues while undergoing
deformation. This stress is supposed to correspond to break of
the majority of adhesion bonds in the system.
ESTIMATION OF ADHESION IN FILLED POLYMER SYSTEMS 435

The value of adhesion strength (A) may be calculated if the

stress-strain dependence is determined cr - f(€) and the stress
value corresponding to transition from the linear dependence to
curvilinear one is known:

where the volume fraction of filler (~v) is expressed as

h = Pp
-PP-L.,-+-:::Pf-('r."l--'~w')'/-1Pw-
Here ~w is the weight fraction of filler (f), p are densities of
filler (f) and polymer (p).

The various methods discussed below allowed us to determine

the starting point of continuity disturbance in filled systems
during their deformation.

1. Compensation Method

The system consists of calibrated ,by tension ,elastic element

joined with the composite sample inves tiga ted ( Fig. 1,2). This
system is elongated at constant rate. The point of force applica-

3
)' .' .. ..
: . . ..
'

Figure 1. The schematic of the device for the determination of

the stress causing the local break of composite material contin-
uity. 1 - elastic element, 2 - mechanism recording the deforma-
tion of elastic material, 3 - experimental sample, 4 - the
mechanism for recording the total deformation of the system,
5 - the stretching device.
436 Yu.S. lIPATOV ET AL.

tion 4 records the total deformation whereas point 2 records the

dIfference curve, which allows one to determine the deformation of
each element.

If the modulus is not changed during deformation, both points

2 and 4 record straight lines ( OB and OA, Figure 2). At a definite
tension the break of adhesion bonds proceeds resulting in modulus
change. Here the curve recorded by point 2 deviates from linearity
(AC,Figure 2). From this curve the tension corresponding to adhesion
bonds breakage may be calculated
The corresponding stress at the deviation point character-
izes the strength of adhesion between the dispersed filler and
the polymer binder.

The samples of rubber (trademark SKMS-30) with various

carbon black (PM-lOO) contents were investigated by this method.
The corresponding stress-strain curves are given in Figure 3 and
calculated values of adhesion strength are summarized in Table I.

o E

Figure 2. The change of deformation with stress or time for

elastic element-composite material element (ODB) and for elastic
element alone (OAC). AE-deformation of elastic element,
AD-deformation of composite material.
ESTIMATION OF ADHESION IN FILLED POLYMER SYSTEMS 437

Table I. Adhesion Strength in the Systems with Various Amounts of

Carbon Black in Rubber.

Sample No. Filler Amount Adhesion

wt. %
10 5 N/m 2

1 5 6.38
2 20 4.37
3 30 4.10
4 35 3.83
5 40 3.72

Table I shows that measured adhesion falls off with higher fil-
ler loading, and this phenomenon may be explained by flocculation of
filler at high concentration. It would seem then that low filler
loading would give the best indication of true adhesion.

0 , 08 0 , 16 0 , 24 0 , 32 0 , 40

Figure 3. The stress-strain curves for samples presented in Table

I.

It is obvious from the results in Table I that the increase

of filler amount causes the decrease of adhesion strength by a
factor of 1.5. This fact may be explained in terms of the floc-
438 Yu.S. LlPATOV ET AL.

culation of carbon black particles with its increasing amount. As

a result the real surface of contact between the filler particles
and the polymer diminishes and adhesion strength drops.

2. Estimation of Adhesion from the Change in Volume

of the Sample while Undergoing Deformation

We have investigated the deformation of polybutadiene rubber

filled with aluminum powder. When the sample was stretched, the
failure of the adhesion contact between the polymer binder and
filler particles was noted. The microphotographs of filled sample
slit suffering continually growing deformation illustrate the
apparent increase in filler particle size due to formation of
some vacuoles around.

. .. '..' .... : ~ .. t. ._ 1
...
...
. .. . .. . .
• I
. .. .
II • ,.
," ..

~. ."
~ . . .... . - '-.."
. .. -'~
....",
. , 4

r
. _ 5
- -' . ..
Figure 4. Microphotograph of the slit of deformed sample at
various loadings (in arbitrary units): 1-0, 2-8, 3-12, 4-16,
5-20.

The initial sample 1, where the filler particles are closely

incorporated into the polymer,transforms after some deformation
in a state where some filler particles are separated from the
polymer. The number of particles separated is increased if the
stress is growing. So the filler becomes surrounded by voids
ESTIMATION OF ADHESION IN FILLED POLYMER SYSTEMS 439

stretched and directed along the stress direction. At high

stresses, the voids collapse and change into slots under the
action of lateral compression. These slots are noted as thin
lines on the photographs. The formation of voids results in an
increase in volume of the sample while it is stretching.

Let us define the initial volume as V = h 2 £ where hand f

o
are lateral and longitudinal dimensions of the sample. Then, the
volume of the deformed sample VI may be calculated by the
following expression:

VI = Vo + AV - (h - Ah)2(f + A£)

where AV, ~ and Af are the increases in volume and geometrical

dimensions of the sample. The relative change in the volume is
expressed by

1 + AV/V = (1 - Ah/h)2(1 - Af/f)

After taking logs:

in(l + AV/V) =2 fn(l - Ah/h) + fn(l + Af/f)

Here fn(l + AV/V), fn(l - Ah/h) and fn(l + Af/f) are the relative
deformation of the volume and sizes in logarithmic form. After
substituting the Poisson's ratio

v = fn(l - Ah/h)/fn(l + Af/f)

in this formula and having done some transformations, we obtain
the next expression:

AV/V = (1 + Al/l)1-2~ - 1

The change in sample volume while undergoing deformation may

now be calculated from the changes in samples size. The calcu-
lated and experimental curves of volume changes for the stretched
filled sample of po1ybutadiene rubber, depending on the value of
deformation, are given in Figure 5.

As is evident, the calculated and experimental curves of

volume changes differ when deformation of the samp1e·is increas-
ed. The deviation of experimental curve from the calculated one
is the result of additional increase in sample volume due to
appearance of discontinuity. At subsequent elongation both
experimental and calculated curves coincide. This is a result of
slot collapse.
440 Yu.S. LlPATOV ET AL.

!Jvjv
0, 08

0, 06
4

0, 04

0,02

2 4 6 8 £.
°
Figure 5. The dependence of volume change on deformation for
filled polybutadiene rubber; 1,2,3,4 - calculated and 5 -
experimental curves.

The experimental dependence t:;, VIv = f (t:;,f If) allows us to

determine the deformation and stress corresponding to the
critical point, where the experimental and calculated curves
deviate. This stress will characterize the adhesion strength
between the polymer matrix and filler particles. A special device
was constructed to determine the volume changes of the sample
during its deformation (Figure 6). It consists of a barrel filled
with inert liquid and is contained in a thermostat. The cover
fits snugly to the barrel and has a stretching device and cali-
brated capillary. When the sample is stretched, the liquid in the
capillary ascends and the volume increase is fixed. After calcu-
lating the stress and deformation corresponding to the sharp
increase in volume we can determine the adhesion strength from
curves of t:;, VIv = f (t:;,f If). For the polybutadiene rubber-carbon
black (45.5% by weight) its value is equal to 1.23 x 105 N/m2 and
for polybutadiene rubber-aluminum powder it is (43.2%) - 3.97 x
105 N/m L •

3. Acoustic Methods

These methods are principally based on the usage of ultra-

sound waves for de£eci~on of discontinuities in the deformed
filled compositionS, • The polymer volume changes because of
the growing porosity in the material due to formation of voids
between the polymer matrix and filler particles. As a result, the
volume concentration is changed after the deformation of the
filled system. The changing concentration of filler is accom-
ESTIMATION OF ADHESION IN FILLED POLYMER SYSTEMS 441

Figure 6. The photograph of device for measuring the sample

volume during its deformation.

panied by attenuation of ultrasound waves and is characterized by

two parameters: al - the parameter involving internal friction,
heat conductivity, elastic hysteresis, plastic flow and relaxa-
tion phenomena and a - the parameter relative to Rayleigh
scattering in the fiiled system. The total attenuation is
expressed as

where f is the frequency. The value a 2 is expressed as:

a 2 = (21T 2 /c 4 )V (llE/E)2

where C is the sound velocity, V is the volume of a scattering

particle, llE is the difference between the filler and matrix
modulus, E. The ratio llE/E is calculated by full solid angle
integration. This value and the volume of a particle allows us to
calculate the concentration of the filler in the system.

At composite sample deformation, as it was mentioned above,

volume fraction of filler changes, leading to changing acoustic
characteristics. The characteristics of transitional process are
the most reliable and simplest; they are obtained when the
442 Yu.S. LlPATOV ET AL.

investigated sample is periodically excited by acoustic vibra-

tions. In our experiments, the vibration periodicity was 50
periods/sec and the frequency 275 KHz. The picture of transition
process changes in dependence on the degree of sample deforma-
tion. The energy of free vibration may characterize the tran-
sition process. The value of free vibration amplitude (b) is
expressed, as a rule, by the exponential function b = exp(-pt),
where p is an indicator of attenuation. V - the rate of
ampli tude decrease is expressed as: \I = -p exp (-pt )=pb.

The energy of the signal is

002 0022
€ = / \I (t)dt =/ p b dt
being identical to kinetic energy. p may be calculated from
the value of amplitude of forced vibration (a):

a = 1 - b = 1 - exp(-pt)

Having carried out a series of expansion of exp (-pt) we obtain

Taking into account the first and the second terms (as the
process is rather quick) we obtain p = a/to This value is used in
the above mentioned equation, which yields:

00 2 2 2
=! (a b )/t dt, t = noLl.t
€
o
where n is the number o,f indications, Ll. t is the time between
measurements. The following equation may be used for graphical
integration:
2 b2
€ = ~ (}b2)/n0Ll.t)2 = 1/(Ll.t)2 a ¥
k k
k=l k=l n 2
For some filled systems we have determined the energy for various
tensions. As a result, the dependence of free vibration energy on
tensile stress was obtained (Figure 7).

It should be noted that the free vibration energy does not

change or changes slightly if the tension is low. At a certain
value of the stress, the energy decreases due to the decrease in
the filler volume concentration; and subsequently the separation
of the polymer matrix from filler particles proceeds leading to
ESTIMATION OF ADHESION IN FILLED POLYMER SYSTEMS 443

6'

Figure 7. The dependence of the energy of free vibrations on the

tension due to stretching.

void formation, which increases the free vibration energy. Having

reached the minimum, the filler concentration begins to grow due
to collapse of voids and the energy is diminished again.

The critical tension, corresponding to minimum of free vib-

ration energy, corresponds also to the beginning of destruction
of adhesion bonds, i.e. this value characterizes the adhesion
strength.

The deformation and tension corresponding to the beginning

of void formation are determined from data resulting from the
dependence of energy on stress.

Adhesion strength may be calculated from this tension. In

order to calculate the adhesion strength the stress value deter-
mined is related to the unit area occupied by the filler in I cm 2
of sample cross-section. Volume concentration is calculated and
the number of filler particles in a unit volume is also estimated as:

n
V
=q,V/ (4/3)'Ifr 3

where ~v is the volume concentration and 2r is the particle

diameter. For regular geometrical arrangement of particles in the rib
of cube of I cm3 :
n = n I/3
I
444 Yu.S. LlPATOV ET AL.

and an area of 1 cm 2 , we shall have

The total area of filler surface is

Sf =
1Tr 2n 2 = {3/4 1Tr 3)1/31Tr 2. ~ v 2/3 ....·.,v
2 O~ 2/3

Having found the tension at which discontinuity appears and

knowing the geometrical contact area between the filler and the
polymer we should be able to estimate the adhesion strength.

The experimental results given in Figures 2, 5 and 7 have

been used to calculate the adhesion strength in the system
polybutadiene rubber - A1 powder using the three values of
stresses found in three types of experiments and calculated con-
tact surface. The results are given in Table II.

Table II. Calculation of Adhesion Strength from Data obtained by

Various Methods.

Method Vol. Frac- Stress Contact Adhesion

No.* tion of 10 N Ar~a2 Stsengzh
Filler 10 m 10 N/m

1 0,432 1,23 0,34 3,60

2 0,432 1,30 0,35 3,72
3 0,432 0,76 0,19 3,98

* 1 - compensation method, 2 - volumetric method, 3 - acoustic

method.

Table II shows that there is a satisfactory agreement

between the adhesion strength (between the filler particles and
the polymer matrix) values obtained by different methods. Thus the
proposed methods might serve as the basis for discussion of practical
application to evaluate the effectiveness of filler treatment.
ESTIMATION OF ADHESION IN FILLED POLYMER SYSTEMS 445

4. Method of Transfer Functions

One of the non-destructive methods of estimating the adhesion

strength is based on the usage of transfer functions of ultrasound
vibrations (275-400 KHz) modulated by rectangular pulses (50 Hz)
and transmitting through the sample I7 ,18. The rectangular ultrasound
signal passing through a sample changes its parameters depending on
the medium characteristics.

The growth of amplitude in time may be represented by a

transitional function which may be found from approximation of
this curve and expressed as an exponential polynom. The tran-
sitional function is the ratio of two polynoms of P operator
which represents the transfer function. It provides the integral
characteristics of filled polymeric systems and consists of
transfer functions of some characteristics. The transfer function
of adhesion interaction in filled systems cannot be separated
from the total transfer function. For its performance an
auxiliary sample identical to the one under investigation is used
where the adhesion bonds are weaker due to filler surface treat-
ment by antiadhesive. Such sample is compared with the sample
under investigation and the transfer functions are obtained.

The transfer function of the investigated sample character-

izes the properties of the system and adhesion bonds present. To
obtain the transfer function which characterizes the adhesion
properties F(P) the transfer function of investigated sample
F"(P) should be divided by function of auxiliary sample F' (p).
The transfer function of adhesion properties then is

A pn + A p n- 1 +
F" (P) n n-l
F(P) = F' (P) or F (p)
B pm + B pm- 1 +
m m-l

where A ,B are coefficients of the polynom, and P is the operator.

n m
Having synthesized this function we can, using electro-
mechanical analogies, transform it into a mechanical model, and
calculate the elastic, viscous and inertial properties, character-
izing adhesion interaction in the filled system. Using this method,
t·he adhesion characteristics were determined for polyurethanes
filled with kaoline
446 Yu.S. LlPATOV ET AL.

Figure 8. The schematics for fixing the discontinuity in polymers

reinforced with carbon fibers during the deformation in a
magnetic field.

5. Estimation of Adhesion Strength in Systems

Reinforced with Fibers

In order to determine the adhesion strength in the polymeric

systems reinforced with fibrous carbon, the method of conductor
vibration may be used. A conductor with current is subjected to a
magnetic field to detect the discontinuities during deformation.

Here, the property of fibrous carbon to be a conductor of

electric current is used. The schematic set up is shown in Figure
8'. For experimental purposes, filaments of carbon fiber were
pressed into the epoxy resin. The filaments were disposed trans-
verse to the deformation direction.

The filament terminals from the sample were connected to an

alternating current source of frequency 20 KHz. The second
diagonal of the bridge was connected with the recorder after the
rectifier. The sample is put into a magnetic field (up to 8000 Oe
and stretched. During this process at a certain stress the
polymer matrix is separated from the filler surface. Under the
influence of the alternating current, the filament starts vibra-
ting in the magnetic field. The opposing electromotive force is
induced in the filaments. This force changes the balance of the
bridge arm, where the sample filaments are included in. As a
result of fixed discontinuity, the electric tension appears in
the second bridge diagonal. This stress is determined both by
the tension at elongation and adhesion bonds strength. This force
converted to unit area determines the adhesion strength.

The adhesion strength in the system mentioned was determined

using this method.
ESTIMATION OF ADHESION IN FILLED POLYMER SYSTEMS 447

Figure 9. The schematic of sample polymer reinforced with fibrous

carbon disposed at an angle of 45°.

The method of shear stresses is simpler for calculation of

carbon filament adhesion to polymer binder. If the fabric is
formed from carbon filaments, the plate with 2-3 layers of fabric
should be cut so that the sheet filaments are disposed at an angle
of 45° (see Figure 9). The prepared sample is stretched and
the force needed for break is measured. Figure 9 shows that the
sheet filaments release from the entanglements, being shifted
relative to each other along the binder. When the area of
destruction is known (it is equal to the sample square width) the
shearing force needed for adhesion failure can be calculated.

The shearing force is related to the breaking force by the

Poisson ratio:

E/2G = 1 + II
where E is modulus of elasticity and G is shear modulus. II is the
Poisson ratio equal to 0.34 in our experiments. So the adhesion
strength aad is expressed in terms of shearing force as

aad = 2.68a sear

h

Using this method the value of adhesion strength was deter-

mined for the composite made of epoxy resin reinforced by carbon
fibers. It w5-s fo~nd that the adhesion strength is equal to
42-37,52 10 N/m.

CONCLUSIONS

The proposed methods of adhesion strength determination in

polymers filled with disperse fillers provide rather accurate
448 Yu.S. LlPATOV ET AL.

idea of adhesion interaction and are widely used in our experi-

mental work. At the same time their usage is based on one
important parameter; i.e. the stress at the failure of continuity
of filled materials caused by polymer separation from the filler
surface is taken as a measure of adhesion strength. As a rule,
the character of adhesion failure (pure interfacial type, or
cohesive or mixed type) is unkn~~. According to the well-argued
viewpoint developed by Bikerman the break of adhesive joints is
seldom interfacial in character, rather as a rule it proceeds in
a weak layer near the interface. Nevertheless, this argument
cannot change the importance of the values determined by the
above mentioned methods. The correctness of these meth28s is
proved by the following theoretical arguments. Earlier a new
correlation between adhesion and cohesion strength was developed,
i.e.,
Wa d = Wc

where W d is the thermodynamic work of adhesion and W the

cohesio8 work for the component with lesser strength,Cthe binder
being such a component. The correlation proves that adhesion as a
thermodynamic value is determined only by the cohesive strength
of the binder when there are no specific interactions at the
interface. However, W represents the cohesive strength of the
binder in the bou21a2¥ layer. The numerous data obtained by us
and other authors ' testify to the considerable changes in
properties of polymer boundary layers in contact with a solid
surface. They are caused by adsorption interaction with the
surface of filler. These changes display the processes occurring
at the interface when adhesion 2~nd is formed. In particular,
some physical and technological factors cause the formation of
"weak" (using Bikerman's terminology) layers at the interface,
and this is where the adhesion bond failure takes place.

From what is mentioned above one can easily conclude that

the real conditions of adhesion failure are very complicated.
Even in the case when the adhesion bond break has a cohesive
character, the principal equation relating the works of adhesion
and cohesion is still valid. This gives a theoretical basis to
adhesion estimation based on the determination of the stress
corresponding to the appearance of the discontinuity in the
sample under investigation. Hence the combination of strict
theoretical approach to adhesion in combination with concepts of
physics and mechanics of polymer fracture makes it possible to
resolve one of the most important problems in the processing of
composite materials, i.e., the problem of estimating the adhesion
interaction without knowing the precise locus of adhesion
failure.
ESTIMATION OF ADHESION IN FILLED POLYMER SYSTEMS 449

REFERENCES

1. A. A. Berlin and V. E. Basin, "Principles of Polymer

Adhesion (Russ.)", Chimija, Moscow, 1974.
2. B. V. Derjagin, N. A. Krotova and V. P. Smilga, "Adhesion of
Solids (Russ.)", Nauka, Moscow, 1973.
3. S. S. Voyutsky and B. V. Derjagin, Kolloid Zh., ~, 626
(1965).
4. K. L. Mittal, Polymer. Eng. ScL, .!.Z., 467 (1977).
5. K. L. Mittal, in "Adhesion Science and Technology", 1. H.
Lee, Editor, Vol. A, pp. 129-168, Plenum Press, New York
(1975).
6. G. P. Sendeckyj, Editor, "Mechanics of Composite Materials",
Academic Press, New York (1974).
7. L. E. Nielsen, "Mechanical Properties of Polymers and
Composites", Marcel Dekker, Inc., New York (1974).
8. A. D. Zimon, "Adhesion of Films andCoatings (Russ.)",
Chimija, Moscow, 1977.
9. P. J. Zubov, et al., Lakokracochnye materialy i ikh
primenenie, No.3, p. 28 (1964).
10. K. L. Mittal, Pure Appl. Chern., ~, 1295 (1980).
11. W. L. Baun, in "Adhesion Aspects of Polymeric Coatings", K.
L. Mittal, Editor, pp. 131-146, Plenum Press, New York
(1983).
12. W. L. Baun, these proceedings, pp. 3-17.
13. Yu. S. Lipatov and L. M. Sergeeva, "Adsorption of Polymers",
John Wiley, New York (1974).
14. Yu. S. Lipatov and P. K. Zarev, Mekhanika polymerow, No.2,
p. 195 (1972).
15. P. K. Zarev and Yu.S. Lipatov, Vysokomolek. soed., A 12, 282
(1970).
16. P. K. Zarev and Yu. S. Lipatov, Vysokomolek. soed., A 17,
717 (1979).
17. P. K. Zarev and Yu. S. Lipatov, Vysokomolek. soed., A 21,
514 (1979).
18. A. M. Yaglom and 1. M. Yaglom, "The Probability and
Information (Russ.)", Nauka, Moscow, 1973.
19. J. J. Bikerman, Uspekhi chimii, 41 , 1431 (1972).
20. Yu. S. Lipatov and A. E. Feinerman, J. Adhesion, 6, 165
(1973). -
21. Yu. S. Lipatov, "Interphase Phenomena in Polymers (Russ.)",
Naukova Dumka, Kiev, 1980.
22. Yu. S. Lipatov, in "Adhesion and Adsorption of Polymers", L.
H. Lee, Editor, Vol. B, p. 601, Plenum Press, New York
(1980).
23. Yu. S.Lipatov, Pure Appl. Chem., 43, 273 (1975).
24. J. J. Bikerman, Ind •. Eng. Chern., 2, No.9, 40 (1967).
 Part IV
Durability or Stability Aspects
ADHESION AND DURABILITY OF METAL/POLYMER BONDS

J. D. Venables

Martin Marietta Laboratories

1450 South Rolling Road
Baltimore, Maryland 21227

This paper reviews the results of a com-

prehensive investigation made at the author's
laboratories to determine those factors responsible
for promoting the integrity and long-term durability
of metal/polymer bonds used in the fabrication of
aircraft and aerospace structures. Using a multi-
disciplinary approach and a variety of surface
analytical techniques, such as extended resolution
scanning electron microscopy (XSEM), X-ray photo-
electron spectroscopy (XPS), ellipsometry, and a
new technique we call surface behavior diagrams
(SBD), we have evolved several important concepts.
First, we have determined that the initial integrity
of metal/polymer bonds used for structural applica-
tions depends critically upon the morphology of the
surface oxide on the metal. In the case of the
metals studied, AI and Ti, we have observed that
certain etching or anodization pretreatment pro-
cesses produce oxide films on the metal surfaces
which, because of their porosity and microscopic
roughness, mechanically interlock with the polymer
forming much stronger bonds than if the surface
were smooth. Second, we have shown that the long-
term durability of metal/polymer bonds is determined
to a great extent by the environmental stability
(or lack ot stability) ot the same oxide which is
responsible for promoting good initial bond strength.
For AI, the effect of moisture intrusion at the
bond line is to cause the oxide to convert to an
nydroxide with an accompanying drastic change in

453
454 J.D. VENABLES

morphology and bond strength. For Ti, the evidence

suggests that the oxides formed on it are much more
stable than those on Al but under certain circum-
stances the oxide can undergo a polymorphic trans-
formation which may lead to bond degradation.
Third, a major finding of these investigations is
that significant improvements in durability of
adhesive bonds to Al can be achieved using an ex-
tremely simple treatment in which monolayer films
of certain organic acids are applied to the adherend
oxide to protect it against the effects of moisture.

INTRODUCTION

Although adhesive bonding has been widely used in the aero-

space industry for many years, the reduction in weight and cost
savings possible with properly designed adhesive joining systems
have recently added impetus to perform sophisticated research in
the field. More widespread use of bonding will necessarily in-
volve its use in primary structures, those responsible for keeping
the major structures intact. The problem of developing confidence
in the reproducibility of bonding operations and long-term reli-
ability of the bonds must be addressed before such use can occur.

Major efforts, such as the Air Force sponsored Primary

Adhesively Bonded Structure Technology (PABST) program,l have
focused attention on pretreatment of the metal parts to be joined
as the most important step in guaranteeing both initial bond
strength and long-term durability. Using new and highly sophisti-
cated surface analysis tools, workers at Martin Marietta Labora-
tories have been investigating what the pretreatments do to metal
substrates, particularly A1 and Ti, since they are the most widely
used metals in the industry. These efforts have lead to a new
understanding of how strong adhesive bonds are formed with metals,
what causes their degradation over time, and how their performance
can be improved by treatments which stabilize the oxides present
on the metal against the effects of moisture.

INITIAL BOND INTEGRITY

The initial integrity of metal/polymer bonds used for

structural applications is critically dependent upon the morphology
of the surface oxide on the metal. In the case of A1 and Ti, the
metals studied, we have observed that certain etching or anodiza-
tion pretreatment processes produce oxide films on the metal sur-
faces which are extremely rough and porous on a microscopic scale.
ADHESION AND DURABILITY OF METAL/POLYMER BONDS 455

For example using XSEM (SEM done on a STEM at ~ 30 A resolution) we

have observed 2 that the Forest Products Laboratories (FPL)3 pro-
cess and phosphoric acid anodization (PAA)4 process used for pre-
paring Al produce surface oxide structures of the type shown in
Figures 1 and 2. The microscopic interlocking roughness exhibited
by the structures is apparently a crucial factor determining ad-
hesion at the epoxy-oxide interface in bonded aircraft structures.
To demonstrate this, we have intentionally added 500 ppm fluorine
to the FPL etch bath and observed that the surface oxide morphology
is drastically modified, so that it is much less interlocking in
nature. This surface, which exhibits undulations, but does not
interlock with polymeric coatings, may be readily separated from
an overlying adhesive or primer coating to allow examination of
the polymer side of the interface. When this is done, it is
observed that the polymer retains a perfect replication of the
original oxide features. This situation contrasts in two respects
with the behavior exhibited when attempts are made to separate a
normally prepared FPL surface from an overlying adhesive. First,
separation at the oxide/polymer interface is much less easy to
achieve if the metal is properly prepared, but can be done if the
AI is first bent very sharply. Second, we have observed that the
separation of good bonds, when forced to occur at the oXide/poly-
mer interface, is accompanied by an extreme amount of deformation
of the polymer which is generally badly torn and ripped. One
consequence of this different mode of separation is that the bond
strength, as measured by a climbing drum peel (CDP) test, for
example, may be as much as a factor of three different, with the
interlocking surface yielding the highest strength levels.

Figure 1. Oxide morphology produced on Al surface by the FPL

process, as observed by stereo XSEM (Ref. 3).
456 J.D. VENABLES

" " - - AI

Figure 2. Oxide morphology produced on Al surface by the PAA

process (Ref. 4).

Another consequence of not having an interlocking morphology is

that the peel strength can be lowered further simply by placing
a drop of water in the crack developed during the test. We inter-
pret this to mean that in the absence of mechanical interlocking,
when the bond strength is determined principally by chemical
forces across the interface (e.g., van der Waals or dispersion
forces), the presence of water can disrupt these bonds readily
thereby reducinp the interfacial strength. A similar effect is
observed for mica5 which is bonded across the layer planes by
Van der Waals forces. When cleaved ih the presence of water,
the cleavage energy (¢ = 250 ergs/cm2 ) is lower than when done
in a dry environment (¢ = 1200 ergs/cm 2 ). For adhesive bonds,
however, water has no significant short-term effect when
interlocking is present. (The long-term effect of moisture on
properly prepared bonds will be discussed in the next section).

It might be argued that fluorine picked up from the con-

taminated bath and deposited on the oxide surface might be respon-
sible for the degradation in bond strength observed in the above
experiment. However, examination by Auger/XPS of surfaces pre-
pared in fluorine contaminated baths indicated that the surface
concentration of fluorine was very low (less than 3% surface
ADHESION AND DURABILITY OF METAL/POLYMER BONDS 457

coverage). Prior work,6 in which fluorine was added to rinse

water (in which case, much higher concentrations of F can be
adsorbed on the surface) has demonstrated that such low concen-
trations of F, per se, do not significantly degrade bond strength
so long as the oxide-morphology is not altered.

Although considerable success can be achieved in bonding

polymers to Al using either the PAA or FPL pretreatment process,
the situation is not as straightforward for Ti. Prior attempts
to develop a pretreatment for Ti that would be as successful as
those for Al have yielded a multitude of processes, but not until
recently have any shown promise. Because of the increasing
interest in using Ti for advanced aircraft structures we felt it
was important to know more about the types of surfaces generated
by these processes with the hope that this information would pro-
vide guidelines for future improvements. Accordingly, using the
techniques we employed for studying Al adherends, Ditchek et al. 7
characterized surfaces of Ti prepared according to a variety of
different pretreatment processes. The results were then compared
with mechanical properties measurements performed by other inves-
tigators 8 in a coordinated U.S. Navy (NAVAIR) program as discussed
below.

Table I lists the various surface pretreatments investi-

gated and provides some comments on the types of surfaces gener-
ated on Ti-6Al-4V alloys. The appearance of these surfaces varied
considerably, but it was possible to classify them into three
groups according to similarities in their surface morphologies.
Group I surfaces, which include those resulting from the PF and
MPF treatments (notation defined in Table I), display little
macro- or micro-roughness.* Group II surfaces, which derive from
the DA, LP, TU and DP treatments, all exhibit a large degree of
macro-roughness and a small degree of micro-roughness on the LP and
TU surfaces. Group III surfaces, which include those generated
by chromic acid anodization 9 (CAA) at 5 or 10 volts are character-
ized by having no macro-roughness, but a high degree of micro-
roughness associated with a porous oxide.

The Group III surface morphologies are of particular inter-

est because they bear a marked resemblance to those produced on Al
by the PAA or FPL process. For example, the CAA surface exhibits
a porous oxide with protruding whiskers similar to the FPL struc-
ture and is approximately the same thickness if the anodizing

* A macro-rough surface is defined as an uneven surface with

characteristic bumps or jagged features about 1.0 ~m or greater.
Micro-rough surfaces have fine structure with dimensions 0.1 ~m
or less.
 oj:>.
TABLE I. Morphological Chara~teristics and Chemical Contaminants Associated with Various UI
(Xl
Ti Pretreatment Processes

Process Oxide Groupa

Process Code Thickness Number Comments
(A)

1. Phosphate fluoride PF :wu ! ¥ contami.nation

2. Motlitied phosphate MPF 80 I F contamination
fluoride
3. Dapcotreat DA 60 II No apparent fine structure;
Cr on surf ace
4. Dry hone DP 100-200 II Deformed surface with
PASA JELL 107 embedded Al203; fluorine
contamination
5. Liquid hone LP 200 II Embedded alumina; fluorine
PASA JELL 107 contaminant; Cr on surface
6. Turco 5578 TU 175 II Fe containing particles on
surface
I. Ghromic acid anodize CAA+F SV:400 III Porous oxide with protrud-
10V:800 ing whiskers; fluorine con-
tamination
~. Alkaline peroxide AP 450-1350 III Porous oxide
depending c....
on chemis try, 0
temperature, <
time m
z
»to
r
a For a definition of group numbers, see text. m
en
 ADHESION AND DURABILITY OF METAL/POLYMER BONDS 459

potential is 5 volts and the anodization time is 20 minutes. When

the anodizing potential is raised to 10 volts (and the time remains
the same), the surface morphology becomes somewhat intermediate
between that of FPL and PAA oxides in both appearance and thick-
ness. Because of these similarities, it would therefore be
expected that the CAA oxides would interlock with polymer coatings
providing interfacial bond strengths comparable to those associated
with Al prepared by the FPL and PAA processes.

Evidence that this is the case is provided by wedge tests

performed by Brown 8 whose results are shown in Figure 3. For these
tests the Ti surfaces were prepared in exactly the same manner as
those samples examined by XSEM. A standard wedge test configura-
tion was employed using the BR127/FM300 primer/adhesive system to
bond the Ti test strips together. The test conditions are indi-
cated in Figure 3.

The test results clearly demonstrate a correlation between

surface roughness and bond strength. Thus, the Group I surfaces
which exhibit no macro- or micro-roughness, exhibit very poor
behavior in the wedge test with all of them failing adhesively at
the primer/metal oxide interface. On the other hand, the Group III

2.0 Wedge Tests

140°F - 100% R. H. ~ _ _ _ _ _ _ _ _ MPF
1.8
I

c:
_ _- - - - - - - - PF

TIME (hr)
Figure 3. Wedge test crack length as a function of exposure time
@ 100% R.H., 60°C for Ti-6A1-4V adherends prepared by
eight different processes described in Table I. Defi-
nitions of Groups I, II and III can be found in the text.
460 J.D. VENABLES

surfaces, which exhibit no macro-roughness but a high degree of

micro-roughness in the form of porous oxides exhibited almost no
crack growth except for a small amount of cohesive failure in the
adhesive during early stages of the test. The wedge test values
corresponding to the Group II category lie intermediate between
those of Group I and III consistent with the fact that Group II
oxides exhibit more macroroughness than Group I but less micro-
roughness than Group III. Evidently, the presence of a microrough
porous oxide which can interlock with the polymer is just as im-
portant a factor determining the strength of polymer bonds to Ti
as is the case for AI.

LONG-TERM DURABILITY

The long-term durability of metal/polymer bonds is deter-

mined to a great extent by the environmental stability (or lack of
stability) of the same oxide that is responsible for promoting good
initial bond strength. For AI, moisture intrusion at the bond line
causes the oxide to convert to an hydroxide with an accompanying
drastic change in morphology,IO Figure 4. The resulting hydrated
material adheres poorly to the Al beneath it and, therefore, once
it forms, the overall bond strength may be severely degraded. The
proposed failure model is shown schematically in Figure 5.

Figure 4. Morphology of aluminum hydroxide (pseudo-boehmite) pro-

duced on A1 surface during exposure to moisture. The
change in morphology which accompanies the conversion
of the original oxide to this hydrated form results in
a drastic reduction in adhesive bond strength (Ref. 10).
ADHESION AND DURABILITY OF MET ALIPOL YMER BONDS 461

Crack extension
• Original FPL Oxide

formed afterhydroxide_~~~~~~~~~~~~~~~~~~~~~~...--
Aluminum crack
propagation

Figure 5. Schematic drawing of the mechanism deduced for crack

propagation during wedge testing. In the humid en-
vironment, the original FPL oxide is converted to a
hydroxide which adheres poorly to the aluminum sub-
strate. The crack propagation rate is much faster
here than in a dry atmosphere, when the crack propa-
gates directly through the adhesive (Ref. 10).

For Ti, the evidence suggests that its oxides are much more
stable than those of AI. However, preliminary evidence l l suggests
that under some circumstances the oxide, which is originally amor-
phous, undergoes a polymorphic transformation to anatase which,
because of volume changes and accompanying morphology changes, may
lead to bond degradation just as the oxide-to-hydroxide conversion
process does for bonds to AI. This transformation, which occurs at
temperatures above 200°C in a dry environment and above 85°C in the
presence of moisture, is being studied further because it could be
an important degradation mechanism in future applications, for
example, in attempts to increase service temperatures through the
use of polyimide adhesives.

The proposed degradation model for AI/polymer bonds, Figure

5, is supported by evidence showing a correlation between incuba-
tion times for the oxide-to-hydroxide conversion process and wedge
test results; the longer the incubation time of oxides prepared
(and treated) in various ways, the better the wedge test results.
This appears to be the first time that such a direct correlation
has been observed between a bond durability test and a measurable
physical parameter of the metal adherend.lO

A new technique, the surface behavior diagram, was devel-

oped by Davis et al. 12 during this work. This type of diagram is
462 J.D. VENABLES

analogous to phase diagrams for bulk phases but is intended to

represent effects that are specific for surfaces, e.g •• reactions
between a surface, an adsorbate, and the environment. In the
present case, the technique was used to show that the greater
stability of PAA surface oxides, relative to FPL oxides, is due
to the presence of adsorbed phosphate (from the electrolyte) which
inhibits the oxide/hydroxide conversion process until the phosphate
itself becomes extensively hydrated and is lost by dissolution.
Thus, as shown in Figure 6, XPS data were used to determine the
composition of the PAA surface, and therefore its position on the
Al203-AlP04-H20 behavior diagram as a function of exposure time to
hot water. The fact that the surface reaction follows the path
shown in the figure, rather than the alternative path of proceeding
first to Al203 and then to the hydroxides, suggests that the effec-
tiveness of the inhibitor is eventually lost not because it goes
directly into solution but because it first hydrates and then de-
sorbs. Evidently, choosing an inhibitor that is more resistant
to hydration and subsequent desorption is an approach that would
provide greater stabilization of the surface against the effects
of moisture. This example suggests that SBD's may be generally
useful for surface science studies particularly in the fields of
adhesive bond durability, metal corrosion and corrosion inhibition.

""'
"' "
""
""
""
""
""
"
" "'
"'0
""

Figure 6. The ternary behavior diagram of fresh and hydrated PAA

Al oxide surfaces. The cluster of unnumbered points
are data taken on freshly prepared surfaces. The
numbers by some points denote the exposure time in
hours to 100% relative humidity at 50°C (solid points)
or at 60°C (crosses) (Ref. 12).
ADHESION AND DURABILITY OF METAL/POLYMER BONDS 463

INHIBITORS

A major finding of these investigations is that significant

improvements in durability of adhesive bonds to Al can be achieved
using an extremely simple treatment whereby monolayer films of
certain organic acid molecules are used to protect adherend oxides
from the effects of moisture. 13 Specifically, we have shown that
an adsorbed monolayer of an amino phosphonic acid, nitrilotris
methylene phosphonic acid (NTMP), can improve the stability of
FPL-treated Al so that its performance in wedge tests is almost
comparable to Al treated by the PAA process (Figure 7). Moreover,

c:
z
0
Vl
Z
LW
I-
c.s 2
:.<:
u
<I:
ac
u

TIME (hr)

Figure 7. Wedge test crack length as a function of exposure time

in a humidity chamber @ 100% R.H., 60°C. The data
demonstrates that the performance of FPL-treated ad-
herends under these test conditions is improved sig-
nificantly when the adherend is treated before bonding
in an inhibitor solution containing just 10 ppm of
nitrilotris (methylene) phosphonic acid (NTMP). In
fact, after the treatment, the performance approaches
that of PAA-treated adherends. (Ref. 13).
464 J.D. VENABLES

those oxides formed by the PAA process, which are normally stabi-
lized to some extent against moisture by adsorbed phosphate, can
be stabilized even more effectively by the amino phosphonic acid
treatment. 14 When this is done, the wedge test performance of
PAA adherends is improved by almost a factor of two under the
conditions of the tests employed (Figure 8). The attractiveness
of the inhibitor treatment is enhanced by the fact that since the
inhibitor is used in monolayer form, very large quantities of Al
can be treated with only small quantities of inhibitor solution.
We also note that another potential approach for improving dur-
ability is suggested by the following observation we have made.
In an attempt to perform some specialized hydration experiments on
bulk aluminum oxide, we were surprised to find that crystalline
A1203, whether in the form of powders or single crystals, is
very much more stable than FPL or PAA oxides, which we determined,
by our electron diffraction studies, to be amorphous. In fact, we
observed that whereas the amorphous oxides hydrate in a matter of
minutes or hours when immersed in 80°C water, no evidence of hy-
dration could be found by XSEM on crystalline material that had
been exposed for over a week in water heated to temperatures up
to the boiling point. Evidently, the degree of crystallinity
of Al203 has a profound influence on its stability against the
effects-of moisture, which clearly suggests another avenue of
investigation for improving the performance of metal/polymer
bonded structures.

2.25

~ 2.00
<.!)
Z
LU
-'

""
u
~ 1.75
u

l.50

10

TIME (hr)

Figure 8. Wedge test crack length as a function of exposure

time in a humidity chamber @ 100% R.H., 60°C showing
the improvement in performance when PAA adherends
are treated with NTMP inhibitor. (Ref. 14).
ADHESION AND DURABILITY OF MET ALIPOL YMER BONDS 465

DISCUSSION AND CONCLUSIONS

In summary, our work indicates that the initial integrity of

polymeric bonds to metals is significantly enhanced if the adher-
ends are prepared in such a manner that the surface oxide is rough
and porous on a microscopic scale. For AI, the PAA and FPL pre-
treatment processes provide porous oxide structures which interlock
with the overlying polymer to form strong adhesive bonds. For Ti,
the CAA process provides similar oxide morphologies and good bond
strengths. In humid environments, we find that the long-term
durability of metal/polymer bonds is determined to a great extent
by the stability (or lack of stability) of the same oxide which
is responsible for promoting good initial bond strength. For AI,
the degradation mechanism involves an oxide-to-hydroxide conversion
process which leads to drastic surface morphology changes and an
accompanying loss of bond strength. For Ti, the evidence suggests
that the oxides formed on it are much more stable than those on Al
but that morphological changes do occur albeit at more elevated
temperature. Since the environmental conditions may be more severe
for applications which require the use of Ti, however, durability
problems could be just as troublesome as they are for Al although
much more work is required to define those circumstances which
would threaten bond integrity.

Considerable improvement in the durability of bonds to Al

can be achieved using a simple treatment procedure in which mono-
layer films of certain organic acids are applied to the adherend
oxide to protect it against the effects of moisture. These exper-
iments demonstrate that organic compounds can be just as effective,
and potentially perhaps much more effective, than inorganic inhi-
bitors (e.g., chromates) which are currently used as additions to
corrosion inhibiting primers. We are currently using surface
behavior diagrams, a new technique developed during this work for
examining the interaction between surfaces, adsorbates and envi-
ronment, to study organic inhibitors further.

Although the emphasis of this work has been on metal/polymer

bonds, the results appear to have implications that go beyond this
technology area. Specifically, the work on durability has clear
relevance to the general field of corrosion and corrosion inhibi-
tion. For instance the reaction between water and a surface oxide
to form a hydroxide cannot be considered strictly a corrosion pro-
cess because the metal itself is not attacked in the initial stages
of the reaction. Nevertheless, we suggest that this reaction is
an important precursor step that leads to eventual corrosion of the
metal. In support of this we have observed that when an Al sample
is exposed to water, the oxide film initially passivates and pro-
tects the metal from attack during an "incubation" period. Follow-
ing this period, intense hydration activity occurs, during which
466 J.D. VENABLES

the original oxide is converted to a hydroxide. We observe evi-

dence of true corrosion only after hydroxylation when the evolution
of gas suggests that the Al metal itself is reacting with the water
and forming hydroxide and hydrogen gas. Evidently, the protection
provided by the oxide layer is disrupted once the oxide hydrates,
which suggests that procedures designed to reduce the hydroxylation
rate would also be effective in corrosion protection of Al at least
in certain pH ranges. If so, we suggest that the concepts and
techniques developed during these investigations could be of con-
siderable benefit if properly applied to the general field of metal
protection against environmentally induced degradation.

ACKNOWLEDGEMENTS

We wish to thank the following agencies for supporting the

indicated aspects of this work: The surface behavior diagram work
was performed under contract F49520-78-C-0097 to AFOSR; the Ti work
under contract N00019-80-C-0508 to NASC; the inhibitor work under
contract N00014-80-C07l8 to ONR/AROD.

REFERENCES

1. E.W. Thrall. in "Proc. 10th National SAMPE Tech. Symp.,"

Kiamesha Lake, NY, Oct. 17-19, p. 73 (1978).
2. J.D. Venables, D.K. McNamara, J.M. Chen, T.S. Sun, and R.L.
Hopping, Appl. Surface Science 3, 88 (1979).
3. H.W. Eichner and W.E. Schowalter, Forest Products Laboratory
Report No. 1813 (1950).
4. G.S. Kabayaski and D.J. Donnelly, Boeing Co. Report No. DG-
41517 (February 1974).
5. B.V. Deryagin and M.S. Mesik, Soviet Phys.-Solid State (Eng.
Transl.) l, 1393 (1960).
6. J.M. Chen, T.S. Sun, J.D. Venables and R.L. Hopping, in "Proc.
22nd Nat. SAMPE Symp.", p. 25 (1977).
7. B.M. Ditchek, K.R. Breen, T.S. Sun, J.D. Venables and S.R.
Brown, in "Proc. 12th Nat. SAMPE Tech. Conf.", Oct. 7-9,
1980, Seattle, WA, Vol. 12, pp. 882-885 (1980).
8. S.R. Brown, in "Proc. 27th Natl. SAMPE Symp.," San Diego,
CA, May 4-6, p. 363 (1982).
9. Y. Moji and J.A. Marceau, "Method of Anodizing Titanium to
Promote Adhesion," U.S. Patent No. 3,959,091, May 25, 1976.
10. J.D. Venables, D.K. McNamara, J.M. Chen, B.M. Ditchek, T.I.
Morgenthaler, T.S. Sun, and R.L. Hopping, in "Proc. 12th Nat.
SAMPE Tech. Conf." Oct. 7-9, 1980, Seattle, WA, Vol. 12,
pp. 909-923 (1980).
ADHESION AND DURABILITY OF METAL/POLYMER BONDS 467

11. M. Natan and J.D. Venables, paper presented at the 2nd

International SAMPE Conf., Stresa, Italy (1982).
12. G.D. Davis, T.S. Sun, J.S. Ahearn and J.D. Venables, J.
Mat. Sci. 17,1807 (1982).
13. J.S. Ahearn, G.D. Davis, T.S. Sun and J.D. Venables, in
"Adhesion Aspects of Polymeric Coatings," K.L. Mittal,
Editor, pp. 281-299, Plenum Press, New York (1983).
14. D.A. Hardwick, J.S. Ahearn and J.D. Venables, unpublished
work.
FAILURE MECHANISMS IN THE BOUNDARY LAYER ZONE OF METAL/POLYMER

SYSTEMS

W. Brockmann, 0.-0. Hennemann and H. Kollek

Fraunhofer-Institut fur angewandte Materialforschung

0-2820 Bremen 77, W.-Germany

A systematic optimization of the long term

stability of polymer/metal systems in the boundary
zone requires the exact knowledge of the mechanisms
producing the so-called adhesional failure particu-
larly under the influence of humidity. Strong and
water-stable adhesion, for example, between organic
primers and aluminium alloys is obtainable if the
metal surface is coated with a porous oxide layer,
which can be produced by special etching and anodi-
zing processes. The relatively small molecules of the
primers can penetrate practically the whole oxide
structure. But despite this, it is easy to demonstrate
that very small differences in the parameters of the
surface treatment, leading to small differences in
the surface morphology, can change the water stability
of the boundary zone considerably. The transition zone
between the oxides and primers in such systems was
investigated by using the wet peel test, transmission-
electron microscopy of ultra-thin cross sections from
the boundary zone before and after delamination and
ESCA on the fracture surfaces. It was found that in
bonds showing adhesional failure in the wet peel test
very near to the oxide layer there is a zone whose
properties appear to be different from that of the
bulk polymer. After an adhesional failure of these
bonds, in many cases, relatively high amounts of
polymer remain on top of the oxides, and on the
polymer surface no metal oxides are detectable, if
an anodized oxide is present. These observations
lead to the conclusion that at least partly an

469
470 w. BROCKMANN ET AL.
adhesional failure in reality is a cohesional
failure in weak boundaries of the polymer, whose
properties are influenced by the state of the
oxide surface.

INTRODUCTION

One of the problems in metal-polymer systems like adhesive-

ly-bonded metal jOints or metals coated with organic coatings is
the stability of the adhesion in humid environments. In many cases
bonded joints showing high initial adhesional strength after ex-
posure to humid environments fail in the form of so-called "adhesional
failures". At first glance it appears as a clear separation be-
tween the metal surface and the polymer. The sensitivity of metal/
polymer systems against humidity penetrating the polymer and
weakening or destroying adhesion can be influenced by special sur-
face treatments of the metals prior to the resin application.

Particularly the aircraft industry uses for the different

aluminium alloys some empirically developed chemical surface treat-
ments to obtain water-stable adhesion in adhesively bonded joints
and coated surfaces. Common are etching processes in chromic sul-
phuric acid solutions at elevated temperatures with different
etching times (in Europe 30 min (CSA) i in USA 10 - 15 min (FPL)).
In some production lines of the European aircraft industry the
CSA etching followed by an anodization process in chromic acid
(CAA) as a surface treatment has been used for nearly 30 years.
In the United States 10 years ago an anodizing process in phospho-
ric acid (PAA) was developed and is widely used today.1,2

These different surface treatmens produce on the aluminium

surfaces oxide layers of very different morphologies as can be
seen in the transmission electron micrographs in Figure 1. The TEM-
micrographs were taken from ultra-thin cross sections of the tran-
sition zone between metal and polymer in bonded aluminium joints.

With all these surface treatments it is possible to pro-

duce highly water-stable adhesion between the polymer and the metal.
But experience shows that in some cases bonded joints of aluminium
alloys in aircraft delaminate as adhesional failures after ser-
vice times of 4 or 5 years if the surface pretreatment was only an
etching process like FPL or CSA. This is not observed with CAA- or
PAA-treated metals bonded by adhesives. Both anodizing processes
are practically equal in achieving bonded joints of high long term
stability under humid environmental conditions. 3

The good adhesional properties of the oxide layers produced

by the anodizing processes are attributed to their high porosity.4
FAILURE MECHANISMS IN THE BOUNDARY LAYER ZONE 471

Figure 1. TEM of cross sections from the transition zone between

metal and polymer in bonded aluminium joints after different
surface treatments.

Compared with oxides produced by CSA or FPL they have greater sta-
bility against hydration,2 and high chemical reactivity.5 Espe-
cially the hypothesis of a micromechanical interlocking between
polymer and oxides as an important mechanism, besides chemical
reactions, has attracted some attention.

THE RESIN-OXIDE TRANSITION ZONE

A micromechanical interlocking can only become effective if

the molecules of the primer or the adhesive during and after appli-
cation and before curing penetrate the oxide layers produced by
the surface treatment. Even after drying of the treated surfaces a
472 W. BROCKMANN ET AL.

relatively large amount of water is present in the oxide layer. 6

So, at first glance, it seems not very probable that organic mole-
cules with molecular weight up to more than a thousand can com-
pletely penetrate the aluminium oxide layer.

Nevertheless such a total penetration seems to be possible

from the two transmission electron micrographs, shown in Figure 2.

Figure 2. TEM of cross sections from PAA-oxides prepared from a

metal surface without primer, right; and with cured primer
(EC 3924, 3M Co. ), left.

If an ultra-thin cross section is taken from close to the

surface of a PAA-treated aluminium, deformations in the oxides
structures can occur as can be seen in the right micrograph. If
the oxide layer is coated by a primer before the cutting process
and this primer is cured as usual, a deformation of the oxide by
the cutting process is not observed, as can be seen in the left
TEM-picture. From the differences in the pictures it may be con-
cluded that the primer strengthens the oxide against deformation
in this preparation process. The same strengthening effect by a
primer is also observed on surfaces pretreated by etching or by
the chromic-acid-anodizing (CAA)-process.

The conclusion that large portions of the oxides on the

aluminium are penetrated by molecules of the primer is confirmed
by results of the electron energy loss spectroscopy (EELS) on
ultra-thin cross sections. One result of such investigations is
shown in Figure 3.
FAILURE MECHANISMS IN THE BOUNDARY LAYER ZONE 473

Figure 3. Polymer identification in CAA-oxide by EELS by analysis

of carbon.

The points of identification in the oxide structure are

marked and are shown in the TEM Figure 3. Today only those analy-
tical techniques with lateral resolution in the range of less than
100 ~ allow identification of organic material in different ranges
of the oxide structure. If the cross section is prepared on an ano-
dized metal without primer coating, carbon cannot be detected by
EELS.

Assuming the presence of a partly or completely cured resin

in the pores of the oxide layer, it is hard to understand how in-
vading water should cause an adhesional failure in this zone.
Especially considering the rapid failures induced by water this
is not easy to visualize. Such fast aging effects are known from
the wedge-test and the wet peel test. 2 ,7 In the last test within
seconds a cohesional crack can be turned into adhesional failure
only by adding water.

For the examination of strength and durability of metal

bonds, the wedge-test has been used as a short-time test method
for some years. 2 Today this test method is accompanied in some
cases by the wet peel test. In this test using a normal roller-
peel-specimen the pulling-off procedure is started in a normal
climate of 20°C and 60 % rel. humidity. After peeling-off half
the length of the bond line, the peel-process is stopped and in
474 w. BROCKMANN ET AL.
the tip of the cohesive crack in the bond-line a drop of water
(20 °C) containing 0.5 % detergent, for complete wetting, is
applied. 10 sec later the peeling process is continued. If the
adhesion between primers or adhesives and oxides on the metal
surface is sensitive against humidity, the peel-strength de-
creases after the water application and the fracture surfaces
show totally or partly adhesional failure, as can be seen in
Figure 4 on peeled specimens, whose metal parts were anodized
under different conditions before primer application.

Figure 4. Fracture surfaces of peeled specimens, aluminium CAA-

anodized with different anode-cathode distances.
FAILURE MECHANISMS IN THE BOUNDARY LAYER ZONE 475

The peel strength measured on these specimens in dry and

wet states is plotted in Figure 5.

1111. III

N/mm
nOry

~Wet
5.111 Peel-Strength
CAAl211l24 cl.

111.111
~ ott ott
~
Cb ~ {> ~
Figure 5. Peel strength of metal bonds after different CAA-
anodizing processes.

In the case of sufficient adhesion stability, the peel

strength remains practically constant in the wet state and no
adhesional failure is detected on the fracture surfaces. This
test has proved to be more sensitive than the wedge-test in our
and other investigations 7 and seems to be one of the best short-
time tests to determine the long term stability of metal-polymer
bonds against humidity.

The roller-peel specimens of the tests mentioned here were

produced using the Alloy 2024 T3 clad, corrosion inhibiting air-
craft-primer EC 3924 containing different polymer components like
epoxy-resins, epoxy-novolak-resins, special curing agents and
strontium-chromate as pigment, and the nitrile modified epoxide-
adhesive AF 126. Prior to bonding, the metal surfaces were de-
greased, etched in chromic sulphuric acid and anodized in chromic
acid at 40°C with a stepwise increasing of voltage up to 50 Volts. 3
For the anodizing process, different anode-cathode distances were
used to reach different current densities, which are shown in
Table 1. The results show that only under relatively low current
densities adhesion of sufficient stability is obtainable.
476 w. BROCKMANN ET AL.
Table I. Dependence of Current Density on Anode-Cathode Distance
during Chromic Acid Anodizing.
Anode-cathode distance Current density
cm A/dm2

8 2
16 1,6
25 1,2
40 0,4

FAILURE MECHANISMS

Because scanning electron micrographs of the surfaces after

different anodizing processes (Table 1) showed practically no
differences in the oxide topography, ultra-thin cross-sections
from the boundary zone of undestroyed bonded joints were prepared
and examined in a transmission electron microscope. Some results
of these investigations are shown in Figure 6.

Recognizable in the micrographs are, at first, differences

in the oxide morphology depending on the current density during
the anodizing process. Second and probably more important is the
presence of bright zones in the primer area near the oxide in the
case of oxides produced by anodizing with 2 A/dm2 and 1.6 A/dm 2 •

Only after anodizing with 0.4 A/dm2 such a bright zone is

not recognizable. A comparison between these micrographs and the
fracture surfaces in Figure 4 lead to the conclusion that a cor-
relation between the appearance of such bright zones and water-
unstable adhesion may exist. A first assumption that these bright
zones in the polymers are due to insufficient wetting during the
primer application does not seem very probable, because if this
was the case then adhesional failure must occur in the peel-test
in.the dry state, too, and not only in the wet state. But the ini-
tial peel-strength of all the bonds is practically the same, and
in the dry state cohesional failures occured in all cases.

A second assumption is that the bright areas in the micro-

tamed samples are due to fractures at the polymer/oxide interface
or extremely thin spots both, for example, caused as artifacts by
the cutting process which can only be carried out in contact with
water. If this is true, we have to conclude that the resistance
of the polymer near the oxides against mechanical deformation in
contact with water must depend on the state of the oxides on which
it was cured because using the same cutting parameters artifacts
occur on some oxides and not on others.
FAILURE MECHANISMS IN THE BOUNDARY LAYER ZONE 477

2 A/dm 2 \,6 A/dm 2

0,4 A/dm 2

Figure 6. TE-micrographs of cross sections from oxide/primer

transition zones. CAA-oxide produced with different current
densities; primer EC 3924.

At last it may be that the bright zones are caused by ma-

terial of much lower molecular weight than the surrounding mate-
rials, oxides and cured polymer. Both, differences in resistance
against the cutting process or differences in molecular weight of
the polymer zone near the oxides may be induced by selective
chemical reactions of the organic molecules with the oxide or
selective adsorption mechanisms of reactive portions of the
primers invading the oxides, whereas portions of lower reactivity
or larger size remain at the surface and cannot crosslink as tight
as the other polymer in the curing process. 8

The question now is whether the so-called "adhesional

478 w. BROCKMANN ET AL.
failures" in the wet peel tests could be correlated directly to
the appearance of the bright zones.

To demonstrate this, ultra-thin cross sections from the

boundary area of a metal/polymer bond after an "adhesional failure"
induced by wet peeling was prepared. The TEM of this cross section
are shown in Figure 7 .

.........

Figure 7. TEM of cross sections from the polymer and the metal
sides of an adhesively bonded aluminium jOint after "adhesional
failure". Surface tri:!atment CAA. Upper Figures: Transition zone
between the primer and the embedding material used for micro-
toming. Lower Figures: Transition zone between CAA-oxides and
the primer. This part was microtomed without embedding material.

In the lower part of the Figure the typical CAA-oxide-struc-

ture is recognizable on which remains, particularly on the left
side, a primer in which again a bright zone in the boundary layer
can be seen. At the right, the crack changes from a cohesional
failure to an adhesional mode, but on the top of the oxides some
material remained, which is perhaps of different nature than the
polymer bulk itself. On the other hand, in the upper part of the
FAILURE MECHANISMS IN THE BOUNDARY LAYER ZONE 479

Figure there is the area of the primer surface which had to be

embedded before the cutting process. Between the primer and the
embedding material practically no dark areas caused by the pre-
sence of oxide particles are to be seen. The picture leads to the
conclusion, that the "adhesional failure" in reality occurred as a
a cohesional failure in the polymer near the top of the oxides.
This interpretation of the transmission electron micrographs is
confirmed by results from ESCA, by investigating the fracture
surfaces after such an adhesional failure. Some of the results
are shown in Table II.

Table II. ESCA investigations of an "Adhesional Failure" in an

Adhesively-Bonded Aluminium Joint.

Element Metal Surface Adhesive Surface

% %

0 21,3 15,3
N 1,1 1,3
C 65,4 78,8
Al 6,1 0,5
S 3,3 2,2
Na 2,9 1,6

Surface Treatment: CAA

Adhesive/Primer: AF 126/EC 3924

These results show that in the case of adhesional failure

on the metal side, relatively high amounts of carbon are detect-
able; and on the primer or adhesive side there are very low
amounts of aluminium or aluminium-oxides.

The carbon containing material on the metal surface may be

partly residual detergent on the oxide surface as a result of the
wet peel test procedure. But the high carbon/sulphur and carboni
sodium ratio makes it probable that at least portions of layer are
remains of the primer.

The situation is very different, if, for example, only the

chromic sulphuric acid etching as a surface treatment was used.
After an adhesional failure, not only the resin remains on the
metal side but oxides particles are also seen on the primer sur-
face, Figure 8.

From the TEM's in Figure 8 it is not possible to decide

precisely, whether the boundary failure occurs only in case an
influenced polymer zone is present. But it cannot be excluded that
between the thin oxide whiskers of the CSA-Iayer, polymer with
properties different from that in the bulk is existing.
480 w. BROCKMANN ET AL.

r-~ 0 ~.
PrImei'
•

L. EmbNdong Malfl"lOI

CSA o.~

Figure 8. TEM of cross sections from the polymer and the metal
sides of an adhesively bonded aluminium joint after "adhesional
failure". Surface treatment CSA.

These statements are true only for the system aluminium

with corrosion inhibiting primers; however some other results show
that on steel surfaces, too, in the case of adhesional failure on
the metal, polymer remain whose chemical and mechanical properties
seem to be influenced by the state of the metal surface. 9

The presence of weak boundary layers in polymers as an im-

portant reason for the water instability of the adhesion in such
systems is contradictory to the conclusions of Venables in the
foregoing chapter of this volume 10 . From his results he concluded
that an adhesional failure in aluminium/polymer bonds is mostly
due to hydrolysis of the oxides followed by crack in the hydroxy-
lated zones. But in comparing his results and the pictures shown
here, the different aging conditions in the two investigations
have to be taken into account.
FAILURE MECHANISMS IN THE BOUNDARY LAYER ZONE 481

It is true that a hydration process in the oxides can also

produce delaminations in an aluminium-polymer system or bond. But
the stability of certain anodic oxide layers against hydration is
relatively high. This is measurable by storing, for example, un-
protected anodized aluminium parts in a humid environment of 40°C
and 95 % relative humidity without condensation of water. Under
these conditions it needs hundreds of hours until hydration starts.
To initiate hydration in CAA-oxides coated with a 0.2 rom thick
primer layer (EC 3924) cured under normal conditions very severe
environmental conditions are needed. Hydration of the oxide layer
under the polymer as shown in Figure 9 was only possible, if the
specimen was stored for over 7 days at 133°C in water vapour with
a pressure of 2 bars.

Figure 9. TEM of cross sections after an aging process from CAA-

oxides coated with 0.2 rom thick primer EC 3924. Aging conditions:
7 days at 133°C in H2 0 vapour.

For example, in adhesively bonded aluminium jOints with

anodized surfaces and epoxide adhesives with corrosion inhibiting
primers aged over a year in humid climate of 40°C and 95 % reI.
humidity we could not detect partly hydrated oxides as shown in
Figure 9. Further we investigated recently an adhesively bonded
aluminium joint from the fuselage of British civil aircraft
Trident. The surfaces were shotblasted, chromic-sulphuric etched
and chromic acid anodized, and the adhesive was Redux 775, a modi-
fied phenolic resin. The fuselage was tested in a pressure tank
under water over 13 years and 139 500 simulated flight load cycles.
After this test the fuselage was stored without protection for
over 4 years on a British airfield. Despite this long aging process
482 W. BROCKMANN ET AL.

no hydration in the CAA-oxide layer under the phenolic resin was

detected, as shown in Figure 10. The specimen investigated was
microtomed in an area which is 5 mm away from the unprotected edge
of a bond between the stringer and the outer cover of the fuselage.

Figure 10. TEM of a cross section from a long time aged adhesive-
ly bonded aluminium joint.
Surface treatment: CAA
Adhesive: Redux
Aging time: 13 years under load in H2 0
+4 years in natural climate.

CONCLUSIONS

Insufficient water stability in the boundary layer of ad-

hesively bonded aluminium joints is easily detectable by the wet
peel test. Particularly, in the case of chromic acid anodized sur-
faces, a water-sensitive zone in the polymer on the top of the
FAILURE MECHANISMS IN THE BOUNDARY LAYER ZONE 483

oxides can exist. The polymer in this zone is of higher water sen-
sitivity than the bulk polymer and its properties seem to be in-
fluenced by the state of the oxides. High stability in the boundary
zone can only be observed if such a weak boundary layer does not
exist. The failure mechanism in the weak polymer zones described
here can only be induced in the presence of water by high plastic
deformations of the polymer. Under these conditions microcracks
can accelerate the migration velocity of water l l ,12. In the dry
state this weak boundary layer is not of importance for the be-
haviour of the bonds. If wet peel specimens after a peeling pro-
cess in the presence of water are redried, the crack again changes
from the boundary layer back into the adhesive as a cohesiona1
failure. This partial reversibility of the strength in the bound-
ary layer in the presence or absence of water is an indication too,
that the weakening mechanism in the boundary layer is not due to
hydration processes of the oxides, which are not reversible under
all aging processes.

All of the statements given in this report are valid only

for aging processes of polymer/metal systems without primary cor-
rosion attack of the metal parts. If primary corrosion occurs, for
example, under condensing humidity at unprotected edges of the
jOints and aging times of more than several days, "bond line cor-
rosion failures" are observed in some cases which are induced by
other failure mechanisms, in which the chemical stability of the
oxides plays an important role.

REFERENCES

1. P.F.A.Bij1mer and R.J. Sch1ieke1mann, SAMPE-Quarter1y 1, 13

(1973) .
2. J.C. McMillan, "Surface Preparation, The Key to Bondment
Durability", in: Bonded Joints and Preparation for Bonding,
AGARD-Lect. Ser. 102, 1979.
3. O.-D. Hennemann and W. Brockmann, J. Adhesion 12, 297 (1981).
4. A. Hartman, Report of Nationaa1 Luchtvaart-Laboratorium
Amsterdam 1961.
5. W. Brockmann, Adhesives Age 20, 30 (1977).
6. W. Wernick and R. Pinner,"TheSurface Treatment and Finishing
of Aluminium and its A11oys,"Fourth Edition, R. Draper,
London 1972.
7. K.K. Knock and M.C. Locke, in "Adhesion Aspects of Polymeric
Coatings", K.L. Mitta1, Editor, p. 301, Plenum Press, New York
1983. .
8. J.A. Marceau, "An SEM Analysis of Adhesive Primer Oriented
Bond Failures on Anodized Aluminium", SAMPE Quarter1y!Q., 1,
(1978)
9. D.M. Hercules, J. Electron Spec. Re1. Phen. ~, 811 (1974).
484 W. BROCKMANN ET AL.

10. J.D. Venables, these proceedings, pp. 453-467

11. C.L. Leung and D.H. Kae1b1e in "Resins for Aerospace",
C.A. May, Editor, ACS Symp. Series 132, p. 419,
Washington 1980
12. R.J. Morgan and E.T. Mones in "Resins for Aerospace",
C.A. May, Editor, ACS Symp. Series 132, p. 232,
Washington 1980.
JOINT DURABILITY STUDIES WITH ABRADED, ETCHED, COATED AND

ANODIZED ALUMINUM ADHERENDS

J. Dean Minford

Product Engineering Division

ALCOA Laboratories
Alcoa Center, PA 15069

The effect of surface pretreatment on the

aluminum oxide surface is critical and intimately
related to the bond permanence of aluminum adhesively
bonded joints. It is difficult, however, to distin-
guish the benefit of different surface treatments in
the absence of some weathering test condition since
the joints will fail cohesively for many different
surface-treating conditions when tested without such
exposure. Aluminum alloy 6061-T6 was pretreated by
various procedures including sanding, grit blasting,
hot acid etching, heat-cured epoxy paint pr1m1ng,
Alodining, and sulfuric acid anodizing. Both room-
temperature-curing two-part and one-part heat-curing
epoxy adhesives were used to fabricate the test
specimens. The weathering evaluation procedures
included both accelerated laboratory-type and natural
atmospheric exposures. The list of weathering
conditions included: immersion in room-temperature
water, exposure to continuously condensing humidity at
52°C (125 F), and hot water soak/freeze/thaw cycling~
as well as, exposure in an industrial-type atmosphere
and a seacoast atmosphere.

485
486 J.D. MINFORD

INTRODUCTION

While many theories have been proposed to explain the

various aspects of the adhesion phenomena, no universal theory of
adhesion exists. l ,2 It is during the adsorption phase of the
bonding process that the forces of adhesion become effective and
whether highly durable joints are formed at this time is
significantly influenced by the characteristics of the adherend
surface.3-S, 10-20

Two types of debonding are generally recognized for

weathered metallic adherend joints, as studied by Gledhill and
Kin1och. 9 The first of these is promoted by the presence of
water in the environment and can be considered a preferential
desorption of the organic polymer adhesive by the highly polar
water molecules at the interface. In this process, the aluminum
oxide interface generally is converted to a cohesively weak
boundary layer of aluminum hydroxide resulting in a joint failure
appearance described as being interfacial in nature to the
adherend. Another effect of the water may be to weaken the
cohesive strength of the cured adhesive itse1f~ in which case,
the breaking load on the joint can be significantly below the
original value, yet the nature of the appearance of the failed
joint may be largely cohesive in nature. The second type of
debonding is where corrosive water is present, and there can be
gross corrosion of the aluminum as the failure mechanism. These
failure mechanisms would indicate that to resist joint failure,
more stable oxides and stronger interfacial forces need to be
forged from better surface preparations.

It is the purpose of this paper to make suitable comparisons

regarding the long-term durability tests between a variety of
surface pretreatment conditions on aluminum including abrasion,
etching, coating and anodizing techniques. The direct comparisons
between surface pretreatments can only be made because the
experimental variables of alloy, adhesives, curing conditions and
weathering exposures were maintained as constants in experiments
conducted over a twenty-year period of investigation.

MATERIALS ANO METlDOS

Tensile-shear 6061-T6 aluminum lap specimens in 3.1S mm

(0.125 n ) thickness were prepared according to ASTM 01002-72. This
adherend has been used for more than 20 years at Alcoa because of
its wide use in aluminum structures both as rolled sheet or
extrusions. It also affords comparable joint strengths in the
sheet thickness used here to those obtained with the prescribed
thickness of Alclad 2024-T3 aluminum recommended in ASTM 01002-72.
JOINT DURABILITY STUDIES WITH ALUMINUM ADHERENDS 487

For comparison purposes between surface pretreatments, the

same two-part, room-temperature-curing epoxy (National Starch
82-0688/84-9023) and one-part, nitrile~odified, heat-curing epoxy
(3M EC-2086) were used to bond each differently prepared adherend
surface. These same adhesives have been used for more than 15
years in Alcoa investigations that have been carried out to
compare a variety of different bonding parameters.

Similarly, the same long-time exposure weathering conditions

consisting of immersion in room-temperature· deionized water,
soaking in continuously condensing humidity at 52°C (125°F),
repeated cycling for consecutive 24-hour periods in 74°C (165°F)
water, -34°C (-30°F) freezing air and 77°C (l70 0 F) hot air oven,
or an industrial or seacoast-type natural atmosphere have been
continued in use.

Surface Preparations

The following surface preparations were investigated:

A. Vapor degreasing in refluxing perchloroethylene vapor.
B. Belt sand abrasion with Aloxite 120 grit cloth followed by
vapor degreasing in perchloroethylene.
C. Blasting with silica grits to uniform appearance followed by
vapor degreasing.
D. Alcoa 3 (chromic-sulfuric acid etching) - The surfaces were
degreased in perchloroethylene vapor and treated by immersion
ih a solution of 35 grams of chromic acid and 100 ml of
concentrated sulfuric acid in a liter of distilled water
followed by a tap water and deionized water rinse and 121°C
(250°F) forced air dry. The solution temperature was 82°C
(180°F) and immersion time was 5 minutes.
E. Alodining - As prescribed by Amchem, the supplier of Alodine
1200.
F. Alumilite 215 (sulfuric acid anodizing) - The surface was
pretreated by cleaning for 5 minutes in Diversey 914 at 66°C
(150°F) and etched 8 minutes in Alcoa 1 (5 percent Pennwalt
AE16 caustic at 49°C (120°F) followed by desmutting three
minutes in 8-10 percent Sanfax MT324 at room temperature.
This thick architectural-type coating was applied by
anodizing in 16 percent sulfuric acid at 21°C (70°F) at
12 amps/ft 2 for 60 minutes to produce a coating about 0.95
mil thick. A tap water rinse was followed by a 10-15 sec.
hot deionized water rinse (66-82°C, l50-l80 0 F) and air dry.
Sealing was in pH 6 deionized boiling water for 30 minutes.
488 J.D. MINFORD

G. Vapor degreasing in refluxing perchloroethylene. Spray

application of Midland Chemical V-80S epoxy-phenolic coating
and heat curing as recommended by manufacturer.

RESULTS AND DISCUSSION

Initial Joint Strengths

The average initial joint strengths obtained for both adhe-

sives and ten different conditions of surface pretreatment are
shown in Table I and have been graphed in Figure 1. A joint code
number which refers to a fixed set of bonding parameters including
alloy, surface pretreatment, adhesive, and curing condition has
also been assigned in Table I. This code number has been used in
subsequent tables along with only metric units of temperature and
joint strength to conserve space in reporting the durability test
results. It can be observed that a wide range of joint strengths
result from the various surface pretreatments. For the room-
temperature~uring epoxy, the range was from 13.90 MPa to 20.19
MPa, whereas the heat~ured epoxy joints were in a significantly
wider and higher joint strength range from 17.02 MPa to 44.65 MPa.

Table I. Joint Strength with Different Surface Pretreatments.

Joint CUring Average Joint

Code No. Adhesive Pretreatment Time Temp,oC Strength, MPa

lC 82-0688/84-9023 Vapor Degreased 7 days 25 14.68

2C EC-2086 Vapor Degreased 15 min. 204 29.68
lA 82-0688/84-9023 Belt Sanded 7 days 25 13.99
2A EC-2086 Belt Sanded 15 min. 204 33.55
3A 82-0688/84-9023 Silica Grit Blast 7 days 25 13.90
lE 82-0688/84-9023 cr0 4/H2S04 Etch 7 days 25 19.50
2E EC-2086 CrO /H2SO Etch 15 min. 204 36.72
lCO 82-0688/84-9023 Al~ine 1~00 7 days 25 7.79
200 EC-2086 Alodine 1200 15 min. 204 17.02
3<P 82-0688/84-9023 Epoxy-Phenolic 7 days 25 20.19
400 F;C-2086 EPoxy-Phenolic 15 min. 204 36.52
lAN 82-0688/04-9023 Alumili te 2l5 l 7 days 25 17.43
2AN EC-2086 Alumilite 215 1 15 min. 204 22.53

1 - Alcoa Anodize Trademark

The magnitude of the joint strengths obtained with the various

surface pretreatments merits some discussion. Any increase in
joint strength due to abrading the adherend is strongly influenced
by the nature of the adhesive. For example, belt sanding or
JOINT DURABILITY STUDIES WITH ALUMINUM ADHERENDS 489

silica gri t blasting obviously produced physical roughening of the

surface. No increase in joint strength resulted, however, as com-
pared with the smoother, vapor degreased surface when using the more
brittle, room temperature-curing epoxy adhesive. A significant
increase in joint strength did occur, however, when the tougher,
more crack-resistant, heat-cured epoxy adhesive was used. Chemical
deoxidizing of the adherend produced significant joint strength
gains using both adhesives because of a significant enhancement in
the chemical wettability of the surface. The conversion coating
presents a unique problem greatly influenced by a combination of
coating thickness and adhesive curing conditions. Increasing
thickness increases the possibility of finding lower joint load-to-
failure values and weak boundary layer conditions present in the
coating cross section. This is most evident in the very low joint
strength of the room temperature-curing epoxy adhesive. A further
dramatic change in physical cohesive strength of the conversion
coating can occur where the adhesive is cured at elevated tempera-
tures, as shown by the EC-2086 heat-cured joints. Heat-cured paints
used as primers are capable of developing joint strengths equivalent
to those of the best deoxidized surface pretreating. The anodized
aluminum surface again presents a thickness-sensitivity situation,
as shown for the conversion coatings. The decreasing joint strength
accompanying increasing anodizing times for aluminum has been shown
by Minford 6 to be an example of variable thickness anodic oxide
layers breaking through planes of weaker structure.

40
(5806)
r-- .--
;--

~ 30
(4354)
- -
&.
~
~
C, -
t: 20
- - -
~ (2903)
en
!E - .--
0
0
r-
.11
:;;
."Q.E ~
:;;
E

.11
-
.I::. ;-- ~
en "C
c3 c3 "
0
0
"0 "0
e: "C :l2
"" ·u
e:
Co 10
"C
il: ~ "
"C
e: 1;;
(/) (/)
'0 ~ .<= "
.<=
« .:J.
-
l-
~m "C
~ '" i -£ 0
« "e:
Q. Q.
>. 0" .Y ~
:;:0
o.~'"
j (1451) 8.", ",iii ~J: >. .t:.~
!::-g"
~J:
'0
c2~
)(
;-- )(
:>"C
~~ ~~ ~~
a; ~:c Q~ 0
«
0
Co
0
Co =0
:>e:
=0
:> e:
(/)«
CD
al Cii<.:l UW UW W W (/)«
COM CO COM CD
CO COM COM CD COM CD COM CD
COM CD
CON CON CON CON CO CON CO CON CO CON CO
lil lil 0 0
lil 0
~~ U
~:;;
U 9~ 9~ N
U
~:;; N
9~ U
~:;; N
N~
co co W
NoO-
co co W
N~
coco N~
coco W
N~
coco
U N~
coco W
No::}
coco
U
Joint W W

Code No. 1C 2C 1A 2A 3A 1E 2E 1CO 2CO 3CO 4CO 1AN 2AN

Figure 1. Effect of adhesive and surface pretreatment on joint

strength.
490 J.D. MINFORD

Bond Durability in Room Temperature Water

The relative joint durability performances after 24 months

of continuous soaking in deionized water are shown in Table II and
have been plotted in Figures 2 and 3. No joints lost as much as
50 percent of their initial strength although the vapor degreased-
only surface pretreated joints degraded 47 percent. Six of nine
surface pretreatment condition joints showed less than 20 percent
strength loss in the two-year period. The highest overall joint
strength retention was 19.29 MPa for the two-part epoxy and
30.32 MPa for the heat cured one-part epoxy. In each case, the
adherend pretreatment was a heat-cured epoxy phenolic paint over
a vapor degreased-only surface.

Table II. Room Temperature Water Immersion Tests.

Average Joint Strength

Joint After Indicated Exposure Time, MPa % Strength
Code No. 3 Mo. 6 Mo. 12 Mo. 24 Mo. Retention

lC 8.96 7.34 8.47 7.79 53

2C 21.15 13.78 10.82 8.06 27
lA 12.73 12.88 12.35 * 88
2A 25.98 23.63 21.36 14.95 45
3A 14.88 * 12.88 14.68 100
IE 20.46 * 17.43 18.40 94
2E 30.80 30.32 28.73 24.32 66
ICC 7.85 6.75 5.99 4.69 60
2CP 17 .02 14.95 14.95 11.51 67
3CO 20.88 20.19 20.88 19.29 96
40:> 34.93 32.38 31.49 30.32 83
lAN 17.57 17.02 15.16 14.95 86
2AN 17 .02 14.59 16.54 16.74 74

*No Test Result

In spite of the fact that the highest strength retention was

obtained with heat-cured EC-2086 joints, one should recognize from
the overall comparatively steeper decline in joint strength for
the heat-cured EC-2086 joints in Figure 3, as compared with the
room temperature-cured epoxy joints in Figure 2, that performance
for this adhesive is more dependent on good surface preparation.
Thorough surface deoxidizing, priming with a paint, anodizing, or
conversion coating are required for good EC-2086 joint durability
even in relatively mild room temperature water-soaking conditions.
JOINT DURABILITY STUDIES WITH ALUMINUM ADHERENDS 491

40 I
(5806 )

_ 30
-
l (4354)

l.
~
4i
~
!
20
(2903)
.
n
u
n
3CO,
II) X-----X _____X 6 1E
.. ----X_ 1AN

3A

~~
-t1A

-
10 -
(1451)
• 1C

• 1CO

I I
6 12 18 24
Exposure TIme In Months

Figure 2. D.1rability of room-temperature-curing epoxy joints in

room temperature water-soaking exposure (for code no., see
Table I).

40
(5806)

4CO
30
~ (4354)
~
~ 2E

=
2' 20 ~.
.
~ (2903)
<II x
2AN

~ 2A
Q.
,j
10 2CO
(1451) 2C

6 12 18 24
Exposure TIme in Months

Figure 3. Durability of heat-cured epoxy joints in room-

temperature water-soaking exposure (for code no., see Table I).
492 J.D. MINFORD

Bond Durability in Condensing Humidity at 52°C (125°F)

The results are summarized in Table III and Figure 4.

Although a limited number of surface pretreatment joints were
exposed to this condition, but some valid comparisons can still
be made.

Table III. Exposure to 100% Relative Humidity at 52°C (125°F).

Average Joint Strength

Joint After Indicated Exposure Time, MPa % Strength
Code No. 3 Mo. 6 Mo. 12 Mo. 24 Mo. Retention

lA 10.32 10.80 11.07 * 79

2E 14.68 8.75 4.55 6.10 17
3CO 19.77 18.60 17.91 17.43 86
400 28.73 25.70 22.25 19.50 53
!AN 14.33 12.40 13.09 15.16 87
2AN 18.33 16.54 18.12 17.91 80

*No Test Result

40
(5806)

~ 30
~ (4354)
~
~

4CO
3eo
2AN x

--~----------x--------------------~~
1AN
10
(1451)
2E

6 12 18 24
E)(posure Time in Months

Figure 4. Durability of heat-cured epoxy joints in condensing

humidity at 52°C (125°F) exposure (for code no., see Table I).
JOINT DURABILITY STUDIES WITH ALUMINUM ADHERENDS 493

Strength retention of 80 percent or higher was shown by at least

one example of an abrasion, etching, coating or anodizing proce-
dure with the room temperature-curing epoxy. Only the sulfuric
acid anodizing treatment achieved this percentage retention using
the heat-curing epoxy. Once again the epoxy phenolic paint-primed
joints remained strongest overall at 17.43 MPa for the two-part
epoxy joints and at 19.50 MPa for the heat-cured epoxy joints.

It should be noted that thorough deoxidizing of the adherend

was not an effective treatment for promoting a high level of joint
durability with heat-cured EC-2086 adhesive in this environment.
Since this laboratory weathering condition is closely related to
accelerated tropical-exposure conditions, one would conclude that
the adherend surface would need to be primed, anodized, or
conversion coated for long-time tropical service.

Bond Durability in Soak/Freeze/Thaw Cycling

The data summarized in Table IV and plotted in Figures 5 and

6 show no surface pretreatment could overcome a tendency for sig-
nificant joint strength loss for the heat-cured epoxy joints. As
expected, the Alcoa 3 (similar to FPL etch) pretreatment afforded
the highest strength retention in this aggressive water weathering
exposure for the two-part room temperature-cured epoxy joints.

Table IV. Soak/Freeze/Thaw Cycle Testing

Average Joint Strength After

Joint Indicated Exposure time, MPa % Strength
Code No. 3 Mo. 6 Mo. 12 Mo. 24 Mo. Retention

lC 0.83 0 0 0 0
2C 13.09 11.51 7.30 3.45 12
2A 12.4 11.23 9.65 0 0
3A 8.82 8.75 6.44 0 0
IE 17.71 13.71 13.57 14.47 74
2E 11.71 10.54 9.16 0 0
lCO 5.72 5.55 5.27 0.83 11
200 17.08 12.94 16.54 0 0
!AN 13.57 12.88 10.54 8.41 48
2AN 15.64 13.99 15.16 6.41 28

An important fact that needs to be recognized when testing a

room temperature-curing epoxy adhesive is that the water tempera-
ture of the test environment or the air temperature in part of a
cyclic weathering condition can have strong influence. When a
494 J.D. MINFORD

room temperature-cured epoxy adhesive is exposed to a signifi-

cantly higher air- or water-temperature condition in test, there
is a tendency for additional cross-linking to take place, and this
occurs before any appreciable degrading action due to other
factors that may be present. In the cyclic weathering environment
used in this investigation, the adhesive, already cured at room
temperature, experiences both 74°C (l65°F) water and 77°C (l700F)
air during each three-day cycle. Wi thout special surface
preparation the joint is still degraded to the point of failure
within three months~ but with special surface preparations, the
room-temperature-curing joint performances, as shown in Figure 5,
are significantly superior to that predicted from the EC-2086
data in Figure 6. One should also note from Figure 5 that a
thoroughly deoxidized surface preparation is equal in performance
to an anodizing preparation using the room temperature-curing
epoxy. In contrast, this surface preparation had been inadequate
even in hot humidity when using the heat-cured nitrile modified
epoxy adhesive.

® r------------r----------~------------,-----------_.
(5806)

_ 30
~ (4354)

~
:iE

lE

-x--______________ ~

10
(1151) lAN

12
Exposure lime in Months

Figure 5. Durability of room-temperature-curing epoxy joints in

soak/freeze/thaw cycling exposure (for code no., see Table I).
JOINT DURABILITY STUDIES WITH ALUMINUM ADHERENDS 495

40
(5806)

30
~ (4354)

~
:E
:5
en
c
e 20
(2903)
<is
tv/I>
.r:
(/)
•
c.
~
10
(1151 )

6 12 18 24
Exposure Time in Months

Figure 6. Durability of heat-cured epoxy joints in soak/freeze/

thaw cycling exposure (for code no., see Table I).

Bond Durability in Industrial Atmosphere

The data are summarized in Table V and plotted as strength-

retention curves over an eight-year exposure time in Figures 7
and 8. It is clear that the industrial atmospheric exposure does
not deteriorate adhesive-bonded aluminum joints to the same degree
as shown in earlier accelerated laboratory-exposure tests. Added
confidence in the acceptable durability of such joints comes from
the fact that this exposure condition represents the real service
condition for the majority of structures. Since no edge or
top coating protection was provided for these joints, there is
the possibility of some acceleration still occurring in this
weathering condition.

From the room temperature-curing epoxy data in Figure 7, it

is clear that all forms of surface pretreatment attempted lead to
durability enhancement as compared to the vapor degreased-only
surface condition. An identical summary statement can be made
after studying the results in Figure 8 obtained with the one-part
heat-cured epoxy.
496 J.D. MINFORD

Table v. Results of Industrial Atmospheric Weathering.

Average Joint Strength After

Joint Indicated Exposure Time, MPa % Strength
Code No. 1 Year 2 Years 4 Years 8 Years Retention

1C 12.20 10.54 9.09 6.25 43

2C 22.05 21.36 19.29 3.10 10
2A 29.63 28.46 25.98 23.43 70
3A 13.78 14.26 13.99 13.99 100
IE 18.60 20.67 20.23 18.71 100
2E 37.21 36.52 35.83 32.59 59
1CO 8.75 8.27 8.96 7.79 100
2CO 18.60 17.91 18.60 19.09 100
3CP * 19.50 20.77 19.70 98
4CO * 33.97 30.56 33.35 91
IAN 16.05 14.68 12.88 10.92 63
2AN 19.29 19.09 19.50 18.09 80

40 r-----,-----_r-----,------r-----~----_r----~----~
(5806)

30
~ (4354)
~
~
~
-5
en
c 20 3eO
~ 0
(i) (3903) A
1E
lij
<11
.r:. 3A
fJ)
c. 1AN
j
10
-x
(1151) 1CO

1C

2 3 4 5 6 7 8
Exposure Time in Years

Figure 7. Durability of room-temperature-curing epoxy joints as

a function of industrial atmospheric exposure (for code no., see
Table I) •
JOINT DURABILITY STUDIES WITH ALUMINUM ADHERENDS 497

40 ,-----~----.-----._----_r----_.----_.----_.----_.
(5806)

2E
~ 30 o 4CO
! (4354)

if.
~ 2A

10
(1151 )

2C

3 4 5 6 7 8
Exposure Time in Years
Figure 8. Durability of Heat-CUred Epoxy Joints in Industrial
Atmospheric EXposure.

Bond Durability in Seacoast Atmosphere

The data are shown in Table VI and plotted in Figures 9 and

10. As indicated in the Introduction section of this paper, we
must consider that bonded metallic joints can fail due to
debonding by water penetration to the interface, or gross pitting
corrosion of the adherend can destroy the adhesion by converting
the metallic surface to corrosion product. In most cases, the
cured adhesive may be left unaffected because the factors
creating the condition for metallic corrosion are not necessarily
aggressive to the cured polymer. As with the industrial
atmospheric exposure just discussed, we must consider that
exposure to a seacoast atmosphere is the real world for joints
whose application brings them into contact with corrosive
saltwater conditions.

The results shown for the room temperature-curing epoxy

joints in Figure 9 fall into predictable and nonpredictable
categories based on prior experience. The experience to which we
refer is whether the surface, as treated, is inherently resistant
to corrosive chloride ion attack in the sheet surface condition.
498 J.D. MINFORD

Table VI. Results of Seacoast Atmospheric Exposure.

Average Joint Strength After

Joint Indicated Exposure Time, MPa % Strength
Code NO. 1 Year 2 Years 4 Years 8 Years Retention

lC 0 (All failed 71 days) 0

2C 0 (All failed 270 days) 0
2A 11.99 (All failed 760 days) 0
3A 16.74 12.40 10.82 9.30 67
IE 17.43 0 (All failed 760 days) 0
2E 34.66 23.22 0 (All failed 0
1440 days)
lCO 7.43 4.58 2.76 (All failed 0
1475 days)
2CO 16.74 14.68 16.26 17.78 100
3eD * 11.09 5.17 (All failed 0
1497 days)
4CO * 31.90 30.04 71
IAN 15.64 14.81 14.29 13.94 80
2AN 19.29 14.67 16.26 17.77 79

Of the various treatment conditions, one would expect both the

vapor degreased and the deoxidized 606l-T6 surface to develop
pitting readily, and these two kinds of joints did degrade first.
We can explain the significantly longer survival time of the de-
oxidized surface joints as due to the much higher order of initial
surface wetting 5 and adhesive spreading on the deoxidized
surface. It was unexpected, and still not completely understood,
to find the silica grit blasted joints surviving eight years
exposure with no significant joint strength degradation. We might
relate this result to the probable deposition of a silicate
barrier film on the surface which resists undercutting pitting on
the adherend. We have many years of experience demonstrating the
effectiveness of anodizing films on aluminum as barriers to
pitting corrosion, and this effect is clearly evident in Figure 9.
The Alodine 1200 dhromate-conversion coating has been an effective
barrier to undercutting corrosion of heat-cured paint coatings in
products like aluminum siding. This protection was only partial-
ly effective through four years of exposure (see Figure 9). We
relate this to the fact that a change in the durability properties
of the conversion coating results from heat curing after an adhe-
sive or coating is put in place, but this change is not realized
where a room temperature-curing adhesive is used. Finally, we
believe the failure of the paint - primed" room temperature-curing
joints was due to the use of only vapor degreasing as a pretreat-
ment under the paint primer and less-than-optimum wetting of the
primer by the room temperature - curing adhesive.
JOINT DURABILITY STUDIES WITH ALUMINUM ADHERENDS 499

~ r----r---'r----.--~r----.----.----'--~
(5806)

~ 30
~ (4354)
~
::iE
-=01
~ 20
u; (2903)
~ K~
~
Q.

!I
3A

1C
7 8
Exposure Time in Years

Figure 9. Durability of room temperature-curing epoxy joints in

seacoast exposure (for code no., see Table I).

~
(5806)

~ 30
~ (4354)

~
::iE
4CO
-=c01 20
~ (2903) 1AN
j '-4-~~~==:j~======~==================~2~C~O
Q.
!I 10
(1451)

7 8
Exposure Time in Years

Figure 10. Durability of heat-cured epoxy joints in seacoast

exposure (for code no., see Table I).
500 J.D. MINFORD

Some of the same comments used to explain the results of

Figure 9 also apply to Figure 10. The poor durability perform-
ances of the vapor-degreased, belt-sanded, and deoxidized surface
joints is expected because the surfaces are all sensitive to
pitting corrosion. One needs to note, however, that the times to
failure are all significantly extended over the room temperature-
curing epoxy results because of the better opportunity of the
adhesive to wet and spread over these surfaces due to heat curing.
The anodized surface joints are virtually unaffected in eight
years, as anticipated. The effect of heat curing of an adhesive
over the Alodine 1200 conversion coating is striking, with the
result that the durability is equal to that produced by anodizing.
Finally, we also note a significant effect produced by heat curing
the adhesive over the heat-cured paint primer, which originally
had been applied over vapor degreased-only adherends. In a
recent technical paper, Minford lO has reported that thi s
long-term durability of EC-2086 adhesive over epoxy-phenolic
paint- primed aluminum has continued over a ten-year exposure
period.

CONCLUSIONS

1. A wide variety of joint strength responses to various

weathering conditions was achieved through surface abrading,
etching, coating or anodizing procedures.

2. All of the pretreatments employed for this investigation

showed acceptable joint strength retention in continuous
water soaking exposures for two years when using a room
temperature-curing epoxy. Only the acid etching, conversion
or paint coating and anodizing pretreatments were acceptable
for a heat-cured epoxy.

3. Only etching and anodizing pretreatments were satisfactory

for room temperature-curing epoxy joints in soak/freeze/thaw
cycling. None worked for the one part heat-cured epoxy.

4. All four kinds of pretreatment were effective in eight year

industrial atmospheric exposure. Anodizing and conversion or
paint coating procedures were most satisfactory in seacoast
weathering.
JOINT DURABILITY STUDIES WITH ALUMINUM ADHERENDS 501

REFERENCES

1. B. o. Bateup, Int. J. Adhesion Adhesives, 233 (July 1981).

2. R. J. Good, J. Adhesion, 8, 1(1976).
3. J. D. Minford, Adhesives Age, 17 (7), 24 (1974).
4. J. D. Minford, in "Treatise onAdhesion and Adhesives", R. L.
Patrick, Editor, Vol. 3, Chap. 2, Marcel Dekker, New York,
1973.
5. J. D. Minford, in "Treatise on Adhesion and Adhesives", R. L.
Patrick, Editor, Vol. 5, Chap. 3, Marcel Dekker, New York,
1981.
6. J. D. Minford, SAMPE Quarterly, 7, 18 (1978).
7. J. D. Minford, Paper presented at the ASM/SAMPE Conference on
Specia:lized Cleaning, Finishing, and Coating Processes, Los
Angeles, February 1980.
8. J. D. Minford, Adhesives Age, 23 (10), 36 (1980).
9. R. A. Gledhill and A. J. Kinloch, J. Adhesion, ~, 315 (1974).
10. J. D. Minford, Paper presetned at the ASM Surtech and Surface
Coating Exposition-82, Dearborn, May 1982.
11. H. W. Eickner, Forest Products Laboratory, Madison, WI, WADC
Technical Report 54-447, Part I, 1955.
12. H. W. Eickner, Forest Products Laboratory, Madison, WI, WADC
Technical Report 54-447, Part II, 1957.
13. H. W. Eickner, Forest Products Laboratory, Madison, WI, WADC
Technical Report 54-447, Part III, 1958.
14. H. W. Eickner, Forest Products Laboratory, Madison, WI, WADC
Technical Report 59-567, Part I, 1960.
15. H. W. Eickner, Forest Products Laboratory, Madison, WI, WADC
Technical Report 59-564, Part I, 1962.
16. R. F. Wegman, W. M. Bodnar, E. S. Duda and M. J. Bodnar,
Adhesives Age, 22 (October 1967).
17. G. F. Carter, ASTM STP 401, 28 (August 1966).
18. R. F. Wegman, Picatinny Arsenal Technical Report 4169, Dover,
NJ, June 1971.
19. J. A. Marceau and W. Scardino, Air Force Materials Laboratory
Report, AFML-TR-75-3, Wright-Patterson Air Force Base, Ohio,
February 1975.
20. J. C. McMillan, AGARD Lecture Series No. 102, Wright-Patterson
Air Force Base, Dayton, Ohio, October 1979.
COMPARATIVE STUDY OF ALUMINUM JOINT STRENGTH AND DURABILITY WITH

VARYING THICKNESS, BOEHMITE-TYPE OXIDE SURFACES

J. D. Minford

Product Engineering Division

Alcoa Laboratories
Alcoa Center, PA 15069

The presence of boehmite oxide on aluminum surfaces

has often been associated with weather-durable bonding.
In this work, a considerable variance was shown in
aluminum joint durability involving such oxide surfaces
with varying thicknesses as generated in the absence or
presence of electrolyte in boiling water7 i.e., deionized
or tap water. The changes in oxide were generated
starting from a deoxidized surface base by varying the
exposure times in boiling water.

Joint strengths tended to decrease with increasing

oxide thickness7 i.e., longer exposure times, and could
be as much as 50 percent lower in strength than the
deoxidized surface control joints. variable durability
patterns resulted from varying weathering conditions7
also, when stress was simultaneously imposed.

Bonding to deionized water-produced oxide surfaces

showed enhanced durability using unstressed joints in
strictly water-soaking conditions or corrosive salt fog
cycle exposures. l)Jrability generally increased as the
oxide buildup in deionized water increased and was
inversely related to initial joint strength. In addition
to generally lower durability in these exposures, the
tap water-produced oxide bonds had a different pattern,
peaking in durability after ten minutes' exposure.

503
504 J.D. MINFORD

With simultaneously imposed stressing conditions, the

tap water-produced oxide surface joints assumed the great-
er durability. These joints peaked in durability with the
five-minute generation oxide condition. Performing overall
with lower durability potential, the deionized water-pro-
duced oxide joints had their best response at ten-minute
oxide generation level.

Very favorable durability responses were found for

these predominantly boehmitic oxide-type compared to
similarly fabricated and tested joints reported by Alcoa
using acid deoxidized, and chromic or phosphoric-acid-
anodized 6061-T6 adherends.

This procedure could have some practical application

for low-cost general manufacturing operations where the
simple boiling water bath treatment would be replacing
more expensive aerospace-type surface preparations, like
modified FPL and chromic or phosphoric acid anodizing
procedures.
IN TRODUCT ION

Wegman et aLI,2 studied how the oxides on 2024-T3 aluminum

affect the bond strength of joints by deoxidizing the adherend in
hot sulfuric acid-sodium dichromate solution and establishing that
a new oxide developed after immersion in hot deionized or tap
waters. Low joint strengths and adhesive-type failures resulted
when bonding to the former. High joint strengths and cohesive-
type failures were found with the latter treating conditions.
They concluded that a thick, hydrated, cohesively weak surface
oxide formed in deionized water from which strong bonds could not
be obtained. The slight acidity present that promotes multivalent
compounds in tap water was said to prevent generation of this
hydrated layer. No durability test data were offered in these
investigations. More recently other investigators, among them Wu
and Bowen3 ,4, have attached specific significance to generating
the boehmi tic form of oxide for most durable bonding wi th high
initial strengths. I t is the intent of this investigation to
study some aspects of both considerations by studying the bond
durability of joints fabricated with surfaces generated from
either hot deionized or tap water immersion processing. The water
temperature was deliberately maintained at lOOOC (2l2°F) in order
that any oxide generated would essentially be the boehmitic type.
Additional variances from the Wegman work included substitution of
606l-T6 for 2024-T3 alloy and deoxidizing in hot caustic (Alcoa 1
Process) instead of sulfuric acid-sodium dichromate solution.
Finally, because Russell and Garnis 5 had demonstrated signifi-
cant differences in joint strength after varying water rinse
conditions, a decision was made to study the boehmite generation
effects using a range of water-exposure times.
ALUMINUM JOINT STRENGTH AND DURABILITY 505

Harrington and Nelson 6 first reported the film substance

formed on aluminum by boiling water treatment was boehmite. They
further found that temperatures above 80°C (176°F) were necessary.
Bryan 7 and Al tenpoh1 8 found that with alloys of aluminum, the
reaction was initially very rapid, with fairly quick slowing of
the reaction afterwards, until virtual cessation after four to
eight hours. The boehmite films formed were intermediate in
thickness between air-formed films and anodic films. In the
immersion times used for this study (one- to 60-minute range) the
oxide film increase would be virtually linear according to
Hart 9 , who also found the rate seemed to be the same whether
oxygenated or deoxygenated water was used.

METHODS AND MATERIALS

Individual lap joints were fabricated and tested according to

AS'lM 01002-72. Aluminum alloy 6061-T6 was vapor degreased and
deoxidized by etching in a 66 DC (150 OF) 5 percent sodium
hydroxide solution for two minutes. The resulting smut was
removed by immersion in room temperature 40 percent nitric acid,
rinsing in cold tap water, and immediately drying in an air jet;
then immersing in boiling New Kensington, PA tap water or
deionized water for either 1, 5, 10, or 60 minutes. Immediately
after hot air drying, joints were made with EC-2086 nitrile epoxy
adhesive cured in the bondline for 15 minutes at 204°C (400 0 F).

In order to make comparisons among such a wide variety of

adherend pretreatments and different weathering exposure times,
it was pertinent to use an Alcoa-developed technique which has
been described. ll In brief, exposure specimens are periodi-
cally tested to 50 percent of their estimated initial joint
strengths based on actual initial test specimen values as shown
in Table I. Joints which continue to pass this criterion
continue to be exposed until they are deliberately stressed to
failure at the end of a sufficiently long exposure time. The
Alcoa Durability Factors shown in the tables of this paper are
based on calculations which consider how much of the initial
strength was retained for how long an exposure time and can be
written as:

Durability Factor - % Retained Joint Strength x Days in Exposure

- 1000
In the evaluation of performance under simultaneous stressing
and weathering exposure conditions the above formula was altered
to read:

Imposed Stress in psi x Joint Failure Time in Days

Durability Factor
1000
506 J.D. MINFORD

RESULTS AND DISCUSSION

Initial Joint Strengths

The average initial joint strengths obtained from 11 different

adherend pretreatment conditions are shown in Table I and graphed
in Figure 1. The deoxidized surface permitted highest joint
strength development as expected. We assume increasing exposure
time in water correlates wi th increasing oxide residual on the
surface. We should expect the overall joints to decrease in
strength as the oxide layer thickens. However, in Figure 1 we
note that the relationship between exposure time and joint
strength is quite different in deionized as compared to tap water.
In deionized water, the average joint strength with increasing
soaking time follows an exponential decay-type curve; whereas
there does not appear to be any predictable relationship between
joint strength level and soaking time through ten minutes for
tap water-treated joints. After an exposure time of 60 minutes,
however, joint strength is significantly lower and similar for
joints derived from either water treatment. Also, we do not
observe, as Wegman did, any striking difference between the types
of bondline failures noted for joints pretreated in the two dif-
ferent waters; for example, joints treated 1, 5, and 10 minutes in
each water averaged 80 to 90 percent cohesive-appearing failure.

45
(6531) ;--

40 r-- r--
(5806)
;--
r--

30 r-
'iii (4354)
~
..I
0>
~ """
~ ..
E
~
-
:E :2
...
III 20 r- .:i. ;--
CII
.c (2903)
C/) 10 0 0
0- -5:£ 0 0
.;: 0
.;:
0
.;: .;: .;:
.;:
~!
j 0. 0.
0.
0 ~
0 ~ 0 ~ 0> 0> 0> 0>
0> 0> .5 .5 .5 .5
0> 0>
-
::I!!::
10 (p ~ .5 .5 ~ :sIII =a
51 '0 =a :g '0 =a =a
I~
III III III
(1451)
.'"
III III III III
~
r:: r:: r:: r:: r:: r:: r::
0> r:: :i! :i! :i!
51"" :i! :i! :i! :i! :i!o
c 00
~:i ~ ~ ON
." ." «IN «IJ:
00. ",8, + + + + + + +~ ;g.
Test
~ ~§ ~ f> f> f> f> ~ f>d r'1-
Group 2 3 4 5 6 7 8 9 10 11

Figure L Initial Joint Strengths Obtained with Varying Surface

Oxide Conditions.
 }>
r
C
s:
z
c
s:
t....
Q
Table I. Initial Joint Strength Comparisons. z
-I
~
BOiling H2O :0
m
Treatment Average Average % of Mill Z
G')
Test Adherend* 'IYPe Shear Strength 'IYPe Failure Finish Joint -I
Group Pretreatment Time MPa psi Adhesive Cohesive Strength :::I:
H2 O }>
(min. ) (% ) (%) Z
o
1 N:>ne N:>ne 0 35.28 5120 50 50 100
o
C
:0
2 A N:>ne 0 34.59 5020 60 40 98 }>
3 A-D N:>ne 0 42.58 6180 100 121 !!!
r
4 A-D Deionized 1 39.48 5730 100 112 =i
5 A-D Tap 1 33.97 4930 15 85 96 -<
6 A-D Deionized 5 32.73 4750 100 93
7 A-D Tap 5 40.86 5930 5 95 116
8 A-D Deionized 10 28.18 4090 40 60 80
9 A-D Tap 10 34.45 5000 10 90 98
10 A-D Deionized 60 20.46 2970 100 58
11 A-D Tap 60 24.63 3575 100 70

*A Degrease: B = Caustic Etch: C = Acid Desmudge: D = Wipe and Rinse under Tap Water

(J'I
o
.....
508 J.D. MINFORD

A treatment of 60 minutes in either water resulted in a

significantly lower joint strength accompanied by 100 percent
failure in the adherend oxide, which could be expected if a
significantly thicker oxide layer was present. Joints produced
by deionized water-treated adherends averaged 30.2 MPa (4385 psi),
only 10 percen t below the tap water-treated joints at 33.5 MPa
(4859 psi) • We believe the fact that Wegman used a two-part,
room temperature-curing epoxy as compared to our use of a 177°C
(350°F) curing epoxy may be a major factor in accounting for the
different observations made in the two separate investigations.

Durability Results in Room Temperature Water

The results of soaking the various joints for up to two years

in room temperature deionized water are shown in Table II, and the
durability factors are plotted in Figure 2. It is clear that the
durability of joints with a deoxidized or built-up boehmite oxide
layer surface are superior to mill finish or vapor-degreased
adherend joints. Increasing the overall time to develop the oxide
surface generally favored increased durability, with the 60~inute
thicker-oxide joints showing the highest durability factors.

100 r- -
0
-£
'" 0
-£ ci -
80 r- "'"~
::I
ci '"
,5 -
'" :s
0
:E'" '.:"
"~9. 0 '0
.9u 60 r-
+:r
.r:;Q.
-£ 0 '".:
~
0
-£
~
~ -
CII
dJ~ ci -£ Q. +
LL
Q.
on r:. ~
~.; :S'" r:. +
-
~ ~
:a'"
'0 ,5
:c ~.s '".: :S'" ;-- 0
I!!
:::l 40 I-
UII:
+"
"," ~
'0
'"r:: '"r:: -£ -
0 "ea ~
H
g>;;:
+ ~ on ci
:S'"
+
9:
ea
.---
0+ '"
~ + ~ '0
20 -
.r:; I!!
"'"
h - '" -
'u::"" 0
(;
~ ,5
::;:
0
~ Q. <D
~

Test
Group
n n
2 3 4 5 6 7 8 10
+
~

Figure 2. Joint Durability Results in Room Temperature Water

Exposure.
 »
r
C
5:
Z
c
5:
c-
O
Z
-l
en
-l
::xl
Table II. Durability Results in Room Temperature Water. m
Z
G)
% Average Average -l
I
Test Retained Average '!ype Failure Durability »z
Group Surface Condition Joint Strength Exposure Time Adhesive Cohesive Factor o
(days) (% ) (%) o
C
::xl
1 Mill Finish 35 182 97 3 6.37 »OJ
2 Vapor Degrease 40 126 95 5 5.04 r
3 Caustic Etched 42 532 50 50 22.34 =i
-<
4 Deionized H20 (1 min) 40 630 85 15 25.20
5 Tap H20 (1 min) 35 518 95 5 18.13
6 Deionized H20 (5 min) 63 714 -50 50 44.98
7 Tap H20 (5 min) 40 560 85 15 22.40
8 Deionized H20 (10 min) 69 714 40 60 49.27
10 Deionized H20 (60 min) 123 714 N.A. N.A. 87.82

(]1
o(.0
510 J.D. MINFORD

This is particularly interesting in view of the fact that these

joints showed the lowest initial joint strengths with highest
apparent adhesive-type failure. We must assume, however, that
the lower stress capable of fracturing the 60-minute oxide, as
compared to the thinner oxide joints, was responsible for the low
strength and the apparent adhesive-type failure. The increasing
percentage of apparent adhesive failure after approximately two
years of water soaking observed for the thinner-oxide joints
indicated either these oxides were now fracturing at lower stress
levels near the interface20 or microscopic fragments of water-
weakened adhesive were being broken out of the surface at a lower
stress level than initially had been necessary to fail the bulk
adhesive cohesively.22 In contrast, the long-timed-soaked,
thicker-oxide adhesive joints required an even higher stress to
fail the joints than initially present, so neither oxide nor
adhesive near the interfacial area seemed to be weakening.

Durability Results in 100 Percent Relative Humidity at 52°C (125°F)

The results of soaking in elevated temperature condensing

humidity for a year are summarized in Table III and the durability
factors plotted in Figure 3. The deoxidized adherend joints did
not perform differently from the mill finish or vapor-degreased
joints in this accelerated tropical-type exposure. We know, from
earlier Alcoa investigations, that deoxidizing with chromic-
sulfuric acid instead of caustic has significantly increased
durability using this same adhesive and adherend. All the
boehmite oxide oriented joints showed superior durability wi th
thicker layered oxide generally increasing the durability
performance. Comparisons can be made between the performance of
joints made from adherends treated for 1, 5, 10, and 60 minutes
in both deionized and tap waters, with the result that the former
averaged 36 percent higher durability factors.

Durability Test Results in Salt Ebg Cycle

It should be mentioned that aluminum adherend joints can fail

by two different mechanisms, and the testing for durability in a
salt fog cycle environment relates to that fact. While the
progression of water from the environment into the interfacial
joint area is the mechanism promoting joint failure in most
weathering conditions, there is also the possibility of joint
failure due to gross corrosion of the adherend. Salt water has a
special tendency to produce a pitting type of corrosion on
aluminum, and the effect of a progression of such pitting across
the adherend interface promotes catastrophic joint failures.
 »
r
C
:s::
z
c
:s::
c-
O
Z
-t
C/)
-t
::tJ
Table III. Durability Results in 100 Percent Relative Humidity at 52°C (125°F). m
Z
G)
Test % Average Retained Average Durability -t
J:
Group Surface Condition Joint Strength Exposure Time Factor »
(days) Z
o
1 Mill Finish 22 140 3.08 o
2 Vapor Degrease 37 84 3.11 C
::tJ
3 caustic Etch 38 84 3.19 »
4 Deionized H20 (1 min) 45 294 13.23 S!:!
r
5 Tap H20 (1 min) 31 196 6.08 =i
6 Deionized H20 (5 min) 54 266 14.36 -<
7 Tap H20 (5 min) 43 224 9.63
8 Deionized H20 (10 min) 58 336 19.49
9 Tap H20 (10 min) 73 364 26.57
10 Deionized H20 (60 min) 91 364 33.12
11 Tap H20 (60 min) 54 308 16.63

0'1
512 J.D. MINFORD

35 r- -
- 0
£
30 I- 0 ~'"
'0
c.
~
-
0 £ m
£ e .!;
- '"
~
'0
25 l- e :S'"
0
"
:i!
m -
£ ":J'"
'0
:S'" '0 £
c.
00
:r.
.!;
0
01
u..
E
~
'0
m 0
£
m
.!; ~
~d
N

"
:i!
20 l- .!;
'" ~ -
~
:c
'0
~~ "
:i!
c.
~ "
:i!
.!;
~
m
r-- +
~
'" -
- '" + .!;
I'!! +:1: .!; 0
::J 15 .t:c. + =a '"
r-=- " £
0
£ -
0 ~~
~.:
J:. m
.!;
:i!
'"+
c.
~ e
10 - "'"
~
acf "
:i! r
r--
:S'"
'0 :am'"
.5
-
+'0
.t:
'" c"'"
",,,,
"C
+ m
.!; .5
:~
5c. m8. J:.
5e-
u.
t~ " " -

n rl rl
:i! :i!
:i! := 0
+ "'+
~ y,
Test
Group 2 3 4 5 6 7 8 9 10 11

Figure 3. Joint Durability Results in 100 Percent Relative

Humidity at 52°C (125 OF) .

Several ways to protect against this type of failure have been to

anodize or apply a surface conversion coating to the aluminum or
use special primers. The dramatic effect on improving joint
durabili ty resulting from even a one-minute exposure to boiling
deionized water is clear from the data reviewed in Table IV and
graphed in Figure 4. It must be remembered, however, that
developing a predominantly bohemitic surface in boiling water is
merely another type of conversion coating. Also, one means known
to the aluminum industry to increase the corrosion resistance of
aluminum surfaces has been to build up boehmite oxide surfaces by
exposing aluminum to water in an autoclave, so it is not totally
unexpected to find high resistance to undercutting corrosion in
these joints.

Deoxidizing of the aluminum adherend produces a highly

signif icant increase in resistance of the joint to undercutting
corrosion, which is the major mechanism for joint failure in the
salt fog cycle or a marine atmospheric exposure.10-1 2 This is
shown by the comparative durability factor average of 0.25 for
mill finish or vapor-degreased surface joints as compared to 7.18
for the deoxidized surface joints. The most significant increase
in joint resistance to corrosive salt water appeared when the
 »
r
C
s::
Z
C
s::
L
o
Z
--l
CJ)
--l
Table IV. Durability Results in Salt Fog Cycle. :0
m
Z
Gl
Test % Average Retained Average Durability --l
Group Surface Condition Joint Strength Exposure Time Factor
:r:
(days)
»
z
o
o
1 Mill Finish 17 28 0.48 C
:0
2 Vapor Degrease 0.5 70 0.04 »
OJ
3 Caustic Etched 27 266 7.18 r
4 Deionized H20 (1 min) 78 364 28.39 =i
5 Tap H20 (1 min) 28 168 4.70 -<
6 Deionized H20 (5 min) 77 364 28.03
7 Tap H20 (5 min) 45 266 11.97
8 Deionized H20 (10 min) 89 364 32.40
9 Tap H20 (10 min) 85 364 30.94
10 Deionized H20 (60 min) 103 364 37.49
11 Tap H20 (60 min) 66 364 24.02

01
CAl
514 J.D. MINFORD

deoxidized surface was exposed to boiling deionized water for

only one minute. The retention of 78 percent of initial joint
strength after one year cycling under conditions where untreated
adherend joints fail wi thin 30 days is impressive. Further
increases in the boehmite oxide thickness induced by increasing
the soaking times did not significantly change the durability
potential. It might be mentioned, however, that the retention of
100 percent of initial joint strength after one year of saltwater
cycling by the joints fabricated with adherends pretreated with a
60-minute soak in deionized water represents perfect protection
against the corrosive weathering conditions.

The relative effects of treating in deionized or tap water

can also be considered. It is clear from Table IV and Figure 4
that the durability potential of the boehmite oxide form created
in tap water does not develop as readily as in deionized water.
Ten minutes in tap water is required to achieve similar joint
durability to one-minute treatment in deionized water. If we
compare the average durability of 5-, 10-, and 60-minute coating
joints, we see that the average durability factor for those
treated in deionized water is 46 percent higher.

r--

35 - -
- roo--
30 - roo--
-
t
:::I
roo-- q,
l:
- J ~
Q.
roo--
-
:!!
.:lq, r'6
- +l:
,t;:.Q.
q,
l:
III
.s
-
dl~ Q.

-
u.5
~~ q,
(!.
.s'" q,
"
:iii
II)
0
£
q,
l:
0
£
q,
l:
-
lla: l: l: +
+ ...
ali a ~
.s d ~ a Q.
{!. a Q.
(!.

10 I- ~i .s'"
~ " I
:iii
'"
.5
~ r ;g'"
c:: ~
·0 -
a ~
III III III

.s
~

+ .s
III
.s .5 .s .s
~
" " " ~ -

rl
5 I- j ~ "
:iii :iii" i
S!
i
S!
i
Ii Ii
It
&
~ II)
+ + + + + +

Test
i
....... ~ !-> f> ~ ~ ~ f>
Group 1 2 3 4 5 6 7 8 9 10 11

Figure 4. Joint Durability Results in Salt Fog Cycle Exposure.

ALUMINUM JOINT STRENGTH AND DURABILITY 515

Durability Data from Simultaneous Stress and Humidity Exposure

The significant added effect of simultaneously imposed stress

on weathering exposure with epoxy bonded aluminum joints has been
discussed in many technical articles. 10 ,11, 13-15 The benefit
of better surface pretreatment in extending the survival time of
highly stressed joints under hot humidity soaking conditions has
often been less than predicted from separate water durability
testing. In Table V and Figures 5 and 6 the data for joints with
adherend pretreatments in boiling tap or deionized water have
been summarized. By plotting the data obtained at two different
percentage levels of initial joint strength for each set of
different adherend pretreatments, we can show the relative
effectiveness of the treatments as compared to mill finished,
vapor degreased, and deoxidized adherend joints.

The most significant observation is that either deoxidizing

the adherends or establishing boehmite oxide surfaces enables the
survival time at the same 35 percent of stress level to be
increased over mill finish or vapor-degreased adherend joints by
approximately 1700 times. The survival time potential for all
the boehmite oxide-type surfaces was, however, equal to or better
than that obtained with deoxidized-only adherends. Finally, the
comparison of the plots in Figures 5 and 6 clearly show that the
tap water-induced boehmite oxide surface joints show increased
joint durability over those induced in deionized water. This is
a reversal from all the previous test results in room temperature
and hot humidity deionized water soaking and corrosive saltwater
testing using nonstressed joints. Comparing the actual durability
factors from Table V, we find that, at the 35 percent stressing
level, the average durability factor for the tap water-treated
adherend joints was 46 percent higher than for the corresponding
deionized water-treated adherend joints.

This result would support the conclusion that the oxide

formed in deionized water permitted jOints with more water and
pitting corrosion resistance, which argues for better wetting and
spreading of the adhesive. 14 However, the oxide deposited in
tap water had the higher resistance to simultaneous stress and
water soaking conditions.

Comparisons with Other Aluminum Surface Pretreatments

Because Alcoa has employed the same aluminum alloy and

e1evated-temperature-curing EC-2086 epoxy adhesive in all its
publications since 1972, it is possible to draw a direct
comparison of the results of this investigation as regards
comparative joint durabi1ities afforded by different aluminum
surface pretreatments. These results are shown in Table VI.
 UI

en

Table V. Durability Data from Simultaneous Stress and Humidity Soak Exposure.

'D:!st Average Joint Stress/Humidity

Group Surface Condition " Stress Fa ilure Time Durability Factors
(days)

1 Mill Finish 35, 15 0.005, 1.12 0.009, 0.859

1 Mill Finish 7.8, 4.9 118, 123 47.20, 30.75
2 Vapor Degreased 35, 15 0.009, 0.519 0.015, 0.391
2 Vapor Degreased 8, 5 104, 120 41.6, 30,00
3 caustic Etch 35, 15 8.5, 53.0 18.39, 49.13
4 De ionized H20 (1 min) 35, 15 9.5, 45.3 19 • 05, 38 • 9 4
5 Tap H20 (1 min) 35, 15 20.0, 115 34.50, 67.85
6 Deionized H20 (5 min) 35, 15 18.5, 62.5 30.75, 44.5
7 Tap H20 (5 min) 35, 15 17.0, 124 35.28, 110.24
8 Deionized H20 (10 min) 35, 15 18.5, 122 26.47, 74.79
9 Tap H20 (10 min) 35, 15 23.0, 135 40.25, 101.25
10 Deionized H20 (60 min) 35, 15 33.5, 135 34.81, 61.43
11 Tap H20 (60 min) 35, 15 42.0, 147 52.54, 78.79

c....
0
s:
Z
"T1
0
:II
0
 Table VI. Comparative Durability Performance with Varying Aluminum Surface Pretreatments »
r
C
EXposure Conditions* s::
A B C D
z
c
Surface Initial EXposure Strength Exposure Strength Exposure Strength Survival s::
c-
Pretreatment Strength Time Retention Time Retention Time Retention Stress Time O
(psi) (days) (psi) (days) (psi) (days) (psi) (lbs.) (min. ) Z
--l
(J)
Chromic- 5330 730 3624 365 885 N.A. N.A. 800 100,000 --l
:II
Surfuric (49.64) 3 (6.06) m
Z
Deoxidize l G)
--l
:::c
Chromic- 5430 364 1629 »
Surfuric (10.92) z
o
Deoxidize 2 o
C
:D
Ch romic 1lc id 5513 730 3198 365 3363 365 4521 900 100,000 »
~
Anodize (42.34) (22.20) (29.93) r
=i
-<
Phosphoric 1lc id 6480 730 4212 365 3551 365 5119 1050 100,000
Anodize (Boeing) (47.45) (20.08) (28.94)

5 Min. D.I. H20 4750 714 2993 266 2565 266 2138 710 100,000
(44.98) (14.36) (11.97)

60 Min. D. I. H20 2970 714 3653 364 2703 364 3059 700 100,000
(87.82) (33.12) (37.49)

*EXposure Condition: 1 Vapor degrease plus 5 min. at 180°F in

A = Room Temperature Deionized Water. chromic-sulfuric.
B 100% Relative Humidity at 52°C. 2 Ridoline 53 alkaline cleaner plus 5 min. at 180°F
C Salt Fbg Cycle. in chromic-sulfuric.
D Stress/Humidity. 3 = All figures in parentheses are calculated U1
durability factors based on the data shown in -..J
this table.
518 J.D. MINFORD

2000 Caustic
Etched-
• 1 Min. in Boiling 0.1. H20
• 5 Min. in Boiling 0.1. H20
o 10 Min. in Boiling 0.1. H 2 0
'iii 1500
Q. K 60 Min. in Boiling 0.1. H 2 0
,!:
"ai
~
..J
<II 1000
~ Vapor J
iii Oegreased

500 6
eo
oe

10 2 10 3 10 4 10 5 10 6
Joint Survival Time in Minutes

Figure 5. Stress-Endurance Curves for Joints Made to Boehrni tic

Oxide Surfaces Generated in Boiling Deionized Water.

Caustic
2000 Etched-
• 1 Min. in Boiling Tap H20
• 5 Min. in Boiling Tap H20
o 10 Min. in Boiling Tap H20
'iii 1500 II 60 Min. in Boiling Tap H 2 0
Q.
.!:
"ai
>
CII
..J
1000
~ Vapor -...i
iii Oegreased

500
eo

oe

10 2 10 3 10 4 10 5 10 6
Joint Survival Time in Minutes

Figure 6. Stress-Endurance Curves for Joints Made to Boehrnitic

Oxide Surfaces Generated in Boiling Tap Water.
ALUMINUM JOINT STRENGTH AND DURABILITY 519

Comparative data involving the deoxidized and anodized surface

joints have been taken largely from an Alcoa publication of 1978. 15
'l1le joints with boehmite oxide generated after 60 minutes in
deionized water were selected for comparison because of their
generally superior performance in most exposure conditions. A
performance comparison is also offered with joints prepared after
only a five-minute treatment time in deionized water as a more
economical and fast procedure for general manufacturing.

The boehmitic oxide surface joints perform very favorably

overall as compared with the higher state-of-the-art surface
pretreatments whether water soaking, corrosive water cycling, or
stress and hot humidity exposure environments are considered. We
do not propose that these relatively simple-to-prepare, built-up
oxide joints are the equal in all respects of the more expensive
and difficult-to-control aerospace procedures, but their perform-
ances are creditable enough for strong consideration of acceptance.

DISCUSSION

When considering how adhesively bonded aluminum joints fail,

consideration must be given to the initial surface condition of
the adherends and how the adhesive interacts with this environ-
ment. A considerable amount of work in recent years has been
published or announced about the micromorphology of aluminum
surface oxides and the stability of those oxides to attack by
moisture. 16 ,17 Additional investigations have focused on ~o~
the polymeric adhesive can interact with this mozphology .18 2
The emphasis in these papers has primarily been focused on the so-
called FPL etch or PAA (phosphoric acid anodize) aluminum surface
pretreatment procedures. The particular oxide surface generated
in the present investigation has not been so specifically studied.
The FPL etched surface condition has been described by Bijlmer 22
as a nmicroetch-type pitted surfacen and is almost uniquely
produced in only a few ratio combinations of the total matrix of
many possible different combinations of chromic and sulfuric acid
that could be considered. The caustic-type deoxidizing used as
the pretreatment for generating boehmite films in this present
investigation is a much more vigorous, gas-evolving, aggressive
attack of the aluminum surface, leaving much larger, low-power-
magnification-viewable craters in the treated surface and
generating considerable smut that is readily removed in a dilute
acid rinse. The oxide generated by the boiling waters is built up
over this pock-marked landscape. At the time these experiments
were conducted many years ago, there was no predisposition to
analyze surfaces as is commonplace today. 1\ccordingly, we must
confess to having left that part of this study uninvestigated.
520 J.D. MINFORD

Because of the remarkable durability results we have obtained

(see Table VI), we appreciate that such a study should be
conducted and compared with the observations already made on
other prepared surfaces as noted above.

If the theory of aluminum bond failure due to generation of a

weak boundary layer of aluminum hydroxide is considered ,23 we
must assume from the hot humidity data in Table VI that most of
our boehmite converted surfaces are less sensitive to that
conversion to hydroxide than found even with FPL-type etched
surfaces. Or we may consider the weak boundary layer in the
polymer theory that the surface morphology influences the
orientation of the polymer or prepolymer molecules during the
adsorption phase as a major factor in establishing resistance to
joint failure under adverse weathering conditions. 22 The
comparatively good durability found for these boehmitic oxide
joints under all the weathering and stress/weathering conditions
would indicate we are dealing with an oxide surface that has not
significantly altered the normal cross-linking of the polymer in
the vicinity of oxide as proposed in this weak boundary adhesive
layer theory.

CONCLUSIONS

1. Significant improvement in joint durability under stressed

and nonstressed weathering conditions can result from
pretreating 6061-T6 aluminum to produce varying boehmitic
oxide surface films as compared to mill finish, vapor
degreased, or alkaline-deoxidized surface adherend joints.

2. FUrther distinctions in joint durability result from using

boiling deionized water or New Kensington tap water as well
as varying exposure times in either.

3. Thicker boehmitic oxide conditions generally produce higher

durability responses in room temperature water or hot
humidity exposure conditions with unstressed joints.

4. Oxide generated in the presence of deionized water produced

the higher durability performances when using unstressed
joints in either water soaking or corrosive saltwater cycling
exposure conditions.

5. With simultaneous stressing and hot humidity exposure testing

conditions present, the oxide produced in boiling tap water
showed the higher joint durability potential.
ALUMINUM JOINT STRENGTH AND DURABILITY 521

REFERENCES

1. R. F. Wegman, W. M. Bodnar, M. J. Bodnar and M. J. Barbarisi,

Technical Report 3495, Feltman Research Laboratories,
Picatinny Arsenal, Dover, NJ, May (1967).
2. R. F. Wegmen, Adhesives Age, 10(1), 20, January (1967).
3. B. B. Bowen, R. E. Herfert and K. C. Wu, AFML Contract No.
F33615-74-C-5-27, Northrop Corp., Hawthorne, CA, March (1975).
4. K. C. Wu and B. B. Bowen, Interim AFML Report IR-854-5 (I),
Contract No. F33615-75-C-5083, Northrop Corp., Hawthorne,CA,
June-November (1975).
5. w. J. Russell and E. A. Garnis, SAMPE Quart. 7(3), 5, April
(1976) •
6. R. A. Harrington and H. R. Nelson, Amer. Inst. Min. Met. Eng.,
Tech. Publ. 1158 (1940).
7. J. M. Bryan, J. Soc. Chem. Ind., 69, 169 (1950).
8. D. Altenpohl, Aluminium, 29, 361 (1953).
9. R. K. Hart, Trans. FaradaY-Soc., 53, 1020 (1957).
10. J. D. Minford, Metals Eng. Quart.~ovember (1972).
11. J. D. Minford, in "Treatise on Adhesion and Adhesives", R. L.
Patrick, Editor, Vol. 3, Chap. 2, Marcel Dekker, New York,
1973.
12. J. D. Minford, Adhesives Age, 17(7), 24 (1974).
13. J. D. Minford, J. Appl. Polym. Symp., 32, 91 (1977).
14. J. D. Minford, in "Treatise on Adhesion and Adhesives", R. L.
Patrick, Editor, Vol. 5, Chap. 3, Marcel Dekker, New York,
1981.
15. J. D. Minford, SAMPE Quart., 18-27, July (1978).
16. J. D. Venables et al., Appl. Surf. Sci., 3, 88 (1979).
17. J. S. Ahearn, G. D. Davis, T. S. Sun and J. D. Venables, in
"Adhesion Aspects of Polymeric Coatings", K. L. Mittal, Edi-
tor, pp. 281-299, Plenum Press, April, 1983.
18. A. V. Pocius, in "Adhesion Aspects of Polymeric Coatings", K.
L. Mittal, Editor, pp. 173-192, Plenum Press, April, 1983.
19. K. K. Knock and M. C. Locke, in "Adhesion Aspects of Polymeric
Coatings", K. L. Mittal, Editor, pp. 301-318, Plenum Press,
April, 1983.
20. D. E. Packham, in "Adhesion Aspects of Polymeric Coatings", K.
L. Mittal, Editor, pp. 19-44, Plenum Press, April, 1983.
21. W. Brockmann, in "Adhesion Aspects of Polymeric Coatings", K.
L. Mittal, Editor, pp. 265-280, Plenum Press, April, 1983.
22. P. F. A. Bijlmer, J. Adhesion, 5, 319 (1973).
23. J. D. Venables et al., in ·;Proc. 12th Natl. SAMPE Tech.
Conf.", Seattle, WA, p. 909(1980).
EPOXY ADHESION TO COPPER, PART II: ELECTROCHEMICAL PRETREATMENT

Jae M. Park and James P. Bell

Department of Chemical Engineering and

Institute of Materials Science
University of Connecticut
Storrs, Connecticut 06268

Electrochemical methods of pretreating copper

surfaces for adhesive bonding were studied. The
objective of the electrochemical pretreatment was to
control the oxide layer under the copper-coupling agent
complex film, since this layer was found to be the locus
of failure of copper/epoxy bonds aged in boiling water.
Both cathodic and anodic pretreatments resulted in im-
proved durability of the jOint after exposure. Experi-
ments were also performed to understand the mechanism
of weakening of copper/epoxy bonds in boiling water.

INTRODUCTION

We previously reported that the durability of copper/epoxy

bonds was enhanced by pretreating the copper surface with substi-
tuted benzotriazole (BTA) coupling agents. l In the previous
studies several BTA coupling agents capable of bonding with both
copper and epoxy adhesive were prepared and the shear strength of
treated adhesive joints was evaluated by a torsional test method.
It was found that joints treated with 5-amino-benzotriazole (5-ABTA)
were 500% stronger than a similarly prepared control after a long
term immersion test in boiling water. Our previous work also
showed that the copper-BTA (CuBTA) complex layer remained on the
epoxy adhesive surface after the jOint was broken. This result
clearly indicated that the locus of failure of the treated
copper/epoxy joint was the oxide layer under the complex film
(Figure 1). The existence of an oxide layer under the complex

523
 -- I
524 J.M. PARK AND J.P. BELL

; :: CuBTA Epoxy
Copper Oxide-
_ ~: Complex Adhesive

Fig. 1. BTA treated copper/epoxy bond.

film ~s previously reported by Ogle and Poling. 2 Mansfeld

at al. explained that an unreacted oxide layer remains under the
complex film; the oxide layer is the substrate for the CuBTA
complex formation and provides the copper ions for the reaction
with BTA. In other words, the oxide layer is partially replaced
with CuBTA complex. Chadwick and Hashemi also found that the CuBTA
layer formed on a cathodically polarized copper surface (oxide free
copper) was too thin to be detected by mUltiple reflection infrared
spectroscopy.4

The objective of pretreating a copper surface electrochemically

is to control the oxide layer under the complex film, which seems
to be weaker than any other part of the bond. According to the
thermodynamic phase diagram of copper in water,S the isothermal
state of copper oxidation in water changes with applied potential
and the solution pH. Copper tends to remain as cuprous oxide at
higher pH (alkaline) and as cuprous ions at lower pH (acidic) when
no potential is applied. If a cathodic potential is applied, the
copper surface is reduced and remains as elemental copper.

Our cathodic pretreatment of copper in an aqueous medium

containing coupling agent consists of two steps, i.e., 1) reduction
of the surface oxide by a cathodic potential and 2) complex forma-
tion after the potential is removed. When an oxide-covered copper
surface is dipped in an aqueous BTA solution the complex formation
starts at the top layer of the oxide surface, but the entire oxide
layer is not completely changed into CuBTA complex. 3 By applying
a cathodic potential, the oxide layer is removed almost completely.
It is, therefore, possible to have the complex layer grow very near
to the elemental copper surface.

The complex formation reaction is a controlled corrosion

reaction. During the complex formation step two reactions are
expected. First, the elemental copper is oxidized into its ion
forms. Secondly, some of the oxide is transformed to CuBTA. The
resulting layer is possibly a mixture of oxide and the complex,
whose composition depends on the relative rates of the two
EPOXY ADHESION TO COPPER 525

reactions. During the complex formation step, it is possible to

apply a small anodic potential in order to increase the dissolution
rate of copper. It was found that a thick CuBTA complex layer
(several thousands of angstroms thick) could be formed by applying
an anodic potential to copper in an aqueous solution of BTA
containing sodium chloride. 2 It was also found that the complex
film grown anodically protected the copper from corrosion more
effectively than the naturally grown film.

In order to understand the importance of corrosion effects

(resulting in interfacial instability) of copper on the durability
of copper/epoxy bonds, an electrochemical aging experiment was
also performed. The glued joints were placed in boiling water in
an electrochemical cell. Various potentials were applied between
the joints and a counter-electrode. The change of joint strength
with the applied potential was examined.

EXPERIMENTAL

The torsional joint geometry designed by Lin and Bel1 5 was

used for the present work. This type of joint was used success-
fully in the previous work 6 and proved to yield more useful infor-
mation on the bond strength than other current methods. 7 The two
parts of the joint were polished with 0.5 micron alumina slurry on
a rotating micro-polishing cloth. The polished joints were washed
with a copious amount of distilled water and dried with reagent
grade methanol. It was important to keep the gap between two parts
of the joint constant (0.254 mm). In order to keep the gap constant
and to prevent the edge from becoming rounded by repeated.polishing,
the joints were remachined on a lathe after they were used three
times. A weight of 44.0 ± 0.1 mg of the adhesive was applied on
the rim-part of the joint (Figure 3) and the flat-part was placed
over it. The joints were tested in shear on an InstronR tensile
test machine equipped with a torsion device as previously
described. 5 ,7

The adhesive used for the present work was a mixture of 4

parts of dicyandiamide and 100 parts Epon 828 R (Shell Chemical Co.).
Reagent grade dicyandiamide (Aldrich Chemical Co.) was used with-
out purification. Dicyandiamide was mixed with li~uid Epon 828
by ball-mill mixing followed by mechanical mixing. The adhesive
was cured at 175°C for 2 hours. Some of the experiments were
performed with Pl08 R curing agent (Shell Chemical Co.). Mixing
solid Pl08 curing agent with Epon 828 was done by mechanical mixing.

The electrochemical cell used for the pretreatment of the

polished joints is shown in Figure 2. Four holes were made in
the cylinder wall of a 350 ml pyrexR Buchner Funnel to fit the
526 J.M. PARK AND J.P. BELL

joints. The funnel contained a F grade glass filter. The two

stopcocks were prepared to bubble gas and to drain liquid. A
rubber seal was used between the hole and the joint to prevent
leaking of the electrolyte. The platinum anode was placed at the
center of the Plexiglas ® cover. Another hole was prepared in the
cover for the reference electrode (saturated calomel electrode,
not shown in Figure 2). For cathodic polarization, a Model 6204
Power Supply by Harrison Laboratories was used. For anodic
polarization, the mini-potentiostat designed by Greene et al~ was
used. A Simpson 260 multimeter and a Keithley 606 Digital Electro-
meter were used to measure the potential and the current, respec-
tively. 0.3% aqueous solution of potassium chloride or sodium
sulfate was used as the electrolyte. In order to minimize the
contact time of the wet joints with the air, the joints were washed
with a copious amount of distilled water while the potential was
applied. The liquid drain was opened while fresh distilled water
was poured into the cell. The washed joints were dried with
reagent grade methanol. It should be noted that the precipitation
of Cu-ABTA complex was observed after several minutes during the
anodic polarization. The anodized surface was covered with the
precipitate when it was removed from the solution. The treated
joints were placed in an ultrasonic cleaner, cleaned for 5 minutes
to remove the precipitate and dried.

Pt electrode

, .
"
':
"
:f:'
,.
polished
joint

rubber seal

glass filter

Fig. 2 Electrochemical cell for joint p£etreatment.

EPOXY ADHESION TO COPPER 527

carbon steel mount

\
~
~
brass bolt
adhesive 304ss spacer

Fig. 3 Demountable joint for aging test.

For the electrochemical aging test, the joint assembly was

modified as shown in Figure 4. The geometry of the modified joints
was the same as previously, except that the two parts in contact
with the adhesive were mounted on steel supports and could be
removed to facilitate exposure to boiling water, etc. The joint
was reassembled before strength testing.

potentiostat
R W
8
reference
electrode 0 . 125

reflux
conde nser -,

[
L--
-----./
cop per wire

teflo n cover

)
- Plexiglas

-cr=
counte r-
·. J-'- upper
fixture

electrod e
(304$$ ) Ii Quid level
/'

·. glued joints
V
- , ' (demountable)
, [/

- :: - I
,

.
::

-
Plexiglas .... heating tape
lower
·.:, 1
If ~
fixture ~
·· , .

Fig. 4 Polarization cell for aging test.

528 J.M. PARK AND J.P. BELL

The glued joints were stacked at the center of an electro-

chemical cell, shown in Figure 4. The cylindrical cell was heated
by a flexible heating tape. The cover of the cell was made of
Tef1onR• Four holes were prepared for the working electrode
(joints), reference electrode, counter-electrode, and ref1uxing
condenser respectively. A 2 1/2" dia. x 3" long piece of stainless
steel (304) tubing was used as the counter-electrode. The
potentiostat designed by Greene et a1 8 was used as the power supply.
By using the electrochemical cell and the power supply, the potential
between the joints and the counter-electrode could be kept constant
within a variation range of less than 2.0%. A 1.0% aqueous sodium
sulfate solution was used as the electrolyte. The joints were
aged in the electrolyte at 100°C for 24 hours while constant
potential was applied. After aging, the joints were cooled to
room temperature and washed with distilled water. The joints were
tissue dried and reassembled for testing.

RESULTS AND DISCUSSION

It was pointed out in the Introduction section that the

oxidation state of copper in water changes with applied potential
and solution pH. The solution pH is, therefore, important during
the complex formation period after the cathodic reduction. Results
summarized in Table I, show that an acidic solution was less
effective for strong adhesion. The reason seemed to be the differ-
ence in thicknesses among the CuBTA layers. Ogle and Poling
defined a limiting pH for BTA solutions, associated with the thick-
ness of CuBTA film formed o~ the copper surface. 2 Thin films with
a limiting thickness of 60 A were formed in solutions ~ith a pH
higher than the limiting pH. Thick films, about 4000 A in thick-
ness, were formed in solutions with a pH lower than the limiting
pH. They suggested that the limiting pH was between 3.5 and 4.0.
It was also found that a large portion of a thick film could be
easily wiped off and that thin films showed better corrosion
protection than thick films. The experiments on the effects of
solution pH on the joint strength were not repeated with 5-ABTA.
The amino group of 5-ABTA was expected to form -NRj in acidic
solution, which would not react with the epoxy adhesive.

Different potentials were applied between the platinum anode

and the polished joints during the polarization process. No
significant change in the joint strength was observed among the
joints pretreated with different potentials (Table II). The
results are consistent with the fact that the cathodic process,
Cu+ + e- = Cu, is rapid. When the potential is increased, the
current also increased. The time required to reduce the oxide
EPOXY ADHESION TO COPPER 529

layer is shortened when a higher current flows between the

electrodes. Assuming the diffusion effect around the electrode is
negligible, the time required for the reduction of the oxide was
calculated according the Faraday's law. It was also assumed that
cuprous oxide has a simple cubic structure with a lattice spacing
of 2.0 A. The result showed that only 2.0 seconds are required
for the reduction of the polished joint surface when the current
is 10 rnA. When a potential of 1.0 volt was applied, the measured
current was 2.0 rnA. A reduction time of 5 minutes seemed quite
enough for a complete reduction even when a potential of only
1.0 volt was applied.

The concentration of BTA may be important if the CuBTA complex-

forming reaction is controlled by the diffusion of BTA into the
complex layer. There was no statistically significant difference
between joints treated in 0.02M and 1.25xlO- 5M solutions (Table
III). It seems reasonable to conclude that the effect of BTA
concentration was not appreciable.

It was also pointed out in the Introduction that the substrate

for the complex formation reaction is the oxide layer, which
provides the copper ions for the reaction. If it is true that the
substrate for the reaction is the oxide, the thickness and structure
of the complex film should be affected by the amount of oxygen
dissolved in the solution. Table IV shows the importance of oxygen
during the pretreatment. The joints treated in a nitrogen-bubbled
solution were weaker than those treated in an air-bubbled solution.
This strengthens the evidence that BTA compounds react with copper
ions, not with the elemental copper. The requirement of oxygen
reduction to meet the charge neutrality of the oxidation reaction
of copper necessitates oxygen in the complex forming reaction.

Table I. Effects of Solution pH on Joint Strength.

pH 3.1* pH 5.9 pH 11.0",,0,

Breakage Force (lbs)*** 264.0 288.8 282.8

Standard Deviation 14.6 19.3 13.2

Electrolyte: 0.3% aqueous KCl solution prepared wi~~ double-

distilled water, containing lxlO M BTA
'Polarization: Cathodic 5 volts for 10 min. and 10 min. without
potential (with air-bubbling)
* pH was adjusted with conc. H]P04
** pH was adjusted with 10% NaOH solution
*** After 24 hours immersion in boiling water
530 J.M. PARK AND J.P. BELL

Table II. Effect of Different Cathodic Potentials for Reduction

on Joint Strength.

Electrolyte: 0.02M aq. BTA solution containing 0.3% KCl

(neutral)

Polarization: Various cathodic potentials for 5 min. and

10 min. without potential (with air-bubbling)

Breakage Force (lbs)*

Joint No. 1 volt 5 volt 10 volt

1 279 289 284

2 286 288 302
3 298 300 282
4 296 304 293

Avg. 289.8 295.3 290.3

S.D. 8.9 8.0 9.2

*After 24 hours immersion in boiling water.

Table III. Effect of BTA Concentrations on Joint Strength.

Electrolyte: 0.3% aq. KCl solution containing different

amounts of BTA

Polarization: Cathodic 5 volts for 10 min. and 10 min.

without potential (with air-bubbling)

Breakage Force (lbs)*

Joint No. 0.02M l.25 x 10-5M

1 287 306
2 295 273
3 292 322
4 266 287

Avg. 285.0 297.0

S.D. 13.1 21.0

*After 24 hours immersion in boiling water.

EPOXY ADHESION TO COPPER 531

Most of the experiments reported thus far were performed with

BTA. Table V shows a comparison of the results of joint tests per-
formed with BTA and 5-ABTA. The results show that there is no
significant difference in joint strengths between BTA and 5-ABTA
treatments when the joints are treated electrochemically. The reason
seemed to be the susceptibility of 5-ABTA to air oxidation. When an
aromatic primary amine is exposed to the air, it acquires a dark
red color. The color is developed as a consequence of atmospheric
oxidation. When an aromatic amine is allowed to react with an
oxidizing agent stronger than air, a variety of reactions, some of
them very complex, occur9. The same phenomenon was observed with
5-ABTA. When an aqueous 5-ABTA solution was exposed to the air,
the solution changed to red in a few hours. Because the experiments
were performed with air-bubbling during the pretreatments, an
extensive oxidation of 5-ABTA could be expected, although the change
in color could not be observed because of the low concentration of
5-ABTA. When all of the amino groups of 5-ABTA were oxidized, the
effect of 5-ABTA treatment would be no better than BTA treatment.

A set of torsional joint tests were conducted to examine the

effect of complex formation time on the joint strength. After
cathodic reduction, the joints were allowed to remain in the
solution for different lengths of time for complex formation. It
was expected that a thicker complex film would be formed on the
copper surface at longer reaction time. The results in Table VI,
showed that the joints reacted with 5-ABTA for 15 minutes were
strongest; longer immersion time did not yield stronger bonding.
The rate of oxide film growth on a copper surface in water was
studied by Kruger. lO It was found that very rapid oxidation took
place initially in unstirred water i~ equilibrium with 1 atm. oxygen,
followed by a leveling off at 65-95 A. This oxidation behavior
was similar to that observed by Rhodin ll for a copper surface
oxidized at room temperature in gaseous oxygen. The only difference
was the limiting thickness, which was greater for the oxidation in
water. Another important feature of Kruger's study was that the
relative phase retardation changed rapidly after 90 minutes. The
change in this optical parameter was associated with the formation
of a new component in the oxide film. Kruger concluded that the
change was due to the formation of cupric oxide. A loosely held
cupric oxide layer, which could be wiped off easily, was observed
after 18 hours immersion. The rate of oxide formation in an
aqueous solution of 5-ABTA containing sodium sulfate should not be
comparable to Kruger's results. However, the formation of cupric
ions explain the lower joint strength at longer immersion time.
It is possible that a loosely held Cu(II)-ABTA complex was formed
on the copper surface in the same manner as the cupric oxide
formation in pure water.
532 J.M. PARK AND J.P. BELL

Table IV. Effects of Oxygen Content of Solution on Joint Strength.

Electrolyte: 0.3% KC1 aq. solution prepared with doub1e-

distilled water, containing 0.02M BTA

Polarization: Cathodic 5 volts for 10 min. and 10 min.

without potential (with a1r- or nitrogen-
bubbling)

Breakage Force (lbs)*

Joint No. air-bubbling NZ-bubbling

1 287 260
2 295 245
3 292 240
4 266 250

Avg. 285.0 248.8

S.D. 11.3 7.4

*After 24 hours immersion in boiling water.

Table V. Effect of 5-ABTA Treatment on Joint Strength.

Electrolyte: 0.3% aq. KCl solution containing 1 x 10-4M

BTA (pH 5.9) or 5-ABTA (pH 6.1)

Polarization: Cathodic 5 volts for 10 min. and 10 min.

without potential (with air-bubbling)

Breakage Force (lbs)*

Joint No. BTA 5-ABTA

·1 276 291
2 264 310
3 310 277
4 305 284

Avg. 288.8 290.5

S.D. 19.3 12.3

*After 24 hours immersion in boiling water.

EPOXY ADHESION TO COPPER 533

Tables VII, VIII, and IX and Figure 5 are summaries of long

term test data from electrochemically treated joints. An anodic
potential of -0.1 volt vs saturated calomel electrode (SCE) was
selected because the corrosion potential of BTA treated copper
was found to be between -0.2 and -0.3 volt vs SCE. 12 A potential
of -0.1 volt seems high enough to enhance the dissolution rate of
copper. However, the rate of copper dissolution, which is directly
related to the rate of complex formation, may change the properties
(i.e. crystallite size, etc.) of the resulting complex film.
Little difference in dry strength was observed among the treatment
methods (dipping, cathodic, and anodic pretreatments). After 24
hours immersion, both the cathodic and the anodic treatments showed
about 12% increase in strength over the dipping treatment. In
immersion up to 100 hours, the cathodically treated joints were
perhaps slightly stronger than those treated by dipping. The
long immersion tests (840 hours immersion in boiling water), showed
that all of the 5-ABTA treated joints were much stronger than the
untreated control. However, after extended water immersion, any
differences between electrochemical treatments and the dipping
treatment were indistinguishable.

The importance of the corrosion effects on the durability of

a copper-epoxy bond was also studied by an electrochemical aging
experiment. The glued joints were placed in boiling water in the

Table VI. Effects of Reaction Time after Cathodic Polarization

on Joint Strength.
Electrolyte: 1 x 10-4M aq. 5-ABTA solution containing 0.3%
Na 2 S0 4

Polarization: Cathodic 5 volts for 10 min. and for various

lengths of time without potential (without
air~bubbling)

Breakase Force (lbs)*

Joint No. 5 min. 15 min. 30 min. 60 min.

1 238 278 273 278

2 258 263 253 272
3 235 284 247 251
4 269 271 258 226

Avg. 250.0 274.0 257.8 256.8

S.D. 14.1 7.8 9.6 10.4

*After 24 hours immersion in boiling water.

534 J.M. PARK AND J.P. BEll

Table VII. Effects of Cathodic Polarization on Joint Strength.

Electrolyte: 1 x 10-4M aq. 5-ABTA solution containing

0.3% Na 2S0 4 (neutral)

Polarization: Cathodic 10 volts for 10 min. and 10 min.

without potential (with air-bubbling)

Adhesive: 4 parts dicy + 100 parts Epon 828

Curing: 2 hours at 175°C

Breakage Force (lbs)

Joint No. Dry 24 48 96*

1 354 294 283 246

2 366 291 256 279
3 339 288 275 255
4 348 279 261 267

Avg. 351.8 288.0 268.8 261.8

S.D. 9.8 5.6 10.8 12.4

*Hours in boiling water.

electrochemical cell of Figure 4. Various potentials were applied

between the counter-electrode and the glued joints. The changes in
joint strength as a function of applied potential were examined.
The results of the joint tests are summarized in Figure 6.
The joint strength showed a maximum at 0.0 volt vs SCE. This
potential seemed to be within the potential region where cuprous
oxide is stable. The low strength at -1.0 volt indicates that the
oxide layer or the complex layer was almost completely reduced to
the elemental copper. The decreased strength at +0.1 volt indicat-
ed that the potential was high enough to oxidize cuprous oxide into
cupric oxide. A potential higher than 0.1 volt was not tried
because an extensive dissolution of copper was observed within an
hour.

It was pointed out earlier that an oxide layer exists under

the complex layer. Because the joints were polished and pretreated
at room temperature 1 the oxide layer is primarily cuprous oxide,
as found by Rhodin11 and Kruger. 10 When a cathodic potential is
applied, the cuprous oxide layer tends to be reduced to the
elemental copper. The reduction of cuprous oxide into the
elemental copper means that a phase transformation of copper occurs
at one part of the bond, which eventually leads to a change in
EPOXY ADHESION TO COPPER 535

specific volume. The volume change will result in the weakening

of joint strength. When a positive potential which is high enough
to oxidize cuprous oxide to cupric oxide is applied, we can expect
another phase transformation leading to the weakening of joint
strength.

The role of oxygen and water in the weakening of copper/epoxy

bonds was confirmed by the following experiments. One set of
joints was aged in boiling water purged with nitrogen for 24 hours.
The joints boiled in the nitrogen-purged water were much stronger
than those boiled in the air-saturated water (Table X). Some
joints were heated in an oven at 100°C for 24 hours. A large
difference in strength was observed between the joints aged in
boiling water and the joints aged in the air (Table XI). The
joints aged in hot air were stronger than the unaged control,
showing that the loss of strength in boiling water is not a thermal
aging effect. The increased strength of the aged sample can be

...... 300
'"
.0

.<:
01
c: ';
Q)
.... ~

-
(/)

c:
-,
0
i
120
'"Co
c
•
a;
~

200
c
40 ..,
0

-0

48 96 800
Immersion Time in Boiling Water(hours)

Fig. 5 Effects of electrochemical treatments on joint strength.

536 J.M. PARK AND J.P. BELL

explained in two ways: First, the frozen-in stress built up during

the cooling period may have been relieved while the joints were
aged in the oven. Second, the curing reaction of the adhesive may
not have been completed in 2 hours at 175°C, and the reaction may
have continued as the postcure of the adhesive. The results of
two experiments, immersion in nitrogen-purged water and aging in
100°C air, showed that both oxygen and water are required for the
weakening of a copper/epoxy bond. The results also indicated the
importance of substrate corrosion on the joint durability. When
a copper/epoxy joint is exposed to air saturated water, oxidation
of copper, Cu+ = Cu++ + e- and Cu = Cu++ + 2e-, seems to occur as
an anodic reaction. The reduction of oxygen, 02 + 4~ + 2e- =
2H20, seems to be required as the cathodic reaction to meet the
requirement of electrical neutrality.

CONCLUSIONS

Electrochemical methods of pretreating copper surfaces for

adhesive bonding have been studied. The durability of a

Table VIII. Effects of Anodizing Copper Surfaces on the Joint

Strength.

Electrolyte: 1 x 10-4M aq. 5-ABTA solution containing

0.3% Na 2 S0 4 (neutral)

Polarization: Cathodic 10 volts for 10 min. and anodic

-0.1 volt vs. SCE for 15 min. (without air-
bubbling)

Adhesive: 4 parts dicy + 100 parts Epon 828

Curing: 2 hours at 175°C

Breakage Force (lbs)

Joint No. Dry 24 48 96*

1 347 296 281 273

2 352 282 252 264
3 338 294 274 241
4 355 300 259 258

Avg. 348.0 293.0 266.5 259.0

S.D. 6.4 6.7 11.5 11.7

*Hours in boiling water

EPOXY ADHESION TO COPPER 537

Table IX. Long-term Immersion Test.

Pretreatment

5-ABTA treated: Dipped in a 2 x 10 -4M aq. 5-ABTA solution

for 10 min. at room temp.

Cathodically Polarized: Cathodic 10 volts in a 1 x 10-4M

aq. solution containing 0.3% Na 2 S0 4
and another 10 min. without potent~al
(with air-bubbling)

Anodized: Cathodic 10 volts for 10 min. and anodic -0.1 volt

vs. SC~4for 15 min. (without air-bubbling) in a
1 x 10 M aq. 5-ABTA solution at room temperature.

Adhesive: 4 parts dicy + 100 parts Epon 828

Curing: 2 hours at 175°C

Breakage Force (lbs)*

Joint No. Control Dipped Cathodic Anodic

1 18 109 158 122

2 35 171 144 151
3 43 181 102 98

Avg. 32.0 153.7 l34.7 124.3

S.D. 10.4 31. 8 23.8 22.5

*After 840 hours immersion in boiling water

copper/epoxy bond in boiling water was enhanced by pretreating the

copper surface electrochemically in a BTA coupling agent solution.
After 24 hours immersion in boiling water, both the cathodic and
the anodic treatments showed about 12% increase in strength over
the dipping treatment. Longer immersion tests showed that all of
the 5-ABTA treated joints were much stronger than the untreated
control. However, after extended water immersion, differences
between the electrochemical treatments and a dipping treatment
were indistinguishable. The presence of oxygen in the solution
during the complex formation period was essential.

The results of the electrochemical aging tests and the required

presence of water and oxygen for the weakening of copper/epoxy
bonds suggested that the chemical stability of interface is an
important factor for the durability of the adhesive bond.
538 J.M. PARK AND J.P. BELL

Table X. Effect of Oxygen Dissolved in Boiling Water on Joint

Strength.

Pretreatment: No

Adhesive: 100 parts Epon 828 + 4 parts P108 (based on

dicyandiamide)

Curing: 2 hours at 175°C

Breakage Force (lbs)

Joint No. Dry Wet* Wet**

1 360 257 295

2 352 255 314
3 360 266 301
4 367 263 321
5 256
6 248

Avg. 359.8 257.5 307.8

S.D. 5.3 5.8 10.3

*Immersed in boiling water for 24 hours

**Immersed in boiling water with nitrogen purging for 24 hrs.

300

:e'"
~

-
.s:: 200
Cl
<:

-'"
...CD
-
<:
0
100

-1.0 -0.6 -0.2 0.2

applied potential (volt)

Fig. 6. Changes in joint strength as a function of applied potential.

EPOXY ADHESION TO COPPER 539

Table XI. Effect of Aging in Hot Air on Joint Strength.

Pretreatment: No

Adhesive: 100 parts Epon 828 + 4 parts PI08 (based on

dicyandiamide)

Curing: 2 hours at 175°C

Breakage Force (lbs)

Joint No. Unaged Aged* Wet**

1 360 435 257

2 352 406 255
3 360 399 266
4 367 388 263
5 379 256
6 248

Avg. 359.8 401.4 257.5

S.D. 5.3 19.2 5.8

*Heated in a 100°C oven for 24 hours

**Immersed in boiling water for 24 hours

ACKNOWLEDGEMENT

Financial support by Shell Development Co. for the present

work is gratefully acknowledged by the authors.

REFERENCES

1. J. M. Park and J. P. Bell, in "Adhesion Aspects of Polymeric

Coatings", K. L. Mittal, Editor, pp. 205-224, Plenum Press,
New York, April 1983.
2. I. C. Ogle and G. W. Poling, Canad. Meta11. Quart., 14(1),
37 (1975).
3. F. Mansfe1d, T. Smith and E. P. Parry, Corrosion Sci., lL,
289 (1971).
4. D. Chadwick and T. Hashemi, Corrosion SCi., 18, 39 (1979).
5. C. J. Lin and J. P. Bell, J. Po1ym. SCi., 16-,-1721 (1972).
6. J. M. Park, Master's Thesis, University oflConnecticut, 1980.
7. W. T. McCarvi11 and J. P. Bell, J. Adhesion, ~, 185 (1974).
540 J.M. PARK AND J.P. BELL

8. N. D. Greene, G. A. Moebus and N. H. Baldwin, Corrosion,

29(6), 234 (1973).
9. R:(f." Brewster and W. E. McEwen, "Organic Chemistry", 3rd Ed.,
p. 238, Prentice-Hall, Englewood, NJ, 1961.
10. J. Kruger, J. E1ectrochem. Soc., 108(6), 503 (1961).
11. T. N. Rhodin, J. Am. Chern. Soc., 73, 3143 (1951).
12. M. Osawa and W. Suet aka , CorrosionSci., 19, 709 (1979).
HYDROTHERMAL STABILITY OF TITANIUM/EPOXY ADHESIVE JOINTS

F. J. Boerio and R. G. Dillingham

Department of Materials Science and Metallurgical

Engineering
University of Cincinnati
Cincinnati, Ohio 45221

y-Aminopropyltriethoxysilane (Y-APS) was shown

to be an extremely effective primer for improving the
hydrothermal stability of titanium/epoxy adhesive
joints when applied to the adherends by adsorption
from dilute aqueous solutions prior to adhesive bond-
ing. The breaking strength of lap joints prepared
from unprimed adherends decreased slowly during im-
mersion in water at 60°C and was only about 960 psi
after 60 days. The breaking strength of joints pre-
pared from adherends primed with y-APS at pH 10.4
and 8.0 decreased very little during similar hydro-
thermal aging and was about 1750 psi after 60 days.
Joints prepared from adherends primed with y-APS
at pH 5.5 were somewhat less durable and had a
breaking strength of about 1180 psi after 60 days in
water at 60°C. The performance of y-APS as a primer
was very different for titanium and iron adherends
even though the molecular structures of the primer
films formed on the two substrates were very simi-
lar. As a result, the performance of y-APS primer
films could not be related to the overall molecular
structure of the films but was related to acid/base
interactions between y-APS and the oxidized surfaces
of titanium and iron.

541
542 F.J. BOERIO AND R.G. DILLINGHAM

INTRODUCTION

It is well known that proper pretreatment of metal surfaces

is essential for the successful use of adhesive bonding for joining
metals. Pretreatments for titanium adherends include a phosphate-
fluoride conversion coating, a modified phosphate-fluoride process,
an alkaline peroxide etch, and chromic acid-ammonium fluoride an-
odization l . The most useful pretreatments, the alkaline peroxide
etch and the chromic acid anodization, produce a relatively thick,
porous oxide 2 .

The use of organofunctional silanes as primers or coupling

agents for improving the wet strength of glass fiber reinforced
plastics is well known 3 . More recently it has been shown that cer-
tain organosilanes are extremely useful as primers for improving
the hydrothermal stability of iron/epoxy4 and aluminum/epoxy5,6 ad-
hesive joints. Comparatively little information is available re-
garding the use of silanes as primers for adhesive bonding of ti-
tanium. However, Schrader and Cardamone 7 have considered the use
of y-aminopropyltriethoxysilane (y-APS) as a primer for improving
the hydrothermal stability of lap joints prepared fro~ titanium-
6AI, 4V adherends and an anhydride cured epoxy adhesiv~., The dry
strength of lap joints prepared from adherends pretreated with l%
aqueous solutions of y-APS prior to adhesive bonding was always
about 25% greater than that of joints prepar'ed from unprimed ad-
herends. After 24 hours immersion in boi~ing water, the shear
strength of joints prepared from primed adherends was about 50%
higher than that of joints prepared from unprimed adherends. The
improved hydrothermal stability was attributed to possible covalent
bonding between the silane and the adhesive and the oxide 7 .

The objectives of this research were to determine the molecu-

lar structure of films formed by y-APS adsorbed onto titanium-6AI,
4V from dilute aqueous solutions and to determine the relationship
between such structure and the performance of y-APS as a primer
for titanium/epoxy adhesive joints. The molecular structure of
primer films formed by y-APS adsorbed onto titanium was determined
using reflection-absorption infrared spectroscopy (RAIR) and ellip-
sometry. RAIR is a technique for obtaining infrared spectra of
thin films on metallic substrates by reflecting radiation polarized
parallel to the plane of incidence from the substrate at a large,
nearly grazing angle of incidence. Ellipsometry is a technique for
determining the thickness and refractive index of a thin film on
a metallic surface by reflecting plane polarized visible light
from the surface and measuring the relative amplitude reduction
and phase shift for components polarized parallel to and perpendic-
ular to the plane of incidence. RAIR8,9 and ellipsometrylO have
been thoroughly described in the literature and no further discus-
sion will be given here.
HYDROTHERMAL STABILITY 543

EXPERIMENTAL

Sample mirrors for infrared spectroscopy were prepared by

mechanically polishing titanium-6Al, 4V coupons. After polishing,
the mirrors were rinsed several times in distilled water and then
dried in a stream of nitrogen. y-APS films were formed on freshly
polished mirrors by immersing such mirrors into 1% aqueous solu-
tions of y-APSfor approximately 30 minutes, withdrawing the mir-
rors, and blowing off the excess solution using a strong stream of
nitrogen.

Infrared spectra were obtained using an external reflection

accessory (Harrick Scientific Co., Ossining, NY) and a Perkin-Elmer
180 infrared spectrophotometer. The reflection accessory was con-
figured to provide one reflection at an angle of incidence equal
to 80°. A silver bromide wire grid polarizer was placed in front
of the entrance slit to the monochromator and oriented to transmit
only radiation polarized parallel to the plane of incidence at the
sample mirrors.

In some cases the sample mirrors were examined with an ellip-

someter (Rudolph Research Model 436) immediately before and after
adsorption of the y-APSfilms. The results obtained indicated that
the adsorbed films were typically about 100A in thickness.

Titanium/epoxy lap joints were prepared according to ASTM

standard Dl002. Titanium adherends (4" x I" x 0.063") were me-
chanically polished, rinsed, and dried in nitrogen. Pairs of ad-
herends were then bonded together using an adhesive consisting of
an epoxy resin (Epon 828, Shell Chemical Co.) and a tertiary amine
curing agent (Ancamine K-6lB, Pacific Anchor Chemical Co.). The
adhesive was cured for four days at room temperature and one hour
at 100°C. Prior to adhesive bonding, some of the adherends were
pretreated by immersion in 1% aqueous solutions of y-APS for 30
minutes and then blown dry with nitrogen. All of the lap joints
were immersed in a water bath at 60°C. At appropriate intervals,
joints were withdrawn from the bath and tested to determine their
shear strength. In some cases, the fracture surfaces were examined
using a scanning electron microscope (Cambridge 600) or an optical
microscope.

RESULTS AND DISCUSSION

The infrared spectrum shown in Figure lA was obtained from

films formed by y-APS adsorbed onto titanium mirrors as described
above. The spectrum is characterized by a strong band near 1070
cm- l and by weaker bands near 1570, 1470, and 1300 cm- l • Infrared
spectra of y-APS monomer are dominated by intense bands near 1105,
1080, and 960 cm- l that are characteristic of SiOC bonds. ll The
544 F.J. BOERIO AND R.G. DILLINGHAM

2000 1100 1100 1400 1200 1000

CM·1

Figure 1. Infrared spectra of thin films formed by y-APS adsorbed

onto titanium mirrors from 1% aqueous solutions at pH 10.4: (A)-
as formed, (B) - after five hours exposure to the laboratory atmos-
phere, and (C) - after three days exposure. Reproduced by permis-
sion from reference 8.

absence of these bands from the spectrum shown in Figure lA indi-

cates that y-APS was adsorbed as highly hydrolyzed oligomers. How-
ever, the band observed near 1070 cm- l in Figure lA is assigned to
an SiOSi stretching vibration, indicating that the adsorbed silane
polymerized on the surface to form a siloxane polymer.

The bands near 1570, 1470, and 1300 cm- l have been the subject
of considerable debate ll - 13 • However, we have recently shown that
these bands are related to the formation of amine bicarbonate spe-
cies by absorbed carbon dioxide 14 • The bands observed near 1570
and 1470 cm- l are, therefore, assigned to the deformation modes of
protonated amino (NHj) groups while the band near 1300 cm- l is
assigned to the symmetric stretching mode of bicarbonate ions.

The as-formed films were stable in a dry atmosphere but seveTal

changes were noted during exposure to the laboratory atmosphere.
HYDROTHERMAL STABILITY 545

(see Figures lB and le). The band originally observed near 1070
cm- l gradually split into components near 1130 and 1040 cm- l , in-
dicating additional polymerization of the adsorbed silane. The
band initially observed near 1470 cm- l decreased in intensity and
the band near 1300 cm- l increased in frequency to about 1330 em-I,
indicating that the bicarbonate species was unstable in the presence
of atmospheric moisture.

The spectra shown in Figure 2 were obtained from films formed

by y-APS adsorbed onto titanium mirrors before and after heat treat-
ment in an oven at 100°C for 20 minutes. The bands originally
observed near 1570, 1470, and 1330 cm- l (Figure 2A) are not observed
after heat treating (Figure 2B), indicating that the absorbed carbon
dioxide is easily removed from the as-formed films. Moreover, the
band originally observed near 1130 cm- l has increased somewhat in
frequency to about 1145 cm- l and the band near 1040 cm- l has become
more prominent, indicating additional polymerization within the film.

The spectra shown in Figure 3 were obtained from films formed

by y-APS adsorbed onto titanium mirrors from 1% aqueous solutions
that were acidified by addition of HCl. These spectra are dominated
by an intense band near 1120 cm- l that is assigned to an SiOSi
stretching mode, again indicating formation of siloxane polymers.

2000
eM"'
Figure 2. Infrared spectra of thin films formed by y-APS adsorbed
onto titanium mirrors from 1% aqueous solutions at pH 10.4: (A)-
before and (B) - after heat treating at 100°C for 20 minutes.
546 F.J. BOERIO AND R.G. DILLINGHAM

A---.. .

2000 1600 1800 1400 1200 1000

CM-'

Figure 3. Infrared spectra of thin films formed by Y-APS adsorbed

onto titanium mirrors from 1% aqueous solutions at (A) - pH 10.4,
(B) '- pH 8.5, and (C) - pH 5.5.

The weaker bands near 1600 and 1500 cm- l are assigned to the defor-
mation modes of amino groups that are protonated as amine hydro-
chlorides (NH~Cl-).

The stability of the air-formed oxide on titanium mirrors dur-

ing immersion in aqueous solutions of y-APS has also been considered.
Several mirrors were polished as described above and then heated in
air at about 300°C for one hour. Results obtained from ellipsometry
indicated that the oxide on the mirrors was about 150A in thickness.
RAIR spectra from such mirrors were characterized by an absorption
band near 850 cm- l that was assigned to the rutile form of Ti0215.
The intensity of the oxide band near 850 cm- l was then determined
as a function of immersion time in 1% aqueous solutions of y-APS
at pH 10.4 (see Figure 4). The results obtained indicated that the
air-formed oxide on the titanium mirrors was stable during immersion
in such solutions. Similar results were obtained at pH 5.5.

These results indicate that there are only minor differences

in structure between films formed by y-APS adsorbed onto titanium
mirrors from aqueous solutions at pH 10.4 and those formed by ad-
sorption from aqueous solutions at lower pH values. The films are
HYDROTHERMAL STABILITY 547

1000 800 CM",800 700

Figure 4. Effect of immersion in 1% aqueous solution of y-APS at

pH 10.4 on infrared spectra of oxide on titanium mirrors: (A)-
before immersion, (B) - after immersion for thirty minutes, and
(C) - difference.

always composed of siloxane polymers. However, at pH 10.4 the

amino groups form bicarbonate salts with adsorbed carbon dioxide.
At pH values less than 10.4, the amino groups are protonated as
amine hydrochlorides.

As indicated above, the use of y-APS as a primer for improving

the hydrothermal stability of titanium/epoxy lap joints has also
been considered. The results are shown in Figure 5. The average
shear strength of joints prepared from unprimed adherends decreased
slowly with immersion time and was only about 960 psi after 60 days.
However, y-APS was an extremely effective primer. The breaking
strength of joints prepared from adherends primed with y-APS at
pH 10.4 and 8.0 hardly decreased during water immersion and was still
about 1750 psi after 60 days. y-APS was considerably less effec-
tive as a primer when applied to the adherends at pH 5.5. The
breaking strength of such joints was about 1180 psi after 60 days
immersion in water at 60°C.
548 F.J. BOERIO AND R.G. DILLINGHAM

2800 • pH 5.5
• • pH 10.4
2400 .... pH 8.0
• =no ,lIln.

,.... 2000 ... I

•
--
<II
...
•
0. 1600
(J)
(J)
• •
1200
W
a:
t-
(J) 800 •
400

5 10 20 30 40
IMMERSION TIME(DAYS)

Figure 5. Breaking strength versus immersion time in water at 60°C

for titanium/epoxy lap joints.

It is very interesting to compare the results described above

with those previously obtained using iron substrates 4 ,11,12. Al-
though some evidence for etching of the oxidized surface of iron in
aqueous solutions of y-APS at pH 10.4 has been obtained 16 , the in-
frared spectra of thin films formed by y-APS adsorbed onto iron
mirrors are very similar to the spectra of films formed on titanium,
indicating that the structures of the films are also similar. How-
ever, the hydrothermal stability of iron/epoxy lap joints was very
different from that of titanium/epoxy lap joints 5 . The breaking
strength of iron/epoxy lap joints always decreased rapidly during
the first ten days of immersion and then continued to decrease at a
much slower rate for times as long as 60 days (see Figure 6). y-APS
was an extremely effective primer for iron/eEoxy lap joints, espe-
cially when applied from acidified solutions. Iron/epoxy lap
joints prepared from unprimed adherends retained about 25% of their
initial strength after 60 days in water at 60°C. Joints prepared
from adherends primed with y-APS at pH 10.4 retained about 50% of
their strength after 60 days but joints primed with y-APS at pH 8.0
retained about 70% of their strength after such immersion (see
Figure 6).

It is evident that the properties of y-APS primer films are

very different on titanium and iron substrates even though the
molecular structures of the films are very similar. The performance
of y-APS primer films on iron improves substantially when the pH
at which the primer is applied is lowered from 10.4 to 8.0 even
though there are only minor differences in structure between films
formed at pH 10.4 and those formed at 8.0. Moreover, the perfor-
mance of y-APS primer films on titanium is essentially the same
HYDROTHERMAL STABILITY 549

100

75

~ C
ci
UJ
Z
;(
I-
UJ
a: 50
8 0
:I:
I-
(!)
Z
UJ
a:
l-
V)

25
•
A
•

o+---_+----~--~--~----+_--_+----~--~--~----+_--~--~
o 10 20 30 40 50 60

IMMERSION TIME, DAYS

Figure 6. Breaking strength versus immersion time in water at 60°C

for iron/epoxy lap joints: (A) - no primer, (B) - y-APS primer at
pH 10.4, and (C) - y-APS primer at pH B.O. Adapted by permission
from reference 4.

when the primer is applied at pH 10.4 and B.O even though some minor
structural changes are observed between films formed at pH 10.4 and
those formed at pH B.O. As a result, it must be concluded that the
performance of y-APS primers for titanium/epoxy and iron/epoxy adhe-
sive joints is not governed by the overall molecular structure of the
primer films. However, it is considered that the performance of the
primers can be largely explained in terms of acid/base interactions.

Bo1ger l7 has suggested that an organic compound having ion-

izable functional groups will be stably adsorbed onto the oxidized
surface of a metal during immersion in water as long as the pH of
the water is between the isoelectric point of the oxide (IEPS) and
the pkA of the ionizable functional group. Otherwise, the organic
compound will be displaced from the surface by water.

As noted earlier, y-APS is hydrolyzed in aqueous solutions

to form si1ano1 species. Both the silanol and amino groups in hy-
drolyzed y-APS are ionizable and have pkA values of about 3.0 and
10.0, respectivelylB. IEPS for the oxidized surfaces of iron and
titanium are approximately 10.0 and 6.0 respectivelylB, and the
pH of the water used for hydrothermal aging of the lap joints was
550 F.J. BOERIO AND R.G. DILLINGHAM

about 6.2. At any pH value less than about 9.0, y-APS should be
adsorbed onto iron (IEPS 10.0) through the silanol groups (pkA'V3.O).
During exposure to water at pH 6.2, joints primed with y-APS at pH
values less than about 9.0 should be stable and that is exactly
what has been observed (see Figure 6). At pH 10.4, y-APS is prob-
ably adsorbed onto the oxidized surface of iron through either the
amino groups (pkA'VIO.O) or the silanol groups (pkA'V3.0). During
immersion in water at pH 6.2, the silanol groups should be stably
adsorbed but the amino groups should be displaced by water. As a
result, y-APS should be less effective when applied to iron at pH
10.4 and that is also observed (see Figure 6).

IEPS for the oxidized surface of titanium is about 6.0. At

either pH 10.4 or 8.0, y-APS should be adsorbed onto the oxidized
surface of titanium through the amino groups (pkA'VIO.O). During
immersion in water at pH 6.2, the amino groups are just stably ad-
sorbed. y-APS should be equally effective as a primer for titanium/
epoxy lap joints when applied at pH 10.4 or 8.0 and that is observed
(see Figure 5). y-APS should be adsorbed onto the oxidized surface
of titanium from aqueous solutions at pH 5.5. through both the amino
and silanol groups. During immersion in water at pH 6.2, the
amino groups should just be stably adsorbed but the silanol groups
should just be displaced by water. As a result, y-APS should be
somewhat less effective as a primer for titanium/epoxy adhesive
joints when applied to the substrate at pH 5.5 and this is also
observed (see Figure 5).

Some support for the mechanism described above was obtained

from visual inspection of the fracture surfaces of titanium/epoxy
lap joints that were tested after immersion in water at 60°C for
long periods of time. Unprimed joints always failed very near the
interface, leaving a small patch of adhesive near one end of the
joint and an apparently bare area near the other end (see Figure
7A). Such behavior is characteristic of lap joints l9 and is re-
lated to the large peeling stresses near the ends of a lap joint.
The fracture surfaces of the primed joints were similar in that
failiure occurred near the interface and that a small patch of ad-
hesive was left near one end of the joint. However, some adhesive
was always observed in the "bare" area near the other end of the
joint (see Figure 7B) and it was evident that failure of the primed
joints was mostly within the adhesive even after long immersion
times.

When scanning electron microscopy (SEM) and optical microscopy

were used to examine the fracture surfaces of titanium/epoxy lap
joints that were tested after 60 days immersion in water at 60°C,
an even better indication of the effects of y-APS primers on the
locus of failure was obtained. Virtually no adhesive was observed
at the "bare" end of joints prepared from unprimed adherends
(see Figure 8), indicating that the adhesive was in fact displaced
HYDROTHERMAL STABILITY 551

Figure 7. Fracture surfaces of titanium/epoxy lap joints that were

tested after 60 days in water at 60°C: (A) - no silane primer;
(B) - y-APS primer a.t pH 10.4.

from the surface by water. However, some adhesive was always evi-
dent in the "bare" area of joints prepared from primed adherends
(see Figure 9), indicating that the adhesive was not readily dis-
placed from the surface in such cases.

CONCLUSION

y-aminopropyltriethoxysilane (y-APS) is an extremely effective

primer for improving the hydrothermal stability of titanium/epoxy
adhesive joints when applied to the adherends by adsorption from
dilute aqueous solutions prior to adhesive bonding. However,
y-APS primers are most effective when applied to titanium adherends
from aqueous solutions at pH values near or slightly below 10.4.
y-APS is also an effective primer for iron/epoxy adhesive joints but
is most effective when applied to iron adherends from acidified
solutions. The performance of y-APS as a primer for improving the
wet strength of titanium/epoxy and iron/epoxy adhesive joints is
not related to the overall molecular structure of the primer films
but is explained in terms of acid/base interactions between y-APS
and the oxidized surfaces of titanium and iron.
552 F.J. BOERIO AND R.G. DILLINGHAM

Figure 8. Optical micrograph of "metal" fracture surface of titan-

ium/epoxy lap joint prepared from unprimed adherends and tested
after 60 days in water at 60°C; 500X.

Figure 9. Optical micrograph of "metal" fracture surface of titan-

ium/epoxy lap joint prepared from adherends primed with y-APS at
pH 10.4 and tested after 60 days in water at 60°C; 500X
HYDROTHERMAL STABILITY 553

ACKNOWLEDGMENTS

This research was supported in part by a grant from the Office

of Naval Research. The assistance of Union Carbide Corp. in provid-
ing y-aminopropyltriethoxysilane and of Timet, Inc. in prqviding
titanium samples is gratefully acknowledged.

REFERENCES

1. M. C. Locke, K. M. Harriman, and D. B. Arnold, in "Proc. of

National SAMPE Symp. and Exhibition," pp. 1-12, 1980.
2. B. M. Ditchek, K. R. Breen, T. S. Sun, and J. D. Venables in
"Proc. of National SAMPE Symp. and Exhibition," pp. 13-24, 1980.
3. P. W. Erickson and E. P. Plueddemann, in "Composite Materials,"
L. J. Broutman and R. H. Krock, Editors, Vol. 6, Ch. 1, Aca-
demic Press, New York, 1974.
4. F. J. Boerio and J. W. Williams, Appl. Surf. Sci. 2, 19
(1981) •
5. F. J. Boerio and C. A. Gosselin, in "Proc. 36th Ann. Tech.
Conf., SPI Reinforced Plastics/Composites Inst.," Sec. 2G,
1981-
6. R. L. Patrick, J. A. Brown, N. M. Cameron, and W. G. Gehman,
App1. Polymer Symp. 16, 87 (1981).
7. M. E. Schrader and J. A. Cardamone, J. Adhesion.2.., 305 (1978).
8. F. J. Boerio, C. A. Gosselin, R. G. Dillingham, and H. W. Liu,
J. Adhesion 13, 159 (1981).
9. F. J. Boerio and C. A. Gosselin, in "Instrumental and Physical
Characterization of Macromolecules," C. D. Craver, Editor,
American Chemical Society, Washington, DC, in press, 1982.
10. F. L. McCrackin, E. Passaglia, R. R. Stromberg, and H. L.
Steinberg, J. Res. Natl. Bur. Stds. 67A, 363 (1963).
11. F. J. Boerio, L. H. Schoenlein, and J. E. Greivenkamp, J.
Appl. Polymer Sci. 22, 203 (1981).
12. F. J. Boerio, L. Armogan, and S. Y. Cheng, J. Colloid Inter-
face Sci. ll, 416 (1980).
13. H. Ishida, S. Naviroj, S. K. Tripathy, J. J. Fitzgerald, and
J. L. Koenig, J. Polymer Sci. 20, 701 (1982).
14. F. J. Boerio, J. W. Williams, and J. M. Burkstrand,. J. Colloid
Interface Sci. 91, 485 (1983).
15. S. Thibault, Thin Solid Films ~, L33 (1976).
16. F. J. Boerio and J. W. Williams, to be published.
17. J. C. Bolger in "AdhesionAspects of Polymeric Coatings,"
K. L. Mittal, Editor, pp. 3-18, Plenum Press, New York, 1983.
18. E. P. Plueddemann in "Composite Materials," L. J. Broutman
and R. H. Krock, Editors, Vol. 6, Ch. 6, Academic Press, New
York, 1974.
19. W. Russell, SAMPE Quarterly, p. 8, July, 1978.
 A STUDY ON ELASTOMER/METAL BONDS APPLICABLE IN

UNDERWATER SONAR SYSTEMS

Robert Y. Ting
Naval Research Laboratory, USRD
P. O. Box 8337
Orlando, FL 32856

Recent emphasis on the long-term performance of

sonar systems requires that investigations be carried
out to improve the durability of elastomer-to-metal
adhesive bonds in water. A screening test of some
commercially available adhesive systems was performed.
ASTM-D429 test methods using both conical button
specimens and peel-strips were employed. The 90-degree
peel test was modified such that the effects of stress
corrosion, water salinity and gas content could be
examined. Most commercial systems were found to give
adequate dry bond strength. Preliminary environmental
aging studies also showed that thermal aging did not
produce any measurable effect in accelerating the
degradation of the elastomer/metal bond. However, chemical
attacks from salt water under stress caused great
reductions in the bond strength, especially when the
oxygen supply was abundant. The weakened bonds were
found to show increasingly large portions of fracture
failure at the adhesive/metal interface rather than
rupture failure of the bulk rubber.

INTRODUCTION

In addition to applications for joining the structural com-

ponents in hydrofoil boats, surface effect vehicles and other
advanced ships, adhesive-bonding technology is also applied iIi
many underwater systems such as vibration damping mounts, acoustic
windows for sonar transducers and other acoustic conditioning
devices. Elastomer-to-metal bonding usually appears in one of two

555
556 R.Y. TING

forms. Either the cured rubber components are bonded to a metal

substrate or the uncured rubber is molded in place onto metal
surfaces which have been prepared in advance with primers and
adhesives. The latter technique is extremely popular, especially
when the area of coverage is reasonably small. Essentially all
the rubber face plates of wet-end sonar transducers are fabricated
in this fashion. For large rubber structures such as sonar dome
windows which contain steel reinforcements, adhesives are often
applied in-between plies during the lay-up process before autoclave
cure. In this case, the adhesive is used to promote adhesion
between the cured rubber surfaces. In the discussion here, however,
the emphasis will be on the elastomer/metal bond prepared by direct
vulcanization of uncured rubber to a metal substrate.

The life-time requirement of many sonar devices may be as

long as ten to fifteen years. Since the device is constantly
exposed to rigorous environmental stresses due to chemical attack
of sea water, temperature extremes, pressure cycling plus mechanical
stresses, it is at first glance very unlikely that an elastomer-to-
metal bond will survive over such a long period of time. Many
factors may contribute to this pessimistic observation. First of
all, the bonds are normally prepared in manufacturers' production
warehouse or in shipyard environments where a "clean room" condition
is probably difficult to maintain, if not impossible. Subtle
changes in bonding procedure or sheer negligence of the workers in
matters such as surface treatment, adhesive mixing and application,
and vulcanization temperature control could easily lead to an
unsatisfactory bond. Furthermore, primers, adhesives and rubber
compounds are all typically complex chemical mixtures. Their com-
positions are often kept by the manufacturer as secretive as possible
under proprietary claims. At the same time, since these materials
usually have narrow profit margins, the formulator will frequently
change suppliers or replace ingredients in order to maintain a
low cost basis. Without extensive studies of the long-term effect
of such changes, these formulation variations, even minor at times,
often lead to compromises in performance. Degraded adhesive bonds
cause serious water-leakage, corrosion, loss of mechanical integrity
and ultimately the failure of the complete sonar system. The result
is ever-increasing cost of maintenance, repair and replacement for
these systems, which also adversely affects fleet readiness. This
makes the reliability consideration an utmost important matter.

In order to improve the reliability of elastomer-to-metal bonds

in sonar devices, it is believed that steps toward stringent quality
control must be taken. It would be necessary to specify the chemi-
cal composition of adhesives and elastomers, and to develop required
bonding and molding procedures. The materials will eventually have
to be developed to replace proprietary commercial products. Quality
control methods need also be investigated, both for chemical analysis
and for performance testing. In this paper, the first effort made
A STUDY ON ELASTOMER/METAL BONDS 557

toward achieving this long-range goal will be described. The

concerns are testing methodology, adhesive screening and preliminary
environmental aging studies.

TEST METHODS

For quality control, the specification of adhesive chemical

composition is necessary, but unfortunately not sufficient. The
performance of adhesives depends not only on their chemical contents
but also on how the ingredients are mixed together and how the
adhesive bond is fabricated. Performance evaluation is therefore
essential. Although there are many mechanical tests in general
use at the present time, there is a need to determine their rele-
vance to field performance, and to standardize the test procedure.
For these reasons, two test methods have been investigated.

The first was the ASTM-D429-73 Method C, using a 4S-degree

conical button specimen. This method, in addition to being a
standard technique, was also cited by ASTM as statistically more
meaningful and more reproducible than other methods 1 • These ASTM
testing methods have also recently been reviewed by Cutts 2 . Figure 1
shows the schematic of the conical specimen. Conical buttons were
prepared from 2.S-cm metal bar stock. They were first vapor
degreased, grit blasted using 120 grit silicon carbide, and then
degreased again prior to coating with adhesive. The manufacturer's
recommendations were followed for each adhesive system used in order
to achieve the suggested nominal coating thickness. After sufficient
drying time, the thickness was checked with a magnetic type film
thickness gauge. An electrically heated transfer mold was used to
fabricate the test specimens. An injection molding technique was
applied for introducing the rubber into the mold for final vulcani-
zation. The fully cured specimens had a controlled distance of
11.S + 1 rnrn between the tips of the conical end pieces. The metal
cones-were pulled away from each other in an INSTRON machine at a
crosshead speed of Scm/min. Stress concentration at the tip of
the cones caused failure to initiate at the tip, then progressed
along the conical face until rupture. The required failure load
was recorded.

The second method employed in this investigation was a 90-degree

peel test in either air or water. The test was designed in a fashion
similar to that given in ASTM-D429-73 Method B, but with some modi-
fications. The method was essentially developed by the Navy3, and
has been applied extensively in the industrial communi ty4,S. It
basically involves hanging a weight on the end of a rubber peel
strip that is bonded to either a steel or an aluminum base, (see
Fig. 2). The specimen may be exposed to hot air for accelerated
life testing, or be immersed in water to study the environmental
effect on· peeling. The rate of peel was measured visually with an
558 R.Y. TING

optical microscope for a given weight. Alternatively, the weight

could be increasingly added to initiate peeling and the rate was
monitored for a specific time period. Besides being a simple method,
this technique has a main advantage in that it is essentially a
stress corrosion test. The chemical action of water accelerates the
debonding process while the bond is under sustained high stresses
from the load. This test method allows one to simulate one of the
more severe service conditions for many underwater rubber-to-rnetal
joints. Effects of parameters such as temperature, stress, salinity
and gas content of water may be easily studied with varying combina-
tions of elastomer, adhesive and metal.

ADHESIVE SCREENING

There are many commercially available rubber bonding agents.

These materials are normally offered as single-source solutions
for handling a broad range of adhesion problems. The idea is, of
course, to supply the fewest different adhesive formulas for
industrial applications covering a wide range of substrates and
elastomer formulations. It is therefore not surprising to find
that many adhesives are described as "general purpose cover coat
or one-coat adhesive". However, this approach is not acceptable
when the optimum long-term performance is sought for Navy's under-
water systems. For instance, an adhesive may be adequate for
bonding many household rubber items without showing any measurable
difference because the performance requirement is very minimal.
But for applications in underwater sonar systems, the performance
requirement is severe and a different combination of elastomer/adhe-
sive/substrate could be very critical in terms of the initial
strength or the lifetime of the bond. Table I is an example to
illustrate this point. Chemlok 304 has been recommended for bonding
rubbers to a steel surface. Four different kinds of rubber were
tested using the static wet peel test in distilled water. For a
2.54-cm wide strip of butadiene rubber, the peel rate was 0.125
cm/day under a 20-kg load. The natural rubber specimen showed a
greater peel rate of 0.25 cm/day at a reduced load of 16 kg. For
neoprene and nitrile rubbers, the load was further reduced in order
to obtain a meaningful measurement. The peel rate was found to be
even greater at 0.5 cm/day in these cases.

It is clear, therefore, that an initial screening test is

necessary. The rubber was chosen to be a Navy neoprene formulation,
and the following commercial adhesives were examined:
1. Chemlok 205/236A
2. Chemlok 205/220
3. Chemlok 205/234B
4. Chemlok 252
5. Thixon P6-1/0SN-2
6. Thixon D21243/508
A STUDY ON ELASTOMER/METAL BONDS 559

7. Thixon OSN-2
The Chemlok materials are the products of Hughson Chemicals, Lord
Corporation, and the Thixon adhesives were manufactured by the
Dayton Coatings and Chemicals Division, Whittaker Corporation.
Single designations mean one-coat adhesives, whereas double desig-
nations are combinations of primer/adhesive systems with the first
number being the primer. These adhesives were tested by using
conical button samples, which were prepared from AISI 4130 steel
bar stock.

Table II shows the result of this series of tests. Each of

the failure load data represents the average of eight measurements.
The fracture surface was visually inspected. The percentages of
bulk rubber tear (R), rubber/adhesive (R/A) and adhesive/metal (A/M)
failures were estimated, following the classification method
developed by Peterson 6 . With the only exception of Chemlok 205/234B,
all adhesive systems led to bulk rubber rupture. This indicates
that the adhesive bond is stronger than the bulk rubber, and the
bond strength should be recognized as greater than that indicated
in Table II.

Static 90 0 peel test was also carried out in distilled water

for a few selected adhesive systems. Not all the adhesives were
examined because the peel rate was generally very low, and a
complete investigation would be very time-consuming. Figure 3 shows
the test result for the same neoprene formulation with Chemlok
205/234B, which was the least effective system as determined by the
conical button tests. The plot shows the observed crack length as
a function of time under a static load of 17 kg. The slope therefore
gives the peel rate. The initial peel rate was always very large,
which was perhaps an artifact of the test. In five days the peel
rate reached a steady state value, which in this case was 0.05
cm/day. For Chemlok 205/236A, a better adhesive according to
Table II, the crack showed practically no growth for five days.
At that point, the rubber pulled apart at the holder of the dead
weight, a masked unprimed portion of the specimen. This forced the
test to be terminated, but gave an indication that the bond strength
was greater than the strength of the bulk neoprene, which was in
agreement with the observation given in Table II.

ENVIRONMENTAL AGING

Since the applications for underwater acoustic systems generally

have very long life-time requirements, the question of aging needs
to be addressed with accelerated life testing techniques. It is
common to assume that elevated temperature translates to longer
time. But such an assumption is valid only if the kinetics of the
aging process is known and the interrelationship between time and
temperature can be established. The long term durability of an
560 R.Y. TING

Table 1. Result of Wet 90 0 Peel-Test with Chem10k 304.

Rubber Load (kg) Peel Rate (em/day)
Butadiene 20 0.125
Natural 16 0.25
Neoprene 4.5 0.5
Nitrile 4.5 0.5

Table II. Neoprene-Steel Bond Strength as Determined by

Conical Button Samples at Room Temperatures.

Failure Load Fracture Surface Estimation (%)

Adhesive (kg) R R/A A/M

Chem10k 252 353 + 22 100 0 0

Chem10k 205/220 339 + 13 100 0 0

Thixon P6-1/0SN-2 336 + 11 100 0 0

Thixon D21243/508 329 + 7 99 0 1

Chem10k 205/236A 321 + 13 99 0 1

Thixon OSN-2 314 + 14 100 0 0

Chemlok 205/234B 310 + 16 86 7 7

A STUDY ON ELASTOMER/METAL BONDS 561

- - Metal

P ......1 - - - - - - 1•• P

Rubber
Figure 1. Conical button specimen for rubber/metal adhesion test.

Water Line

Figure 2. The schematic for static wet 90°-peel test.

(mm)
40

-oo--D---O--Oo--
___.-0--0-
30

20

W 17 kg

(days)

2 4 6 8 10 12
Figure 3. Crack length as a function of time in wet peel test
for ChernIok 205/234B.
562 R.Y. TING

adhesive bond in sea water probably involves at least two distinct

processes: (1) bond line stress corrosion and (2) oxidative or
other chemical degradation. Obviously, these processes are them-
selves very complex and may even be closely related to each other.
As a first step, simple attempts were made to obtain some under-
standing of the aging effect on adhesive bonds, and preliminary
results are presented here.

The result of the screening test shows that most of the existing
adhesive systems give an adequate dry bond strength in that the
bond strength is greater than the tear strength of the rubber itself~
The conical button specimens may then be used to examine the effects
of thermal aging and chemical degradation. Specimens were exposed
to either distilled water or 3.S% salt water at elevated temperatures,
then tested at room temperature for their bond strength. By doing
so, any effect of high temperature or chemical oxidation due to
salt water attack may be isolated, free from any stress-loading
complication. Up to this point, specimens aged at SO°C for 200
hours have been tested, and the results are shown in Table III.
The effect of SO°C aging in distilled water is almost unnoticeable
except for Chemlok 20S/234B and 20S/236A, in which case the failure
loads were reduced by about 13% and 11%, respectively. The deterio-
ration of the Chemlok bonds may be attributable mainly to the effect
of water rather than heat, because in a separate static 90°-peel
test for the same bond placed in a 7SoC dry air oven under a load
of 10 kg no crack growth was found in 14 days. The slight increase
in failure load in some cases may be related to the post curing
effect on rubber due to SO°C aging. The immersion in 3.S% salt
water at SO°C had a more pronounced degradation effect in that the
failure loads for Thixon D2l243/S08, Chemlok 20S/236A and 20S/234B
were reduced. Coupled with this reduction, it was observed that
the fracture surface showed increasing adhesive/metal interface
failure as a result of the salt water immersion. No observable
degradation effect was found for the Chemlok 2S2, 20S/220 or Thixon
P6-l/0SN-2 systems. As of the writing of this report, the aging
time was still too short to manifest any such effect yet.

The oxidative degradation effect of water was further examined

by using the wet 90°-peel test. The water tank was saturated with
either nitrogen or air, and the crack length measured as a function
of time. Table IV shows the result of a Chemlok 20S/234B bonded
neoprene/steel specimen. Under the same loading, crack growth was
enhanced by the saturated air in salt water. As the load was
increased from 7 to 9 kg, the crack growth was further accelerated,
as another indication of the stress corrosion phenomenon in this
type of adhesive bonds.

CONCLUDING REMARKS

Adhesive bond evaluation techniques using conical button

 :l>
CJ)
--I
C
0
-<
0
Table III. Effect of 200-hr. Wet-aging at 50°C on the Bond Strength Failure Load z
m
of Conical Button Specimens. r
:l>
CJ)
--I
Control'~ Distilled 3.5% Salt 0
Adhesive water (kg) Weit_er_ (kg1 ~
~ m
::IJ
Chemlok 252 353 356 351 ~
--
m
--I
:l>
Chemlok 205/220 r
339 347 350 OJ
0
Z
0
Thixon P6-l/0SN-2 336 346 350 CJ)

Thixon D2l243/508 329 328 295

Chemlok 205/236A 321 285 307

Thixon OSN-2 314 327 310

Chemlok 205/234B 310 271 302

*Room temperature, dry

C11
0)
w
564 R.Y. TING

specimens and peel strips were found to be useful in determining

the strength of elastomer-to-metal bonds under various conditions.
The static wet peel test has the main advantage of incorporating
the stress corrosion effect with other variables such as water
salinity and gas content in the same test. Many commercially avail-
able adhesive systems were found to provide adequate dry bond
strength. But chemical degradation by salt water under stress
loading was shown definitely to reduce bond strength. This degra-
dation effect seemed to be enhanced when the oxygen content in water
was high. These failures mainly occurred at the adhesive/metal
interface.

Table IV: Effect of Gas Content in

3.5% Salt Water on Neoprene-Steel Bond.
Static Crack Length in 4 Weeks (em)
Load
(kg) Nitrogen Air

7 o 0.2

9 0.15 0.8

REFERENCES

1. ASTM Standard D429-73, Part 37, p. 82 (1981).

2. E. Cutts, in"Development in Adhesives-2," Ed . A. J. Kinloch,
Appl. Sci. Publ. London, pp. 367-404 (1981).
3. Mare Island Naval Shipyard, Rubber Laboratory, Rept. MI-93-l5,
Feb. 1967, Mare Island, California.
4. General Dynamic Corporation, Electric Boat Division, Rept.
C443-77-082, 1977, Groton, Connecticut.
5. Bolt, Beranek and Newman Inc., Tech. Memo NL-055, May 1982,
New London, Connecticut.
6. C. H. Peterson, Adhesives Age, 2, 30 (1964).
 Part V
Stress Analysis and Performance Aspects
LIFE PREDICTION METHODOLOGY FOR ADHESIVELY BONDED JOINTS

J. Romanko 1 , K. M. Liechti 1 and W. G. Knauss 2

1 1
Genera ·
Dynam1cs 2Ca1ifornia Institute
Fort Worth Division of Technology
Fort Worth, TX 76101 Pasadena, CA 91125

A comprehensive integrated methodology for adhe-

sive bonded joint life predictions is outlined. The
guiding assumption is that the useful life of bonded
joints is determined by failure in the adhesive inter-
layer, and this is the basis for a systematic analysis
of information and the techniques required to provide
valid predictions. Emphasis is placed on time-Dependent
fracture mechanics procedures, including detailed through-
the-adhesive thickness viscoelastic finite element cal-
culations of the stress-strain distributions, and on
analytical methods involving the constitutive relations
of the adhesive interlayer. Use is made of instrumented
bonded joint data obtained from structural overtest
methods under a number of U.S. Air Force-sponsored pro--
grams. A logical program rationale is outlined which
has sufficient generality to apply to any adhesive bonded
joint structure including metal and composite adherends,
various loading/environmental conditions, which apply to
high-performance aircraft. The five main modules of the
integrated methodology are: (Definition of) Loads/Envir-
onments; (Adhesive) Material Properties; Stress Analysis;
Fatigue Mechanisms and Failure Criteria; and Structural
Failure Analysis (Life Predictions). The necessary con-
tents of each module, including input and output data,
are specified and a systems approach is taken inter-
connecting the modules into a logical sequence suitable
for making service life predictions for bonded structures.

567
568 J. ROMANKO ET AL.

INTRODUCTION

The use of polymeric adhesives to structurally join metals and

to join or repair composite aircraft structural components offers
significant cost and weight savings. Initial costs are reduced
through design simplicity and reduced part count as proven recently
in at least two advanced technology programs. 1,2 Operational costs
are potentially reduced through longer service lives. Maintenance
costs are reduced through demonstrated damage tolerance and fatigue
endurance far superior to that of riveted joints. Fuel savings can
result from reduced weight. To properly evaluate the cost savings
through improved service life of the bonded structures, and to be
able to more accurately estimate the total cost of ownership prior
to commitment to hardware, several programs have been sequentially
initiated as described below. They are highly representative of
fundamental understanding of bonded joint fatigue endurance and
damage tolerance. The backbone program of one such series of closely
coordinated programs conducted sequentially is the "Integrated Meth-
odology" program discussed herein.

INTEGRATED METHOD

Figure 1 shows the five major elements or modules (I through V)

of the predictive method and their inter-relationships as applied to
adhesively bonded structures. This integrated methodology parallels
the systems approach for solid rocket service life prediction first
proposed by Kelley and Trout 3 • Extension of Kelley's ideas from
solid rocket propellants to adhesively bonded structures is possible
in view of the time/temperature Cand moisture) dependence of the
constituent material (polymer) properties and similar failure modes
(crack propagation) in both cases. The various feeder sub-programs,
which must be accomplished for full implementation of the CD Inte-
grated Methodology (1M) program, are labeled @ through ® in
Figure 1 and are being carried out under AFWAL/MLBC-sponsored con-
tracts F336lS-: 79-C-Sl17 @(Residual Stress Analysis; Texas A&M
Univ.); BO-C-S093 (D(Basis for Accelerated Testing; TRI, Austin,
TX); BO-C-S167~(Viscoelastic Stress Analysis; General Dynamics,
Fort Worth Division, TX); Bl-C-Sl14 CD (Time Dependent Fracture;
United Technologies, Chemical Systems Div., CA); B2-R-S067 ~
Fracture Analysis Boundary Value Problems;CD Fracture Data Base;
and ® Constitutive Relations with Damage.
LIFE PREDICTION METHODOLOGY 569

I LOADS AND ENVIRONMENT

DEFINITION ..... III CD
STRESS
0
IGE"'ETRY~
ANALYSIS

--.11 MATERIAL PROPERTIES

V
~
MECHANICAL
RESPONSE CD I STRUCTURAL I
I
CORE
O(dY I I INTEGRITY
aT,aH20, Tg
STATE,
INTERFACE ---------
---- - - - --
QUALITY, !TULT ~

FRACTURE CD
ETC
STRENGTH LIFE
PROPERTIES (j) ANALYSIS
(M.S.)
~ PREDICTION
(MS=O)
daJdN, K, Kc 7

~-
~

IV FAILURE
CRITERIA
FAILURE
/
(INSTRUMENTED ANALYSIS
TESTING) STRENGTH OF
STRENGTH
4 FAILURE
MODE ~ THEORY MATERIALS
DEFINITION -FATiGUE&" FRACTURE AND
FRACTURE CRACK GROWTH
THEORY

Figure 1. Modules and Flow Path of Predictive Method: G)

"Integrated Methodology for Adhesive Bonded Joint Life Predictions".

The format of the 1M program highlights setting up the frame-

work and its constituent modules which would be used to make the
life prediction(s). The 1M program would deal with the individual
modules to some extent, with the specific details of the modules to
be developed by the feeder sub-programs. In any case, priorities
were established on which of the elements of each module were
necessary with an indication of the gaps in the availability thereof
for the life prediction.

The contents of the five main modules (viz., analytical methods

and/or questions that the module needs to address) and their inputs
and outputs have been established, and are duly recorded in great
detail in Reference 4. Included therein is an outline of the experi-
mental test plan for developing the required data, in addition to
570 J. ROMANKO ET AL.

an overall integration test plan for effecting the life predictions

of a bonded jo.int of a real structural aircraft component.

In view of the time and space limitations of this paper, we

will endeavor to highlight some of the more important modules,
supplemented by pertinent comments concerning the remaining modules
which are of basic importance to the methodology. First. we will
make general observations regarding certain of the modules which
set them apart from the others.

We note that physical laws govern the contents of a number of

the modules, with only engineering approximations poss·ible in des~
cribing the phenomena involved. Also, oasically there are no modifi-
cations possible for these except for economic trade~·offs in these
approximations. Modules in this category commanding top priority
include "Stress Analysis" and "Failure Mechanisms and Criteria".

Modules which allow descriptions not directly governed by

physical laws allow system choices whicft affect tfte final result.
Two such modules include "Loads and Environments" and "Material
Properties" whicft are of top priority in our scheme. Geometry, if
it can be considered a module at all. commands secondary· emphasis
at this time.

Material properties represent the only element sensitive to

aging time although there is naturally a time element inherent in
the environmental loading rate, duration and frequency. Since
storage time is a primary variable in the service life prediction
scheme, the overall rationale should apply as well to the zero-time
design point which considers all material properties in the unaged
condition. Another important breakout of material properties. when
considering the impact of aging, is the separation of response prop-
erties from the fracture properties; response properties being a
principle input to the stress analysis and fracture properties
applying primarily to the strength analysis.

The response and fracture characteristics of bonded structures

are known to be dependent on loading rate and on history in such a
complicated fashion that analytical procedures must be supplemented
with adequate laboratory testing to provide the necessary design
confidence. The guiding assumption in this integrated methodology
scheme is that the use~ul life of a bonded structure is determined
by the initiation and relatively slow propagation of cracks emanating
from corner singularities and reaching a critical size at which time
catastrophic failure occurs.
LIFE PREDICTION METHODOLOGY 571

. The emphasis in this integrated methodology scheme is on the

"deterministic" approach, with full recognition that a "probabilis-
tic" treatment will eventually have to be incorporated in the method-
ology to realistically predict mixes of missions and mission profiles
and to account for experimental error in the data used. However,
since a deterministic approach provides the basis for a statistical
representation and in view of the time and budget resources of this
program, the deterministic approach was the first priority.

We proceed with a description of the contents of a number of

the five main modules and the corresponding input and output infor-
mation required. Of special interest is the central position of the
Stress Analysis module (III) and the dominance of the Failure Mech-
anisms and Criteria module (IV) in focussing all related activities.
Accordingly, we will highlight these two modules in the latter stages
of this discussion.

The experimental data required to validate the concepts in

the modules is indicated herein and the overall integration of the
method for the analysis of a test joint or an "aircraft", with
associated Management Plan is also presented.

Order of priority of some items-, insofar as being accomplished

in the Integrated Methodology program is indicated by the designa-
tion pen) after the time in the TaBles, with n=l, 2, 3 ••••• in order
of decreas-ing priority. Of course, the aforementioned feeder sub-
programs will eventually develop the necessary inputs for full imple-
mentation of the methodology for a general case.

Loads/Environments (MOdule I)

Often the types of loads under consideration are directly

related to failure modes (as far as the way engineers do it); one
tends to forget that often loads may produce different, possibly
longer term failures. It is important to delineate the kinds of
"loads" so that the types of histories are accounted for in material
characterization (stress and failure response).

Table I summarizes the contents of this important module,

emphasizing the time, temperature. moisture, stress level and
sequence of events in defining the impressed loads. whatever their
origins may be. The magnitude and frequency of the periodic mech-
anical load-and :steady and monotonic non-periodic mechanical loads
are of concern in the Integrated Methodology program.
572 J. ROMANKO ET AL.

Table I Loads/Environments, Contents * .

A. MECHANICAL LOAOS

LOAD HISTORIES (DEDUCED MISSION PROFILES)

MISSION PROFI LE

MANEUVER SEQUENCE

FLIGHT ENVELOPE

(a) PERIODIC (FATIGUE) - MAGNITUDE

FREQUENCY

(b) NON-PERIODIC - STEADY

MONOTONIC

B. THERMAL ENVIRONMENT

(a) EFFECT ON A (THROUGH STRESS ANAL YSIS) UNDER

CONSIDERATION OF DIFFUSION TIMES

(b) EFFECT ON MATERIAL BEHAVIOR

C. MOISTURE ENVIRONMENT

(a) EFFECT ONA (THROUGH STRESS ANAL YSIS)

UNDER CONSIDERATION OF DIFFUSION

(b) EFFECT ON MATERIAL BEHAVIOR

D. SEQUENCE OF MECHANICAL LOADS, THERMAL MOISTURE ENVIRONMENT

*Details are recorded in Reference 5.

Table II summarizes the input and output information rela-

ting to this module. Traditionally, the Loads Group has developed
an expertise in these types of problems for rate independent struc-
tures, with some modification effected by "composite" needs. For
adhesives, we envisage the requirements indicated for failure anal-
ysis purposes (output from the module and raw loads data input).
It must be noted that the listing of input and output items is not
necessarily complete at this time, although we have tried to be as
inclusive as possible.
LIFE PREDICTION METHODOLOGY 573

Table II *
Loads/Environments, Inputs and Outputs.

INPUT OUTPUT

MECHANICAL * TIME DEPENDENCE DF LDADS

AIRPLANE RECORDS ON ACCELERATION
WITH EXPLICIT TIME (A) STEADY

(A) GROUND REST (B) CYCLIC WITH WHAT

FREQUENCY
(B) TAKE-DFF
(C) "BLDCKS"
(C) LANDING FOR STATISTICAL
ANALYSIS
(0) IN-FLIGHT MANEUVERS

DETERMINISTIC

STATISTICAL (SPECTRA)

THERMAL * (TIME DEPENDENT)

GROUND STO RAGE TIME SCALE TEMP.ERATURE DISTRIBUTION

FLIGHT: STAGNATION-INDUCED TIME SCALE TEMPERATURE GRADIENTS

ENGINE - DUCT INDUCED TIME SCALE (INCLUDING SKIN TEMPERATURE)

ATMOSPHERIC TIME SCALE RADIATIDN OF SUN

MOISTURE (AND OTHER SOLVENTS) * TIME DEPENDENT

DIFFUSIDN AS A FUNCTION OF TEMPERATURE H20 (AND/OR SOLVENT)

GROUND STORAGE STRESS CONCENTRATION AND

DEPENDENCE GRADIENTS

FLIGHT DRY-OUT PDWER SPECTRAL RESPONSE

*Details are recorded in Reference 5.

Material Properties (}fodule II)

The local stress distribution developed in a bonded joint will

be greatly influenced by the material properties which define their
deformation characteristics. The "response" properties of the adhe-
sive, such as the creep compliance and Poisson's ratio will dictate,
by and large, these deformational characteristics of the bonded
joint and in general will exhibit time, temperature, moisture and
stress level dependence.

After cyclic loading of the environmentally conditioned (tem-

perature and humidity) bonded joints, the "fracture" or "failure"
properties of the adhesive will control the residual strength
574 J. ROMANKO ET AL.

capabilities of the structure. Typical fracture properties of the

adhesive (joint) include the crack growth rate with cyclic frequency,
viz., da/dn = at, or the equivalent da/dt = arepresentation. Fail-
ure or fracture "properties" are dependent on the current knowledge
of failure analysis (properties are determined in tests prescribed
by failure analysis).
A comprehensive review of these time-dependent response and frac-
ture properties of adhesives required for this program is summarized
in Reference 6. The physical and mechanical behaviors considered
included constitutive relations under monotonic and cyclic loading
conditions, dilatations during cure, cooling and moisture infusion,
aging, mechanisms of damage accumulation, fracture toughness and
ultimate strength spectra, among others.
Table III and Table IV summarize the response and fracture
properties contents of the Materials Property module, and their
sensitive dependence on time, temperature, moisture, and stress
level, among other parameters 6 .

Physical, chemical and structural aging effects are distin-

guished and these effects are being covered in the program (D on
"Basis for Accelerated Testing," currently in progress.

The importance of the dependence on response and fracture prop-

erties of the state of the adhesive, whether it is neat-neat (without
scrim), or neat (with scrim) is acknowledged.

The fracture properties of the adhesive layer describe the rela-

tionship between a fracture parameter and the crack growth rate in
the adhesive layer. Thus, since the fracture properties are known,
the history of the crack growth can be obtained by integration of
the crack growth rate. Failure can then be defined by the achieve-
ment of some critical length crack. Since the geometry changes as
a crack grows in a joint, the relationship between the crack length
and the chosen fracture parameter must also be determined in order
to evaluate the integral. The variation of fracture parameter with
crack length is determined by stress analysis which is the subject
of module III. The choice of fracture parameter and a discussion
of failure mechanisms and criteria comprise module IV.

Stress Analysis (l1odule III)

A thorough, through-the-thickness stress analysis is vital to

the success" of the Integrated l1ethodology program. It allows point-
LIFE PREDICTION METHODOLOGY 575

Table III Materials Properties,* Contents (Response).

RESPONSE

Dhl,or Ehlor Jhl

k(t) or U (tl (WHAT MINIMUM ACCURACY REQUIRED?)

TRANSITION TEMPERATURE (GLASS OR MElT TEMP FOR CRYSTALLINE COMPONENTS

IN AN ADHESIVE)

SHIFT FACTORS cP DUE TO TEMPERATURE CPT

MOISTURE <Pp
STRESS LEVEL <Pa-
COEFF OF EXPANSION DUE TO TEMPERATURE aT

MOISTURE a p (x, tI
STRESS LEVEL ( :: kit))

__ VOLUME
CONSTITUTIVE FORMULATION; NON-LINEAR VE ...... OCTAHEDRAL SHEAR

AGING: PHYSICAL (STRUIK)

CHEMICAL (DARK EDGE)

STRUCTURAL (MICROCRACKING)

NEAT vs NEAT-NEAT BEHAVIOR*

*NEAT TAPE IS WITH MATTE DACRON SCRIM_

NEAT-NEAT TAPE IS WITHOUT SCRIM_

*Details are recorded in Reference 6.

by-point description of the displacements and stresses (both peel

and shear) on a "micro level" in attempting to account for material
displacements (creep) and damage mechanisms occurring in the adhesive
inter layer on a realistic basis. Averaging of stresses through the
thickness of the adhesive interlayer leads to overly-conservative
results.

Table V summarizes the requisite content of the Stress

Analysis module, including the validation of the stress analysis by
holographic interferometry. This validation (described in Reference
7, Appendix 1) is an important part of the methodology since it will
allow first-hand confirmation of the finite element procedures
adopted in this program.
576 J. ROMANKO ET AL.

Table IV Materials Properties,* Contents (Fracture).

FRACTURE

/COHESIVE
t
r= n il p(1) d.
i dt
"-ADHESIVE •
/COHESIVE
~ooooooo§
r., r (I') P(1) d.
a'
"ADHESIVE
!iii

WHAT IS THE "PLASTIC ZONE " SIZE a , AS RelATED TO

BOND LINE THICKNESS; AFFECTED BY MODE INTERACTION? P(1)

EFFECT OF PLANE (]" vs PLANE E

IS FATIGUE BEHAVIOR GOVERNED BY K DR ~K? P(2)
IS SCRIM CLOTH IMPORTANT TO FIRST ORDER?

(I) SCRIM IN A BUn JOINT UNDER NORMAL LOAD }

P(3)
~ SHEAR LOAO
<00000001
(b) SCRIM ORIENTEO IN TEST PLANE (LABO RATORY

·:
TYPE TEST & NEAT)

8,- <...:,<- \'~

. _ ,_ ., J
\
P(2)

(e) NO SCRIM, NEAT·NEAT MATERIAL

IS TIME·TEMPERATURE ·WATER (SOLVENT) "STIFF " BEHAVIOR

SYNERGISTIC P(3)

EFFECT OF CURE LEVEL P(3)

EFFECT OF AGING (CHEMICAL, PHYSICAL) P(3)

ULTIMATE TENSI LE PROPERTIES (SMITH PLOTS) P(3-4)

* Details are recorded in Reference 6.

Table VI details the input and output information required

for and expected from the Stress Analysis module. No te the
selection of the appropriate cost-effective, finite element stress
analysis program for the calculation at hand. In some cases, the
more extensive VISTA program with generalized boundary conditions
being developed under program ®
the Viscoelastic Stress Analysis
program is prescribed. In other cases, the original MARC program
or its recent PRONY series modification may be more cost-effective,
LIFE PREDICTION METHODOLOGY 577

Table V Stress Analysis, Contents.

HIERARCHY OF REQUIREMENTS AND CAPABILITIES: WHICH PROGRAM:

A. MATERIAL REPRESENTATION

LlNEARL Y ELASTIC MARC VISTA

ELASTIC - PLASTIC (WITH AND WITHOUT STRAIN HARDENING) M

LIN VE M V

NON·LlNEAR VE V

VISCOELASTIC NON-LINEARITY (GEOMETRY UPDATING) M V

DILATATIONAL BEHAVIOR OF MATERIAL DUE TO TEMPERATURE

AND DR WATER (SOLVENTS) M V

B. GEOMETRY REPRESENTATION:

CRACK TIP ELEMENT

INTERFACE CRACK TIP ELEMENT \
I (///////////

-=- .....-=
/////////////
DEAL WITH INTERACTION
M V

END TERMINATION I• M V
(CORNER SINGULARITY)

VOIDS AND POROSITY

C. DISTRIBUTION OF V
I) TEMPERATURE
b) WATER

-VALIDATlON-

(HOLOGRAPHIC INTERFEROMETRY)*

*Details are described in Appendix 1 of Reference 7.

and, of course, the much more cost-effective MacNeal Schwendler Corp.

NASTRAN program is recommended for gross analyses.

At the present time. the MARC finite element program is the most
flexible and powerful program for performing the stress analysis of
adhesive bonded joints. I t has many significant capabilities that
are directly applicable to the Integrated Methodology program includ-
ing: plasticity, viscoelasticity. large deflections. and J~integral
calculation. The addition of the generalized Prony series expansion
for viscoelastic material behavior makes MARC one of the few programs
readily available for performing viscoelastic analysis. Finally.
MARC is well supported. documented, and relatively easy to use con-
sidering that it is a very sopIiisticated finite element program.
578 J. ROMANKO ET AL.

Table VI Stress Analysis, Inputs and Outputs.

INPUT OUTPUT

A. GEOMETRY INCLUDING MARC VISTA ENERGY RELEASE RATE OR J INTEGRAL

FRACTURE PATH _ (THICKNESS)
(a) INTERFACE AVERAGED
(b) COHESIVE - DETAILS

V STRESS INTENSITY FACTOR

B. MATERIAL RESPONSE CHARACTERISTICS V CORNER SINGULARITY STRENGTH

CONSTITUTIVE BEHAVIOR ACCOUNTING

FOR TEMP AND WATER M V STRESS DISTRIBUTION MAPS FOR STRESS
COMPONENTS PLUS

C. BOUNDARY CONDITION ON (a) PRINCIPAL STRESSES AND THEIR

(b) ORIENTATIONS
(a) LOADS (e) OCTAHEDRAL
(b) TEMPERATURE (d) DILATATIONAL (a'1 :!: <TZ:!: <T 3 )
(e) WATER

O. EXTREME EXAMPLE: TIME MARCHING

PROBLEM M V NODAL DISPLACEMENTS

V ENERGY RELEASE FOR MINIMUM CRACK

P(Z)

DETERMINATION OF CRACK PATH TO BE

A
ACCOMPLISHED BY TRIAL AND ERROR FOR MAXIMUM
ENERGY RELEASE RATE AS A FUNCTION OF INCRE-
MENTAL CRACK EXTENSION

NASTRAN WILL BE USED FOR GROSS ANALYSIS

Currently a more sophisticated version of MARC, called the MARC

Prony program has become available. It permits a more accurate
representation of the viscoelastic behavior of adhesives. The more
general program VISTA is being developed with more generalized boun-
dary condition capabilities.

An example of MARC finite element modeling of the adhesive

interlayer of a bonded joint, e.g., a structural lap joint, SLJ, is
shown in Figure 9 of Reference 8. Details of the mesh refinement
at the free edges of the adhesive where stress concentrations are
possible are shown in the enlarged views therein. As cracks grow
during the tension-tension fatigue cycling, the corresponding finite
element models can be modified so that the J-integral calculations
can be performed. This' requires local refinement of the mesh similar
to that done to the model joint (MJ) in an early phase of the Inte-
grated Methodology program.5 An r-~ strain singularity is created
by using a ring of isoparametric quadrilaterals with collapsed sides
at the crack tip. The J-integral is calculated by determining the
LIFE PREDICTION METHODOLOGY 579

change in stiffness due to a differential change in crack length and

multiplying this by the displacement vector created by the applied
load. This gives the change in energy due to crack advancing. When
divided by the length, a value for the J-integral can be determined.

Failure Mechanisms and Criteria (Module IV)

The contents and inputs and outputs of this important module

are shown in Tables VII and VIII, respectively.

There are a number of reasons that crack growth in an adhesive

bond could differ from that in metals. First, we recognize that
adhesives are polymers and therefore potentially viscoelastic and
time dependent. Second, the adhesive layer is heavily constrained
by the metal adherends, giving rise to the possibility of geometric
and material nonlinear behavior. Finally, fatigue crack growth must
be understood under conditions of varying degrees of lIIode interac-·
tion.

The fracture parameter chosen to characterize fatigue failure

in adhesively bonded joints in this program is the J-integral, as
described in the previous section. Not only can it be compared
directly with the linear elastic fracture mechanics (LEFM) approach

Table VII Failure Mechanisms and Criteria, Contents.

CRITERIA OF FAILURE - BRITTLE, "OUCTILE"; VISCOELASTIC

(A) CRACK GROWTH RATE EXCESSIVE pm
(B) AN INSTABILITY DEVELOPS (DISCONTINUOUS GROWTH RATE OF PIZ)
CRACK COALESCENCE)
(C) CRACK SIZE EXCEEDS ALLOWABLE VALUE P(3)
VIA (A) THICKNESS AVERAGED ENERGY RELEASE
(B) LOCAL ENERGY RELEASE OR STRESS INTENSITY
INCLUDING LOCAL CRACK TIP DISPLACEMENT
~"ACK P(I)

CRITERIA FOR GENERATION OF SMALL BUT

FINITE CRACK
(C) MINIMAL ENERGY PATH FOR ADHESIVE ANO P(Z·3)
COHESIVE FRACTURE

VECTORIAL ALONG INTERFACE,

DISPLACEMENT CRITERION: Kf + Kfl = r OR ALONG SCRIM PLANE
580 J. ROMANKO ET AL.

Table VIII Failure Mechanisms and Criteria, Inputs and Outputs.

INPUT OUTPUT

ENERGY RELEASE RATE G (A) CURRENT RATE OF CRACK

GROWTH

STRESS INTENSITY FACTORS KI, KII (B) CRACK SIZE AS A FUNCTION

ANO A KI, A KII OF TIME

FRACTURE PROPERTIES, r (i), r (a') (C) INTRODUCTION OF MULTIPLE

CRACK LEADING TO INSTABILITY

FRACTURE PATH (COULD BE CALCUALTED

AT GREAT COST)
da
i =dt
da
a' =iiiI

commonly used for metals, but it is also defined for cases of non-
linear elastic, nonlinear elastoplastic and, more recently, nonlinear
viscoelastic behavior 9 . The J-integral has also been found to be
equivalent to the vectorial crack opening displacement, a fracture
criterion which was successfully us~d to characterize interfacial
crack propagation in an adhesive joint subjected to loading normal to
or tangential to the bondlin~~ A data base for the predictive meth-
odology is being developed using three specimen geometries; namely,
the thick adherend model lap joint (MJ} , the cracked lap shear joint
(CLS) and the structural lap joint (SLJ) based on the plan of Table
IX. Preliminary results of measurements conducted in Programs Q)
and @ are recorded in Reference 7.

Slow fatigue crack growth portion of the crack growth history

is common to all three geometries. It is characterized by a crack
growth law which relates the rate of crack growth per cycle (da/dN}
to the change in J integral (6J) during a cycle. Catastrophic fail-
ure following slow fatigue crack growth has been observed in the MJ
and SLJ. The criterion for this final mode of failure is that the
shear strength of the adhesive has simply been exceeded due to the
shorter overlap produced by an increase in crack length. Prediction
of the crack growth history is thus obtained by a cycle-by-cycle
LIFE PREDICTION METHODOLOGY 581

Table IX Experimental Data Requirements.

R= 0.1

I MODEL ~K & KTESTS FOR FAILURE "K"vs a' THESE TESTS USED TO FOLLOW
JOINT JOINT CRACK FRONT INVESTIGATION OF ~K & KEFFECT
LOCATION alN) SINCE THIS INTERPRETATION AND
APPLICATION OF I & II TO III DEPENDS
(MJI ON THAT

FROM COMPUTER DETERMINE

K vs CRACK LENGTH
IKI& KII)

II CRACKED (1) CHOOSE h1/h~ SO THAT THE KI2 + KII2 =rIa'

LAP SAME KI & Kn A FOR MJ
SHEAR
RESULTS: DOES IT LEAD TO SAME CHECKED
CRACK GROWTH LAW?
(CLSI
(2) IF CHOICE IS MADE WITH MORE "IF NOT, WHAT WORKS?"
DIFFICULTY THAN THAT, ASK
WHETHER A DIFFERENT K~-Kn
COMBINATION PRODUCES HE
"S~ME" FRACT.URE RATE BY
KI + KII2 = rIa)

III STRUC· NEED STRESS ANALYSISTO DETERMINE da

TURAL LOADS SUCH THAT KI & KII IKI2 + K1I2) i =iii
LAP ARE IN THE SAME BALL PARK AS FOR
JOINT CLS & MJ. THEN PROCEED TO PREDICT a' =!!!
NUMBER OF CYCLES TO FAILURE FOR dn
(SLJI GIVEN LOAD LEVELS -
K =AVERAGE K PER CYCLE
USE DIFFERENT LOAD HISTORY

integration of the crack growth law for the fatigue crack growth por-
tion subjected to the critical crack length for catastrophic failure.

The procedure outlined above combines the finite element stress

analysis, which calculates' the J integral as a function of crack
length, a, and the fracture property measurements, viz., da/dN versus
~J, for a specific joint geometry. This yields a prediction of the
crack growth history, viz., crack length, a, versus number of cycles,
N, as shown schematically in Figure 2. The joint failure point, Nf'
is thus determined where the slope of the a versus N curve approaches
infinity.

Structural Failure Analysis (Module V)

Concerning the contents of this module, if a criterion exists,

i t contains the onset of failure; so problems of structural failure
analysis really become problems of reducing structural loads to
582 J. ROMANKO ET AL.

(a) FINITE ELEMENT ANAL YSIS

J·INTEGRAL

CRACK LENGTH,_

(b) FRACTURE PROPERTIES

LOG da/dN
:~ = C (I1J)b

LOG (I1J)

(e) CRACK GROWTH PREDICTION

11_ =J!! I1N

dN
CRACK LENGTH
N
ai=~.
i =1 I

NUMBER OF CYCLES,N

Figure 2. Crack Growth Prediction Scheme.

LIFE PREDICTION METHODOLOGY 583

local stresses causing bond failure. Once the failure has occurred,
one must determine whether the overall structure will fail. Of
course, the extent of bond crack growth must be considered, including
the question of whether the crack stops.

The inputs and outputs of this module are outlined in Table X.

Finally, the overall integration scheme for service life pre-

diction of an aircraft bonded structure, including the management
plan, is shown in Figure 3.

PRELIMINARY RESULTS

The crack growth history prediction scheme outlined above and

shown schematically in Figure 2 has been computer programmed, with
the code validated in Reference 11 using results obtained from the
fatigue testing of center cracked panels of neat FM-73M adhesive in
the Fatigue Behavior program. 12 Further, preliminary predictions of
lifetimes for the SLJ geometry, and comparisons with the correspond-
ing measured lifetimes, are recorded in Reference 7.

Table X Structural Failure Analysis Module, Inputs and Outputs,

INPUT OUTPUT
(1.) F'AILURE CRITERION MARGIN OF SAFETY: BONO OR JOINT

LET SERVICE LOAD BE L

(A) LIFE
(2.) OVERALL STRUCTURE (Airplane) L GIVES RISE TO iL
AND PROBLEM AREA LOCATION (B) YES OR NO ON FLY
LOOK FOR OVERLOAD Lo
(C) MARGIN OF SAFETY
REDUCTION TO
STRESSES AT
BONDED JDINT } WHICH PRODUCES THE

CRITICAL GROWTH RATE io

AT THE EXISTING GROWTH

(3.) LOADS GEOMETRY

MS=!!-1
L
 (11
00
~

II
MATERIALS OF AIRPLANE MANAGEMENT PLAN
AVERAGE BONO
STRESSES ON
REDUCED GEOM.
.i CRITICAL
YES OR NO

GROSS GEOMETRY OF AIRPLANE

BONO AREA TO BE INTERROGATED MARC VERSION IV FAILURE
FOR BONO Al TERNATE EVALUATION
G, J, K MECHANISMS &
CRITERIA OF JUDGING PLANE SAFE
[ VERIFICATION TEST]
GEOMETRY - SLJ T .. S·I , 9.I OR NOT
II.

V FAIL SAFE
ANALYSIS
VIA
a T,ap,(h h , PI MASTRAN VERSION

I
LOADS I!o ENVIRONMENTS .. 1___..... MATERIALS OF
BON D I RESPONSE
r(i ), r io')
' : w . n _ .. __ • , I TEMP
ENVIRONME NT
1: H20 , P
i ' d./dt
. ' = do ldn
GROUND fliGHT TEMP C,;..
STORAGE & RelATIVE ACCELERATED TESTING
HUMIDITY :D
o
Figure 3 Overall Integration: Service Life Prediction of Adhesively Bonded Joints. s:
»
z
'"om
-t
»
r
LIFE PREDICTION METHODOLOGY 585

The eventual aim of this program is to predict the lifetime of

a circumferentially bonded splice taken from station 703 of the PABST
fuselage (Figure 24 of Reference 13). This real structure is similar
in geometry to the SLJ geometry used in generating the above-mentioned
baseline data.

ACKNOWLEDGEMENTS

This work was sponsored by Air Force Systems Command, Air Force
Wright Aeronautical Laboratory, Wright-Patterson Air Force Base, OH,
under Contract F336lS-79-C-5088, "Integrated Methodology for Adhesive
Bonded Joint Life Prediction." The authors especially wish to ac-
knowledge the interest and guidance of Dr. Wm. B. Jones, Jr., AFWAL/
MLBC project engineer on this program. Mr. L. R. Collins of the
Stress Analysis Group of Materials and Structures Technology, General
Dynamics Fort Worth Division conducted the MARC analyses. Also many
scientists of the Materials and Structures Technology group contrib-
uted to the efforts in this program, including: Messrs. M. E. Tohlen,
F. C. Nordquist, C. P. Fisher, and B. O. McCauley.

REFERENCES

1. Contract F33615-75-C-5178, "Laminated Wing Structures",

sponsored by USAF AFML/FIBC, Wright-Patterson AFB, OR 45433,
and conducted by General Dynamics, Fort Worth Division,
Forth Worth, TX, June 1975 - June 1980.
2. Contract F3361S-76-C-3138, "Advanced Technology Wing Struc-
ture", sponsored by USAF AFML/FIBC Wright-Patterson AFB, OH
45433, and conducted by Vought Corporation, Dallas, TX,
November 1976 - March 1980.
3. F. N. Kelley and J. L. Trout, Paper presented at the AlAA/SAE
8th Joint Propulsion Specialist Conference. Paper No. 72-1085.
New Orleans, LA, Nov. 29-Dec. 1, 1980.
4. Quarterly Progress Report No.5, General Dynamics, Fort
Worth Division Tech. Report, FZM-6952, December 1980.
5. Quarterly Progress Report No.2, General Dynamics, Fort
Worth Division Tech. Report, FZM-6892, March 1980.
6. "Materials Properties Handbook, FM-73", General Dynamics,
Fort Worth Division Tech. Report, FZM-7055. July 1980.
7. John Romanko, K. M. Liechti and W. G. Knauss, "Integrated
Methodology for Adhesive Bonded Joint Life Predictions",
AFWAL-TR-82-4139, (Final Report for Period July 1979 to
July 1982), Wright-Patterson AFB, OH 45433, November 1982.
8. Quarterly Progress Report No.9, General Dynamics, Fort
Worth Division Tech. Report, FZM-7010, December 1981
586 J. ROMANKO ET AL.

9. R. A. Schapery, in "Proc. Japan - U.S. Conference on

Composite Materials", Tokyo, January 1981.
10. K. M. Liechti and W. G. Knauss, 1981 Advances in Aerospace
Structures and Materials, AD-01, p. 51 (1981).
11. Quarterly Progress Report No.8, General Dynamics, Fort
Worth Division Tech. Report, FZM-6990, September 1981.
12. J. Romanko and W. G. Knauss, in "Developments in Adhesives -
2", A. J. Kinloch, Editor, p. 173, Applied Science Publishers,
London, England, 1980.
13. Douglas Aircraft Co., Tech. Report AFFDL-TR-79-3129, Wright-
Patterson AFB, OH 45433, November 1979.
STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS

Ronald A. Kline*

Engineering Mechanics Department

General Motors Research Laboratories
Warren, Michigan 43090

In this study, the effect of adhesive bondline thick-

ness on the stress distribution in an adhesively bonded
joint is examined. A linear variation in bondline stresses
through the adhesive thickness is used to model this effect.
In this way, an important influence on bond behavior, here-
tofore neglected or incompletely modeled, is studied. Also
presented is a parametric study of the effect of various
joint parameters on bondline stress distribution when thick-
ness effects are included.

INTRODUCTION

Adhesive bonding has long been recognized as an attractive

alternative to conventional fastening techniques due to a greater
uniformity in load distribution as well as reduced weight and pro-
cessing ease. This is particularly true for the composite mater-
ials which are being used with increasing frequency in the aero-
space and automotive industries as these materials do not lend
themselves easily to conventional fastening techniques. Thus,
there is a need for a firm understanding of the nature of the
stress states which are found in adhesively bonded joints. Much
research has been directed towards this end.

*New Address:
School of Aerospace, Mechanical and Nuclear Engineering
University of Oklahoma
Norman, Oklahoma 73019

587
588 R.A. KLINE

The basic features of the adhesive bond problem were examined

in the classic analysis of Goland and Reissner1. In their work,
the adherends were assumed to deform as thin plates coupled by an
elastic adhesive layer. Particular attention was paid to the fact
that, in addition to the resulting shear stress (T xz )' an appreci-
able transverse tensile stress or peel stress (ozz) is developed
in an adhesively bonded joint. Also, for single lap shear joints,
the eccentricity of load application produces a significant bend-
ing moment at the joint edges. Later work has extended this re-
search to examine the effects of plasticity in the adherends 2 ,
plasticity in the adhesive layer 3- 5 , viscoelastic adhesives 6 - 7 ,
and anisotropic adherends 8 - 9 •

One of the major drawbacks to these works is that the adhe-

sive layer is assumed to be very thin in comparison to the adher-
end layer and bondline thickness effects are often neglected. The
resulting stress distribution from these models represent the
average of the stress state in the adhesive joint. However,
thickness effects are quite important, particularly if one is con-
cerned with failure mechanisms in composite joints. As is illus-
trated in Figure 1, failure in a single lap joint with composite
adherends initiates at opposite edges of the joint (sites 1 and 3
rather than at opposite edges of the joint (sites 1 and 3 rather
than 2 and 4). Thickness independent bond theories are insensi-
tive to this effect. Ojalvo and Eidinoff 11 recognized this short-
coming and developed a model incorporating this effect. They mod-
ified the Goland-Reissner analysis to include a linearly varying
shear stress distribution through the adhesive layer. Although
exhibiting the desired thickness dependent behavior, this model
has several serious deficiencies associated with it. While allow-
ing shear stress to vary linearly through the thickness, the peel
stress was modeled as being invariant in the adhesive layer.
Poisson's ratio was assumed to be zero in this analysis. This
represents a significant mathematical simplification but is some-
what unrealistic. In addition, moment and horizontal force re-
sultants in the adhesive layer were neglected. As a result, their
solution failed to satisfy equi~ibrium relations for the adhesive
layer. In a later work, carpen'ter12 attempted to modify the
Ojalvo-Eidinoff analysis to reflect these concerns. However,
Carpenter's modifications fail to satisfactorily resolve these
problems. In this work, a new model for determining the stress
distribution in an adhesively bonded joint, incorporating bondline
thickness effects, is presented.

THEORETICAL MODEL

The basic geometry of the single lap shear joint is shown in

Figure 2. Sign conventions are indicated in this figure. The
STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 589

- C[==-_---i-lt ---tIt-'--=::11-
LOW LOAD LEVEL

~il.i_RE_S_S_CO_NC_E_NT_R_A_Tl_ON_
- [-r-====~=~~-rt::::==== :J -
DEFORMA TlON OF ADHERENDS UNDER MODERATE LOAD

_ .-----, -~r--------"1 -
"'NTERLAMINAR" FAILURE OF FILAMENTARY COMPOSITE AOHERENDS

Figure 1. Composite Joint Behavior.

tL,
z

ADHEREND 1 x

ADHESIVE

Figure 2. Adhesive Joint Geometry.

590 R.A. KLINE

origin of the co-ordinate system is chosen so that z = ± ~

corresponds to the adhesive-adherend interfaces and x = O,2c
corresponds to the joint edges.

In this analysis of adhesively bonded joints, the following

assumptions are employed:

1. The adherends and adhesive are linearly elastic. For

simplicity the adherends are assumed to be identical.
(With the proper choice of material constants, the
theory can be easily modified to include differing
adherends.) The constitutive equations for the
adhesive layer, unlike most previous investigations
in the adhesive layer 1 ,3-5,8,9,11, 5,8,9,11,12. This
allows for a non-zero Poisson's ratio which is more
realistic for structural adhesives than the
simplified model used in these studies.

2. The adherends deform as thin plates in cylindrical

bending. A plane strain condition in the xz plane is
assumed. The effects of transverse shear forces and
deformations are included in this analysis.

3. The displacements in the adhesive are assumed to

vary linearly through the adhesive thickness. Most
earlier models assume that these displacements are
constant through the adhesive layer 1 - 10 ,13,14.

The starting point for the development of this model is the

consideration of static equilibrium of a differential element of
each of the joint constituents. The force and moment resultants
(Ni,Qi,Mi) acting on each are illustrated in Figure 3 along with
the shear (Txz)i and peel (azz)i forces which develop at the
adhesive-adherend interfaces. In this analysis the subscripts 1,
2, and a will refer to the upper adherend (or upper adherend-
adhesive interface), the lower adherend (or lower adhesive-
adherend interface) and the adhesive layer, respectively. Static
equilibrium (neglecting width effects) of the upper adherend re-
quires that:

aN1 Horizontal
N1 + - - dx - (T xz ) 1 dx - N1 0 Force Equilibrium
ax
..
N1 = (T xz ) 1 T1 (1)

dQ1 Vertical
Q1 + - - dx - (a zz ) 1 dx - Q1 0 Force Equilibrium
. = (azz ) 1
Q1
dx
a1 (2)
STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 591

It.,
MI + MI' dx

H'~ ADHEREND 'H,' d,

°1
r1 +1
01 + ° 1' dx

Ma+Ma'dx
Ma
rl
Na~ ADHESIVE t t - N a +N'dx
a
rZ ° a + 0'
a dx
°a

MZ + M2' dx
M2 iLrz

"~ I
°2
ADHEREND 2
Itt
.. °
.,'
,02 +
.,'"
2' dx
dx

Figure 3. Balance of Forces for Adhesive Joint Element.

592 R.A. KLINE

- (Lxz )1 dx (tJ + M1 o Moment Equilibrium

Q1 - L1t/2 (3 )

where quadratic terms in dx have been neglected and ~ is used to

. 1 d
d enote th e part1a · .
er1vat1ve a •
ax

In a similar fashion, static equilibrium conditions for the

adhesive and lower adherend require:

Adherend 2

N2 -L2 (4)

Q2 -0"2 (5)

~
t
M2 Q2 - L2 "2 (6)

Adhesive

Na L2 - L1 (7)

Qa 0"2 0"1 (8)

~ ( L1 + L2) h
Ma = ~- (9)
2

From assumption 2, the following relations between the force

and moment resultants and the displacements for identical thin
orthotropic adherends (again neglecting width effects) are known
to be 10

1-\112 \121
u01 N1 (10)
E1t

1-\112 \121
u02 N2 ( 11)
E1t

6
w01 - --Q (12 )
- 5tG31 1 - 13 01

~
6
w02 - --Q
- 5tG31 2 - 13 02
( 13)
STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 593

13 01
.. 12( 1-'V12'V21)
M1 (14)
t3E1

13 02
.. 12(1-'V12'V21)
M2 (15 )
t3E1

where the constitutive equations for the adherend are given by

-'V21 -'V31
e:1 0 0 0 a11
E1 E2 E3
-'V 12 -'V32
e:2 0 0 0 a22
E1 E2 E3
-'V 13 -'V23 1
e:3 0 0 0 a33
E1 E2 E3

e:23 0 0 0 0 0 '23
G23

1
e:31 0 0 0 0 0 '31
G23

e:12 0 0 0 0 0 '12
G12

uoi' woi are the midplane displacement components and 13 0 i are the
average midplane rotations.

For cylindrical bending we can write the displacements at the

adhesive-adherend interfaces (ui' wi' 13i) as 16

t
u1 = U01 + "2 /301 (16 )

-
u2 u02 - "2
t
13 02 ( 17)

-
w1 w01 (18 )

w2 w02 (19)

a1 13 01 (20)

a2 13 02 ( 21)
594 R.A. KLINE

To complete the model, it is necessary to consider the con-

stitutive relationship for the adhesive layer. For convenience,
La~e notation is used for the adherends. From assumption (1) we
have

( aua + aWa) (22)

'xz ~a az ax

(23)

Using assumption 3 and the compatibility of the displacements

at the two interfaces, we can write the adhesive displacements in
terms of the adherend displacements as

ul + u2 z
ua +- ( ul - u2) (24)
2 h

wl + w2 z
wa = +- (Wl - w2) (25)
2 h

Combining Equations (22-25) yields

ul - u2
'1 'xzl-h/2 l1a + wl (26)
h

ul - u2
'2 'xzl-h/2 l1a + w2 (27)
h

- .. wl - w2
0"1 <1zzlh/2 Aa u l + (A + 211 a ) h
(28)

<12
-.
Aa u 2 + (A a + 2~a)
wl - w2
(29)
O"zzl-h/2 h

These 32 equations can be algebraically reduced to 15 independent

equations which must be solved to determine the stress distribu-
tion in the bonded joint subject to appropriate boundary condi-
tions.

The necessary boundary conditions were determined from the

applied loads at the ends of the joint; this procedure yields:
STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 595

BCl Ql -Qo

BC2 Ml -Mo

BC3 Nl No at x=c

BC4 Ql = 0

BC5 Ml 0

BCG Nl 0 at x=c

BC7 Q2 0

BCa M2 0

BC9 N2 = 0 at x=c

BC10 ~ 0

BCll Ma 0

BC12 Na 0 at x=-c

and the need to fix the displacements at one point to eliminate

rigid body motions which yield:

BC13 ul 0 at x=-c
BC14 wl 0
BC15 13 1 0

The remaining boundary conditions follow from equilibrium and

symmetry considerations.

This problem was solved for a variety of jOint configurations

using an iterative technique in conjunction with the method of
shooting points 17 • Thickness dependent terms in the governing
equations were multiplied by an iteration variable a. With a = 0,
the thickness independent solution (a la Goland and Reissner 1 et
al.) was obtained. Using this solution as the basis for the next
iteration, the parameter a was increased slightly and the next
solution was obtained. This process was repeated until a = 1,
where the full problem was recovered and the final solution
determined.
596 R.A. KLINE

One principal drawback to this method is that the resulting

shear stress distribution is not zero at the joint edges as re-
quired. This deficiency is common to virtually all linearized
plate theory analyses of the problem, including that of Goland and
Reissner1. Renton and Vinson8 ,9 have shown that this shortcoming
can be eliminated using a somewhat more elaborate model for adher-
end behavior with a parabolic rather than linear variation in ad-
herend shear stress through the thickness. Their results show
that, instead of reaching a maximum at the joint edges as predict-
ed in Goland and Reissner type analyses, the actual stress dis-
tribution reaches a maximum a short distance from the edge of the
joint and decreases rapidly to zero at the edge. As computing
space limitations precluded incorporation of this aspect of bond
behavior, this fact should be taken into account in interpreting
the results presented here.

RESULTS AND DISCUSSION

Bond stress predictions based on the model described above

are presented in Figures 4-6. For comparison, the values of the
°
nondimensionalized peel (No;~C) and shear (-'xz
/ 2) stresses are
No c
shown as a function of the normalized positio~ = 7. Stress
distribution along both the upper ('1,01) and lower ('2,02) bond-
lines are presented to illustrate the variation of stress through
the adhesive thickness. The symmetric nature of the problem is
also reflected in the solution with '1 (x) = '2(2-x), 01(x) =
02(2-x). The basic parameters used to characterize joint geometry
and material composition are shown in Table I. These parameters
correspond to a typical joint used in automotive applications.

Qualitatively, there is a great deal of similarity between

the stress distributions predicted by the earlier models and that
found in this study. Both shear and peel stress distributions are
roughly parabolic with the maximum stresses being developed at the
joint edges and minimum stresses being observed in the central
region of the joint. Thus, the major portion of the joint exper-
iences relatively light loads while small regions at the joint
edges are more severely stressed. It is precisely in these high
stress regions at the joint edges that the principal differences
in the various models emerge. While peak shear stress predictions
are quite similar for the models presented, larger differences in
peel stress maxima were observed principally due to the more com-
plete constitutive equation used in this study for the adhesive
layer.
STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 597

o
•
F£EL SlllfSS

~
,.~----~----~----~------~--~
0.0 0.3 0.. 0.1 0.1 1.0
DlSTfN:E

...o

..,.
o
"'.~~------~---------r---------r---------.---------'
0.0 o.a 0.. 0.1 0.1 1.0
DlSTfIa
Figure 4. Stress Distribution in Adhesively Bonded Joints (Low
Modulus Adhesive).
598 R.A. KLINE

rm.. STIBS
..'"
rr SIG'VI I

SIC:fl' 2.

.....

"!
I+----------r---------,----------~--------,_--------_,
0. 0 0. 2 O. t 0.. 0.. 1. 0
DISTANCE

..
0

.....
SHEAR STI£SS

'"!
VI-
VI
W
~
II>
,.

..
0

0
d
0. 0 O. ~ 0 •• 0. 1 0.. 1. 0
DISTANCE

Figure 5. Stress Distribution in Adhesively Bonded Joints (High

Modulus Adhesive).
STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 599

~
PEa STRESS
...0

en o
•
en
~
en
0

.
~

..
~
0.0 o.a O.t 0.. I.. 1.0
DISTANCE

0
N
SffAR STRESS

:l

en':
..
en
~
en
• TAU 2 J
Ii

•0

0
d
0.0 o.a O.t 0.' I.. 1.0
DISTANCE
Figure 6. Stress Distribution in Adhesively Bonded Joints (Thick
Adhesive Layer).
600 R.A. KLINE

TABLE I

Parameter ~ Value

Adhesive Thickness h .038 cm (.015")

Adherend t .635 cm (.250")

OVerlap Length 2c 2.54 cm (1.00" )

Tensile Modulus (Adherend) 13.24 GPa t (1.92 x 10 6 psi)

Poisson' Ratio (In-Plane, Adherend) 0.25 t

Poission's Ratio (Out-of-Plane, Adherend) Y13'Y31' 0.25°

.
Shear Modulus (Adherend)
Lame Constant (Adhesive)
G31
Aa
0.53 GPa* (7.6 x 10 5 psi)
.230 GPa~ (3.25 x 10 4 psi)
Lam~ Constant (Adhesive) ~a .0259 GPa~ (3.76 x 10 3 psi)

t R. Heimbuch and B. Sanders, "Mechanical Properties of Chopped Figer Reinforced

in 'Composi te Materials in the Automotive Industry," pp. 111-139, American

Society Mechanical Engineers, 1978.

* Measured ultrasonically using pulse-echo superposition technique.

T. Wang, B. Sanders and U. Lindholm, "A Loading Rate and Environmental Effects

Study of Adhesively Bonded SMC Joints," GMMD Report Number 80-044, General

Motors Manufacturing Development, Warren Michigan, 1980.

Estimated.
STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 601

Stress variation through the thickness of the adhesive was

found to be highly dependent on the joint geometry and properties
of the materials used. Typical results are presented in Figure 4.
At the edge of the joint, the difference in peel stress between 01
and 02 was approximately 9% for a relatively thin adhesive (h =
0.010"). The variation in shear stress was somewhat less, approx-
imately 4%. Utilization of a higher modulus, higher strength
adhesive tends to increase the stress variation. This is illus-
trated in Figure 5 for the same joint as in Figure 4 except that
the higher modulus adhesive (Ea = 1.02 x 10 5 psi) has been used.
The peel stress variation was found to increase in this case to
approximately 17% with a corresponding increase to 9% in shear
stress. Extreme variations in stress distribution were observed
for joints with thick adhesive layers. Of course, this is not
surpr~s~ng. Figure 6 illustrates this for the case of the base
joint with h being increased from 0.010" to 0.250". Under these
conditions the peel stresses differ by an order of magnitude at
the joint edges. The variation in shear stress was significantly
less. It is not known how accurate these results are for extreme
cases such as this, as the assumption of a linear variation in
stress through the adhesive thickness may no longer be applicable.
However, it does indicate that the variation in peel stress can be
considerable for certain geometries.

It is important to note that the effect of this variation in

stress on composite joints is significant for predicting the fail-
ure site in a bonded joint. Thickness independent models predict
that failure is equally likely to occur at any of the joint edges
(sites 1,2,3,4 in Figure 1). This is in contrast to the predic-
tion of the thickness dependent models which indicate that sites 1
and 3 should be the preferred fracture initiation sites.

Experimentally, this is precisely what is observed. Figures

7(a-d) follow the behavior of a single lap shear joint (SMC com-
posite) during the course of a test from the beginning of the test
through failure. The initial specimen damage appears (at opppo-
site edges as predicted using this model unlike thickness indepen-
dent models) as small matrix cracks in the composite at the joint
edges (Figure 7b). As the load is increased, the cracks grow in a
highly non-uniform manner, pricipally via fiber-matrix debonding.
This is attributable, in all likelihood, to the weakness of the
fiber-matrix interface 23 • In any event, this behavior is indica-
tive of the high stresses (particularly peel) which develop at the
joint edges and the inability of composite materials to withstand
out-of-plane loads. For bonds with metal adherends, a somewhat
different sequence is observed 19 • The onset of fracture again
occurs at adhesive-adherend interfaces at opposite edges of the
joint. However, due to the (statistically) isotropic nature of
the polycrystalline metal, they are not as susceptible to adherend
peel stress failure as the composite material. Failure is confin-
602 A.A. KLINE

Figure 7. Failure Mechanism in Adhesively Bonded SMC Composite.

STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 603

ed to the adhesive-adherend interface (adhesive failure) or in the

adhesive layer itself (cohesive failure) depending on their rel-
ative strengths.

A parametric study of the effect of pertinent geometric and

material factors was conducted to examine possible methods of re-
ducing the high stresses which develop at the edges of the bonded
joint. The results of this effort are illustrated in Figures 8-12
where nondimensionalized bond stress distributions for variations
in adhesive modulus (E a ), adhesive thickness (h) adherend thick-
ness (t) and joint overlap length (2c) are shown. Adhesive modu-
lus was found to have the most pronounced effect of the four
parameters investigated, the use of a lower adhesive modulus re-
sulting in a large decrease in peak stress values at the edge of
the bonded joint. Peak stress reduction can also be effected by
decreasing the adhesive overlap length and increasing adherend
thickness. Increasing adhesive thickness tended to reduce peak
stresses provided that h « t . This is illustrated in Figures 11
and 12 for a relatively thin adherend (t = 0.100") and a relative-
ly thick adherend (t = 0.250"). For adhesive thicknesses compar-
able to or greater than the adherend thickness, increasing the ad-
hesive thickness increased the peak stresses due to the increase
in the applied bending moment Mo = No (t + h)/2. It should also
be mentioned that applied load can also affect the stress distri-
bution observed due to the bending of the joint. Increasing load
on a joint tends to reduce the peel stress observed as illustrated
in Figure 13. This figure represents the stress distribution in
the basic composite joint at three load levels (1 lb., 100 lb,
1000 lb). The reduction in peel stress at the joint edge was
found to be approximately 13%. It should be mentioned that the
peak stresses which develop at the ends of a bonded joint are only
one factor in determining overall joint efficiency. The effect of
changing any of these parameters on the strength of the adherend,
the adhesive and the adherend-adhesive interface must be taken
into account in determining whether the performance of the bonded
structure will be improved. In any event, it should be stressed
that it is erroneous to conclude that simply because a given joint
fails in the adherend, it is well designed. Conversely, it is not
necessarily true that failure of the adhesive layer in a bonded
joint implies that the joint is poorly designed. In assessing the
performance of a given joint, many inter-related factors (stress
distribution, constituent strengths, etc.) must be considered.
This is particularly true for fiber reinforced composites due to
their inherent weakness when subjected to out-of-plane loads.
with proper choice of bonding agent and specimen geometry, it is
felt that improvements in the efficiency of adhesively bonded
joints can be achieved.
604 R.A. KLINE

'"
'"
EA = 1 X 105 PSI

1 X J.dl PSI

II!
j"
o.D D.a D•• 0.' 0.. 1.0
DISTANCE

D
9£AA STIBS
'"
..
...

i~ 1 x 103 PSI
II)

•d

D
d
o.D La D•• D.' 0.. I.D
DISTNCt

Figure 8. Effect of Adhesive Modulus on Stress Distribution.

STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 605

.
0

ffil. S1RESS

0
.;
T = .(8)"

0 .150"
~~ .250"

0
0

0.. 0..
~
j'
IU o.t 1.0
DlS1flC£

o
.;

...•
T = .rIJJ'

o
d+-________
0.0
~------~~------~--------~--------~
0.. 0.. 0..
11.3 I.'
DISfIIN:E

Figure 9. Effect of Adherend Thickness on Stress Distribution.

606 R.A. KLINE

..
..: Fill SllBS

.
til

C = .250"
.
~M .50)"

,.
.
1 0.11
II." 0•• ..I .. I I.'
DJS1ftNC£

SIfAA S1Tf.SS

·'" C".~

.50)"

l.(xr

•d

•
dt..~.--------r-------~~------~---------r--------~
0.3 0.. ..I 0.1 I.'
DJS1ftNC£

Figure 10. Effect of Adhesive Overlap on Stress Distribution.

STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 607

..!!
..
a

B . .010"
PEEL STRESS

. 050"

~:i .100"
VI
a
,.{

..
d

a
~
0.0 0.3 0.. 0.1 0. 1 1.0
DISTPICE

;:i

SHEAR STRESS
..• H . .010

.050"
•
~~ . 100"

•d

D
d
La a.a 0. 4 0. 1 1.1 1.0
OJ sma

Figure 11. Effect of Adhesive Thickness on Stress Distribution

(Thin Adherend t = 0.100") •
608 R.A. KLINE

.•
.
of

.. H= .010"
.(60"
;~
.250"
~

....
"!
j"
0.0 La L. L. L. I.'
DIS1PICE

.
.c
SI£AR STfBS
•
'"
H" .0lD"

~~
.(E()"
.250"
lin

.
d

D
d
0.0 La I.. L. D.I 1.0
DlsnICE

Figure 12. Effect of Adhesive Thickness on Stress Distribution

(Thick Adherend t = 0.250").
STRESS ANALYSIS OF ADHESIVELY BONDED JOINTS 609

1 LB.
o
,.;

~
i+---------,----------r---------r--------~--------_,
0.0 0.2 0.4 0.. 0.. 1.0
OISTfla:

Figure 13. Effect of Applied Load on Stress Distribution.

ACKNOWLEDGMENT

The author wishes to thank his colleagues Martin R. Barone,

Robert L. Frutiger and Mark F. Nelson for their many helpful tech-
nical discussions throughout the course of · this work.

REFERENCES

1. M. Goland and E. Reissner, J. App. Mech., l!., A17 (March

1944).
2. M. Chang, J. Adhesion, 4, 221 (1972).
3. L. Hart-Smith, "Analysis and Design of Advanced Composite
Joints", NASA 1974. Available through NTIS.
4. L. Hart-Smith, "Adhesive-Bonded Double Lap Joints", NASA
CR-112235, 1973. Available through NTIS.
5. L. Hart-Smith, "Adhesive-Bonded Single Lap Joints", NASA
CR-112236, 1973. Available through NTIS.
610 R.A. KLINE

6. Y. Weitsman, J. Adhesion, 11, 279 (1981).

7. F. Delale and F. Erdogan, ~ Appl. Mech., 48, 331 (1981).
8. W. Renton and J. Vinson. J. Appl. Mech., 4~ 101 (1977).
9. W. Renton and J. Vinson, J. Adhesion, 7, 175 (1975).
10. F. Delale, F. Erdogan and M. Aydinoglu: J. Composite Materials,
12.., 249 (1981).
11. I. Ojalvo and H. Eidinoff, AIAA J., 16, 204 (1978).
12. W. Carpenter, AIAA J., 18. 350 (1980):"
13. U. Yuceoglu and D. Updike. J. Eng. Mech., 106, 37 (1980).
14. U. Yuceoglu and D. Updike, J. Eng. Mech., 107, 55 (1981).
15. W. Ewing, W. Jardetzky and F. Press. "Elastic Waves in
Layered Media", McGraw-Hill Book Company, 1957.
16. S. Timoshenko and S. Woinowsky-Krieger, "Theory of Plates and
Shells". 2nd ed., McGraw-Hill Book Company, 1959.
17. G. Dahlquist and A. Bjorck, "Numerical Methods", Prentice-
Hall, Inc., 1974.
18. R. Kline. "The Effect of Microstructure on the Mechanical
Properties of SMC", paper presented at ASTM Symposium on
Producibility and Quality Assurance of Composite Materials ,
St. Louis, MO. 1981.
THE IMPACT STRENGTH OF ADHESIVE LAP JOINTS

J.A. Harris and R.D. Adams

Department of Mechanical Engineering

University of Bristol, University Walk
Bristol, BS8 lTR, U.K.

Structural adhesive joints may be required to

withstand high loading rates. The effect of impact
loading on the strength and energy absorption of single
lap joints has been investigated. A rubber-toughened
epoxy adhesive, in bulk form, shows significant changes
in mechanical response, when tested in shear over a wide
range of strain rates. Despite large changes in the
response of the adhesive, the strength of lap joints is
only slightly reduced by impact loading. When high
strength adherend materials are used, energy absorption
is relatively small; however, it is increased by using
lower strength adherend materials, as significant plastic
deformation can take place in the adherend prior to
joint failure.

The finite element method has been used to model

and predict joint strength for the various conditions
from a knowledge of the bulk adhesive mechanical response
under those conditions.

611
612 J.A. HARRIS AND A.D. ADAMS

INTRODUCTION

With the development of modern structural adhesives, the use

of adhesive bonding as a joining technique has become increasingly
attractive. In particular, the joining of aluminium alloys is
made possible where the use of conventional spot welding methods
is not feasible on a high volume production basis. One possible
area of application is in the automotive industry, where the use
of aluminium alloys and plastics for car bodies represents a con-
siderable weight saving advantage. This is the background to the
problem that is being addressed here, that is, that the automobile
designer must be confident that structural bonded joints will have
sufficient strength under normal loading conditions and also ret-
ain that strength under high loading rate conditions, as would
occur in a crash situation. The energy absorbing capability of
the structure to protect the passenger is important here, and the
view is taken that joints must retain sufficient strength such
that the main volume of material that is between them will plasti-
cally deform and buckle. Hence, the two parameters of interest
are strength retention and energy absorption under impact.

The feasibility of using adhesives in structures in which the

joints may be required to be highly deformed or crumpled under an
impact load has been investigated l • Cylinders were fabricated
from sheet material by forming a tube with an overlapping seam
running the length of the cylinder. The seam was either bonded or
spot welded, and the tube then crushed axially under impact con-
ditions in a drop weight rig. Both cylinders of steel and
aluminium alloy which were tested showed no significant difference
in energy absorbing performance between bonded and spot welded
seams. The bonded seams were fractured at various points in the
crumpled zone because of the continuous nature of the joint,
whereas the spot welds remained intact, because of the flexibility
of the joint between the welds.

The majority of test work that has been carried out on

adhesives and adhesive joints has been at low loading rates or
'quasi-static' conditions. It is thus necessary to know what
effect high loading rates or impact conditions have on adhesives
and adhesive joint strength. The existing ASTM standard test for
the impact testing of adhesive bonds (ASTM D 950-72) was not
suitable. The test, which consists of shearing off a steel block
bonded to another steel base block, is difficult to reproduce in
the form of a quasi-static test for comparison and, because of the
high stiffness of the specimen used, very much higher loading
rates result than would be experienced in a realistic bonded
structure, which would be very much more compliant. Thus, the
single lap joint test piece was chosen to be tested under impact
conditions, as it was reasonably compliant and was already commonly
used as a test piece for quasi-static testing of adhesives.
THE IMPACT STRENGTH OF ADHESIVE LAP JOINTS 613

The problem of understanding the effects of high loading

rates on bonded joints is part of a more general problem of being
able to predict adhesive joint strength. The approach taken here
is that if the mechanical properties of an adhesive are knownar,d
there exists a suitable analysis of the critical stresses for a
particular loaded joint then, with a suitable failure criterion,
joint strength can be predicted. In order to gain an understand-
ing of the effects of impact loading on an adhesive joint, it was
therefore necessary first to investigate the effects of loading
rate on the adhesive as a bulk material. Then, using the finite
element method, the analysis of the joint could be carried out for
both quasi-static and impact loading, and compared with test
results.

EXPERIMENTAL

Adhesive properties

A rubber modified epoxy material has been used, which was a

diglycidyl ether of bisphenol-A, with 15 parts (by weight) per

lal

.
. IL...-p-)I--====~I(---...JI - .~.
Ibl tOmm
~

Figure 1 Bulk adhesive specimens for quasi-static torsion tests.

(a) Annular; (b) Solid.
614 J.A. HARRIS AND A.D. ADAMS

hundred of resin of carboxyl-terminated butadiene-acrylonitrile

rubber and 5 parts per hundred of piperidine as the curing agent.

The rate dependent properties of the material were

investigated by a series of tests, in pure shear, of bulk specimens
machined from cast blocks of the adhesive. Over a range of rela-
tively low strain rates, specimens were tested on a variable speed
torsion testing machine. Two types of bulk specimens were tested:
an annular specimen and a solid circular specimen, both shown in
Figure 1. For the solid specimen, when a torque is applied,
because the shear stress is greatest at the outside, yield of the
material begins here and spreads inwards through the radius. The
torque-twist curve from this specimen can be converted to a shear
stress-strain curve by applying the correction due to Nadai 2 ;
however, this is only approximate since it is not applicable to
rate sensitive materials. The use of an annular specimen avoids
the use of the Nadai correction as the variation of shear stress
across the annulus is small. Also, the shear strain rate varies
little across the section. However, these specimens failed due
to buckling of the gauge length so the ultimate shear strain could
not be derived. In these tests, torque was measured by a strain
gauged load cell, and the shear strain was measured by a rotary
capacitative transducer. The twist was transmitted to the trans-
ducer by strings passing over pulleys attached to the specimen
at each end of the gauge length. On this machine strain rates in
the range 10- 3 min- 1 to 1 min- 1 were achieved.

For tests at higher rates of strain, a shear impact type

device was used. This together with the instrumentation used is
illustrated in Figure 2. The operation of the machine was based
on the running up of a flywheel to a controlled speed. A spring-
loaded arm was then released to positively engage into the fly-
wheel resulting in the application of an abrupt torque which
in turn was transmitted through the specimen to the load cell
attached to a rigid base. Again, an annular specimen was used as
shown in Figure 3. It has castellated ends by which it is
located into the cruciform ends of the load cell and loading shaft.
The shear strain at the outside of the specimen was measured from
strain gauges bonded to the surface. In order to record the
torque and shear strain signals over the short periods of testing,
a two-channel transient recorder system was used, from which the
stored data could be 'replayed' on a flat bed plotter. On this
instrument strain rates in the range of 10 2 min- 1 to 10 3 min- 1 were
achieved.

Joint strengths

Single overlap shear specimens were manufactured to dimensions

as specified in ASTM D1002-72 for the standard 25.4 mm wide,
12.7 mm overlap length test piece, using the rubber modified epoxy
THE IMPACT STRENGTH OF ADHESIVE LAP JOINTS 615

RELEASE LOADING MOTOR AND

LOAD CELL SPECIMEN ARM SHAFT FLYWHEEL BELT DRIVE

/
,--_-<A......._---.,

e
I I
TORQUE STRAIN OS CILlOSCOPE

00
I
AMPLIFIERS
L _ _ _J I
L- _ _ _ _ _ J
I

TRANSIENT
I

IL. ____J I
L ____ ..J
, i

X-V PLOTTER
RECORDER

Figure 2 Impact torsion test rig.

616 J.A. HARRIS AND R.D. ADAMS

. ,·t·,' .
~
Figure 3 Bulk adhesive specimen for torsion impact test.

adhesive. A glue line thickness of 0.125 rnm was maintained

throughout. Initially, the adherend material used was a high
strength aluminium alloy designated 2L73 which has a 0.2 per cent
proof stress of 430 N rnm- 2 ; but, subsequently, lower strength
aluminium alloys designated BB2hh and BB2s were used with 0.2 peT
cent proof stresses of 220 N rnm- 2 and 112 N rnm- 2 respectively.
Batches of six specimens were tested at low and high loading rates.
For the low rate tests, a screw driven tensile testing machine was
used with a crosshead speed of 2 rnm min-I. A typical time to
failure of the 2173 joints was 4 minutes, with an almost linear
rise in the load to failure. For the lower strength adherends,
after an initial linear rise, the loading rate rapidly decreased
as the adherends yielded and considerable plastic deformation
occurred in the adherends prior to failure which, for the lowest
strength adherend, typically took 10 minutes to occur.

For the high loading rate tests, an existing Izod pendulum

impact machine was modified to incorporate the geometry of the
single lap specimen. At one end, the specimen was clamped into a
piezo-electric force link which was rigidly attached to the base
of the machine. At the other, a block clamped on to the specimen
carried a bar which lay across the direction of the pendulum swing.
The impact load was thus applied by an impactor attached to the
pendulum, which had two striking heads which picked up on each side
of the bar. In order to prevent dynamic oscillations being set up
during a test, a layer of lead was interspersed around the impact
bar which provided a cushioning effect in the initial part of the
impact loading. As a result, reasonably linear loading to failure
could be achieved. The energy absorbed in the test was measured
from the load-end displacement response of the specimen, the end
THE IMPACT STRENGTH OF ADHESIVE LAP JOINTS 617

displacement being measured using a capacitative displacement

probe. For large energy absorption, the loss of pendulum energy
derived from the loss of height of the swing was used. Again,
the two-channel transient recorder system was used to measure the
dynamic load and displacement. For the 2L73 specimens, an impact
velocity of about 1 ms- 1 resulted in a time to failure of about
5 ms. For the lower strength materials, higher impact velocities
were required to fracture the specimens, such that for the BB2s
specimens an impact velocity of 4 ms- 1 resulted in a time to
failure of about 10 ms.

Finite element analysis.

The use of the finite element method in the analysis of the

single lap joint enables more accurate modelling of the joint to
be made compared with existing closed form analyses. Adams and
Peppiatt 3 and Crocombe and Adams 4 have used the finite element
method to model the single lap joint, including the geometry of
the adhesive fillet formed at the ends of the overlap. The
inclusion of the spew fillet served to significantly reduce the
stress concentration in the adhesive at the ends of the overlap
so that, compared to closed form analyses with a square edged
adhesive layer, more realistic stress predictions resu1ted 4 • The
direction of cracks in failed lap joints was perpendicular to the
predicted maximum tensile stresses running down from the corner
of the unloaded adherend at an angle of about 4SDto the adherend
surface. It was conc1uded 3 that failure of a lap joint is 1n1t1-
ated by tensile failure of the adhesive within the spew fillet.

In this analysis, it was required to include non-1inearities

due to the yielding and plastic deformation of the adherend mater-
ials, and also due to the rotation of the overlap under load which
makes the problem one of large displacements. Adams, Coppenda1e
and Peppiatt S have also shown, using the finite element method for
double lap joints, that, in order to predict lap joint strength,
the elastic-plastic response of the adhesive must be included in
the analysis.
The finite element program FELDEP, written by Dr. A.D.
Crocombe, was used for the analysis. In this both large displace-
ment effects and material non~linearities could be included. The
program is based on the incremental theory of plasticity. The
yield criteria used were the Von Mises criterion for the adherend
and a modified Von Mises criterion for the adhesive. The modified
yield criterion is of the form

F = ( J1 (S-l) + (Ji (S-1)2 + 12 J 2 S)~) /2S - YT =0

where Jl and J 2 are the first and second stress invariants, YT the
618 J.A. HARRIS AND R.D. ADAMS

yield stress in uniaxial tension and S the ratio of the yield

stress in uniaxial compression to uniaxial tension. A value of
1.3 for S is assumed, which is typical of a rubber-modified epoxy
type material. The adhesive properties for both quasi-static
loading rates and impact loading rates are represented by uniaxial
tensile stress strain curves, which have been measured. As has
been noted, failure in the adhesive occurs under the action of
primarily tensile stresses. For this rubber modified epoxy, a
significant amount of ductility is shown; thus, a failure criterion
based on a maximum principal strain is assumed and is made equal
to that measured in the uniaxial tensile test. Under quasi-static
conditions, this is 0.15 and, under impact conditions, 0.06.

The analysis was carried out for the high strength adherend
joints for both quasi-static loading and impact loading, and for
the lower strength materials under quasi-static loading.

RESULTS AND DISCUSSION

Adhesive properties

In Figure 4, the shear stress-strain curves for the range of

strain rates over ·which the tests were carried out are shown. As

50
~ - ------- - __ - - --6UO~IN-'

~/ _________ =-_ - --7"

....... ::::.--- - ____ - - - ---1470MIN-'
MIN-'

~ ---21'~IN-'

40
,,~

~
- - ___________ O·~·_~I.!.I~ __ ou ~IN-'
_ - - - - - - _ _ _ _ _ _ _ _ _ _ _ _ _ _ ~J~~2_~.!..N:: _ 0:0'6 ~IN -'
:?.: .,," 0 ,010 MIN-'

30 ~I ,,/. ~ '" 0,0015 ~IN"

~
"
:;a."
011
;;% II

~
- - STATIC TUBUlAR SPECIMEN
~ 20 - - - -- STATIC SOLiO 5PECI~EN5

"I'
- --IM PACT TUBULAR SPECI~ENS
- TEST STOPPED

~~--~--~~~~~~~--~--~
o 0.04
0.02 0.06 0.08 0.10 0.12 0.14 0.16
Shur strain

Figure 4. Shear stress-strain curves for bulk adhesive over range

of strain rates from various tests.
THE IMPACT STRENGTH OF ADHESIVE LAP JOINTS 619

50

"

v
08
v
10

oL--------------------------------c----~
-2 0
I N Strain rate (MIN ..... '

Figure 5 Variation of shear modulus and yield stress of bulk

adhesive with logarithm of strain rate.

can be seen, the rubber-modified epoxy shows significant rate

dependent properties over the complete range of strain rates.
Both the initial shear modulus and maximum flow stress are in-
creased by increased strain rate, indicating the material response
to consist of an initial visco-elastic region and a visco-plastic
region after yielding. In materials such as this the distinction
between elastic and plastic deformation is not well defined, thus
the definition of a yield point is not straightforward. Some
polymers deforming at a constant strain rate have a stress strain
curve which passes through a maximum point which often approxi-
mates to the onset of permanent plastic deformation. 6 Here, there-
fore, this maximum point, often defined as the "intrinsic yield
point" is assumed to be the yield point for the material. In
Figure 5 the values of initial shear modulus and shear yield stress
are plotted against the natural logarithm of the strain rate for
the annular specimens. Both show a reasonably linear relationship
which, for the yield stresses, is in agreement with the flow
model of Eyring. 7

Another effect of increasing strain rate is to reduce the

strain to failure of the material. The failure strain of the
solid specimens at a rate of 0.04 min- l is 0.55 compared with a
value of 0.175 at a rate of 1470 min- l in the impact test.

Thus, the effect of impact rates of loading on the mechanical

properties of the adhesive material is significantly to ~ncrease
the shear modulus, yield stress and failure stress, and to decrease
the failure strain.
620 J.A. HARRIS AND A.D. ADAMS

Joint strength

The results of the lap joint strengths from both the quasi-
static and impact tests are shown in Figure 6 for each of the
adherends. As can be seen, the effect of impact loading on each
of the joints is similar in that joint strength is reduced. For
the high strength 2L73 adherends, the reduction is about 11 per
cent, for BB2hh adherends 6 per cent,and for BB2s adherends 4 per
cent. Thus, the effect of the impact loading would appear to be
less significant for lower strength adherends but in all cases the
reduction is small.

The data from the impact tests have been replotted in Figure 7
as a function of the proof stress of the adherend material. Also
shown is a line indicating the onset of plastic deformation in the
adherend under the combined tension and bending that the adherend
undergoes. This has been calculated from the results of the
analysis of the single lap joint configuration by Goland and
Reissner 8 which, for the adherends, is based on plate bending theor~
Also shown is a line indicating the load at which the cross-section

,,.---,,
MEAN-I I

r----. ---l--"f SCATTER

16 : : --1 BAND

14

12

10
0
<t
0
...J

W z 8
0:: .x
::::>
...J

«
u.
6

0
2L73 BB2hh BB2s

ADHEREND MATERIAL

Figure 6 Comparison of single lap joint strength between static

and impact loading, for various adherend materials.
THE IMPACT STRENGTH OF ADHESIVE LAP JOINTS 621

16

12

..
o
9 10

~~
3
~ ,/
/
/
/

--------
-------"-'-" --'-'-~
100 200 JOO
o2 PERCENT PROOF STRESS OF ADHEREND
Nrnm-2

Figure 7 Variation of single lap joint strength under impact

conditions with adherend proof stress.

150 t AOHERENOS
I '\ o U7l
I I D BB'2ht!
-.......... I I 6 882s
,,' I
'J I

1 '~1~.
100 I ..'\

>
/ /\ I \

\
C)
0:
'"
Z

'" 50

-- '.\t"
o
.,.".
,/

o 10 12 16
FAILURE LOAO
kN

Figure 8 Variation of the energy absorbed by single lap joints

under impact conditions with joint strength for
various adherends.
622 J.A. HARRIS AND R.D. ADAMS

of the adherend will reach the proof stress under uniaxial tension.
Thus it can be seen that, in each case, some plastic deformation
must take place due to bending in the adherends, and for the BB2hh
and BB2s materials considerable plastic deformation in the adher-
ends takes place prior to failure. This was confirmed by the sig-
nificant elongation of the failed specimens that was observed.

The failure loads from the impact tests are again replotted
with the energy absorbed by the joint for each of the tests in
Figure 8. These results indicate that energy absorption of
adhesive joints is very much dependent on the adherend material.
Since the fracture energy of the adhesive itself is relatively
small, significant energy absorption can only be achieved through
plastic deformation of the adherend materials. It would appear
that the joints can withstand large amounts of adherend plastic
deformation, and thus the result of using a lower strength adherend
material is to enhance the energy absorbing capability of the joint,
despite the fact that joint strength is significantly reduced.
The rapid increase in energy absorbed with failure load for the
BB2hh and BB2s materials is again an indication of the large
plastic deformation that is occurring.

From the surfaces of failed joints, it was observed that for

both the quasi-static and impact tests, the locus of failure was
different for the high strength 2L73 adherends as compared with
that of the lower strength materials. For the high strength
adherends, failure initiated with a crack in the fillet o~ adhesive
at the edge of the overlap, running down approximately 45 to the
adherend surface as in Figure 9 (a). For the lower strength
adherends, failure was initiated at the very edge of the adhesive
fillet running along the interface, as in Figure 9 (b). For
convenience, these will be referred to as failure types A and B.
This change in failure mode is assumed to be attributable to the
plastic deformation of the adherends.

Finite element analysis

In Table I, a summary of the finite element predictions is

compared with the practical results obtained. In Figure 10, the
results of the analysis for the high strength adherend under quasi-
static loading is illustrated by the development of the plastic
zone within the adhesive layer. Initial yielding takes place at
only 1.5 KN load and, before failure, the whole of the adhesive
layer has yielded. Peaks in the stress and strain fields occur
around points 'a' and 'b' which, as can be seen, are points where
initial yielding takes place. For this particular joint, the
maximum strains occur around point 'a', thus leading to the pre-
diction of a type A failure. The same is true for the impact
joint also. For these high strength adherend joints, although
THE IMPACT STRENGTH OF ADHESIVE LAP JOINTS 623

,------------

(a)

(b)

Figure 9 Locus of failure for various joints: (a) type A failure;

(b) type B failure.

KEY
1·5kN
4·0kN
6·0kN
S·OkN
lO·OkN
12·0kN

Figure 10 Finite element prediction of the growth of the plastic

zone through the adhesive layer of a single lap joint
with applied load.
624 J.A. HARRIS AND R.D. ADAMS

Table I. Comparison of Finite Element Predictions and Experimental

Results.

Finite Element Experimental

Adherend Predictions Results
Rate
Type Load (KN) Type Load (KN)

2L73 Impact A 10.3 A 14.3

2L73 Static A 14.7 A 15.9

BB2hh Static B 11.6 B 10.1

BB2s Static B 8.1 B 7.9

some plastic deformation of the adherend takes place (a maximum of

about 2 per cent strain in the quasi-static joint at failure) this
does not have a significant effect on the state of stress in the
adhesive layer.

By applying the 0.15 strain to failure criterion to the quasi-

static joint, a fairly accurate prediction of joint strength is
made. For the impact joint, the 0.06 strain to failure does
correctly predict a reduction in the joint strength because of the
change in adhesive properties due to the change in loading rate.
However, the prediction is somewhat low, perhaps because of error
in the value for maximum strai~or because, for much lower strains
in the impact case, the maximum strain to failure criterion is not
as applicable.

In the analysis of the quasi-static joints with lower

strength adherends, significant adherend plasticity takes place
which has two effects on the adhesive stress field. Firstly, the
rotation of the overlap is much enhanced, as the bending deforma-
tion in the adherends is increased. This results in a reduction
i.n the offset of the applied load at the edge of the overlap and
hence a reduction in the bending moment at the edge of the overlap,
leading to a reduction in the concentration of stress around point
'a' in Figure 10. Secondly, as adherend strains become large,
particularly around the region at the edge of the adhesive fillet
where the adherend has tensile bending stresses superimposed
on the applied tension, so then the adhesive around region 'b ' in
Figure 10 is constrained by the adherends such that the strains
build up rapidly with applied load. As a result of this, the type
B failure is predicted to occur with lower strength adherends, at
a reduced load depending on the yield strength of the material.
THE IMPACT STRENGTH OF ADHESIVE LAP JOINTS 625

Again using the maximum adhesive failure strain criterion, reason-

able predictions of the failure loads for the lower strength adher-
end quasi-static joints are made as indicated in Table I.

CONCLUSIONS

The effect of a realistic rate of impact loading on a realis-

tic bonded joint has been investigated for a rubber-toughened
epoxy with adherends of various strengths.

The effect of strain rate on the adhesive properties has been

demonstrated by the shear stress-strain response of the adhesive
over a range of strain rate of 10- 3 min- l to 10 3 min-I. Both the
initial modulus and yield strain are increased linearly with the
natural logarithm of the rate, and the failure strain is siginfi-
cantly reduced by higher rates of strain. Thus, the effects of
impact loading rates on the strength of bonded joints is attrib-
uted to the siginifcant change in response of the adhesive material.
In practice, only about a 10 per cent decrease in joint strength,
due to impact loading, was measured with high strength adherends;
thus, as far as joint strength is concerned, the reduction in
ductility of the adhesive is compensated for to a large extent
by the increase in yield and failure stress. For lower strength
adherends, even smaller reductions in strength occur, for similar
reasons.

The joints, both under quasi-static and impact loading rates,

could sustain considerable amounts of adherend plasticity prior to
failure. In fact, the lower the adherend strength, the greater
the energy absorbing capability of the joint even though joint
strength is reduced. Thus, it would appear that adhesive joints
are suitable for the joining of structures from which high energy
absorption is required; and, for a given adhesive, the energy
absorbing capability is strongly dependent on the adherend
strength.

The finite element method is an important tool in the analysis

of bonded joints, as both geometric and material non-linearities
can be included in the model. Using this method, reasonable pre-
diction of joint strength and the effects of impact loading on
it have been made. Through the prediction of the change in the
position of failure initiation in the joint due to the adherend
plasticity, the effeGt of using lower strength adherends on joints
strength can be assessed. Experimental evidence is in keeping
with these predictions.

From a knowledge of the adhesive properties, together with an

analysis of the stress-strain state that the adhesive is under in
a loaded joint, a greater understanding of the mechanism of failure
626 J.A. HARRIS AND R.D. ADAMS

in bonded joints can be achieved. By application of a suitable

failure criterion the effects of various parameters on failure load
have been predicted, thus, to some extent, verifying the validity
of this approach to the analysis of bonded joints.

REFERENCES

1. J. A. Harris and R. D. Adams. Unpublished data, 1981.

2. A. Nadai, "Plasticity. A Mechanics of the Plastic State of
Matter", pp. 128-130, McGraw-Hill, New York, 1931.
3. R. D. Adams and N. A. Peppiatt, J. Strain Anal., ~, 185,
(1974).
4. A. D. Crocombe and R. D. Adams, J. Adhesion, 13, 141, (1981).
5. R. D. Adams, J. Coppendale and N. A. Peppiatt:-;--in "Adhesion
2", K. W. Allen, Editor, pp. 105-119, Applied Science
Publishers, London, 1978.
6. D. B. Bowden, in "The Physics of Glassy Polymers", R. N.
Howard, Editor, pp. 279-339, Applied Science Publishers,
London, 1973.
7. H. Eyring, J. Chem. Phys., 4, 283, (1936).
8. M. Goland and E. Reissner, J. Appl. Mech., Trans. ASME, 66,
A17, (1944).
THE PERFORMANCE OF ADHESIVE-BONDED THIN-GAUGE SHEET METAL

STRUCTURES WITH PARTICULAR REFERENCE TO BOX-SECTION BEAMS

A. Beevers and A. C. P. Kho

Oxford Polytechnic
Headington
Oxford, U.K., OX30BP

In an effort to reduce vehicle weight, the auto-

motive industry is developing car body structures made
from light-weight materials such as composites, plastics
and aluminium alloys. Fabrication of these materials
in high volume automotive applications using traditional
welding techniques is not feasible and adhesive bonding
is now being investigated as a potential assembly method.
In order to assess performance characteristics of bonded
vehicles, thin-gauge sheet-metal box-section beams have
been used to simulate structural details in automotive
applications such as car bodies and commercial vehicles.
Beams were fabricated from flanged strips by different
joining methods to form box-section structures approx-
imately 1m. long x 6Omm. square. Tests were carried out
to determine torsional and flexural rigidity and ultimate
flexural strengths, and in the majority of tests,bonded
structures gave better characteristics than the equivalent
riveted or spot-welded beams. The failures of beams under
3-point bending have been related to buckling of the
side webs and further experimental tests have shown that
collapse is critically dependant on flange-bend radius.
Finite element techniques have been used to analyse
stress distribution in the beam section and this confirms
the experimental observations of beam collapse.

627
628 A. BEEVERS AND A.C.P. KHO

INTRODUCTION

The motor industry has for some time used adhesives in semi-
structural applications particularly for the attachment of stiff-
eners to bonnets and boot lids. As well as providing a uniform
load distribution, adhesive bonding provides a smooth surface
finish and thus eliminates the need for expensive secondary oper-
ations prior to painting. More recently the feasibility of using
adhesives in fully structural automotive body applications has been
studied. Although the concept of a bonded, unitary construction
car body is applicable to conventional mild steel assembly, it is
potentially mOre relevant to the fabrication of bodies from light
alloy and composites. These materials are becoming increasingly
attractive for automotive applications because of their light
weight but they cannot be joined by the traditional welding methods.
Bonding technology is ,therefore ,expected to make a significant
contribution to future developments in the motor industry and a
considerable amount of research effort has already been committed
to establish characteristics of bonded joints for such structures.

A large volume of data has been established on properties of bonded

joints using relatively small test pieces 1-4 and new analytical
techniques are providing a greater understanding of bonded joint
performance 5,6 However, there is little information available
on the performance of structurally bonded large scale elements
which might be encountered in automotive applications. It is
envisaged that box sections and "top hat" stiffener configurations
will feature prominently in fully-stressed body shells and these
forms of structural elements are also applicable in other types of
vehicle construction such as transport containers, coaches,
commercial vehicle cabs, and rail carriages. The work described in
this paper was carried out to determine flexural and torsional
properties of box-section beams formed from thin gauge sheet metal.

The general design of the fabricated beam structure adopted for the
tests was chosen to give a reasonable representation of a typical
detail of a car body such as a sill or door pillar. It was expected
that the test method would also provide a simulation of a variety
of other sheet metal structures and the external flange arrangement
enabled the beam to be made by a number of different joining methods.

BOX SECTION BEAM TESTS

Beam Manufacture

Each beam was manufactured to the same nominal dimensions using

four pieces of sheet metal 910mm long by 90mm wide. To form the
'box', two pieces were shaped into a shallow channel section by
bending 15mm flanges on to the edges of the strip using standard
workshop bending bars. The beam was completed by joining the edges
ADHESIVE-BONDED THIN-GAUGE SHEET METAL STRUCTURES 629

Figure 1. Configuration of box-section beams fabricated from thin

sheet (dimensions in mm).

of the flat strips to the flanges so that the main section formed
a 60mm square, as shown in Figure 1. The beams were made from
0.9mm thick deep drawing quality low carbon mild steel and 1.2mm
thick 2.5per cent Mg. aluminium alloy.

For the adhesive bonded structures all the surface areas to be

bonded were grit blasted and thoroughly cleaned and degreased with
trichloroethylene. The adhesives were then applied uniformly to
both surfaces. In the case of the toughened acrylic adhesive,
the resin was applied to one surface and the activator to the other
joint surface. The parts were then located, assembled together
and held with spring clips while the adhesive cured. The single-
component high-temperature-curing, epoxy adhesive was cured at 180°C
for 30 minutes, while all the other adhesives were cured at room
temperature.

For the spot-welded and riveted beams the welds and rivets were
placed at 25mm intervals along the centre line of the flange using
standard practices as far as possible. The rivets were standard
3.lmm diameter steel-bodied blind rivets which were inserted with a
pneumatic/hydraulic rivet-setting power tool.

Flexural stiffness was determined by three-point bend tests as

shown in Figure 2. The beam was supported by two radiused supports
630 A. BEEVERS AND A.C.P. KHO

710mm apart on the bed of a compression testing machine and the load
was applied at a third "centre" point on the top of the beam by a
similar radiused anvil. To reduce the risk of sheet indentation
and premature collapse small steel pads were placed between the
beams and the support radii. During testing the bending load was
applied incrementally at intervals of IkN and the midspan deflection
was measured by means of a dial test indicator.

Torsional stiffness was measured by mounting the ends of the beams

into special end fixtures which rigidly clamped the four sides of
the box-section to a central core. The fixtures were then located
into grips on the torsion testing machine which enabled the
torsional load to be measured for incremental angles of twist at
0.25 degree intervals.

Ultimate strength tests were carried out using the same machines
and fixtures by continuing the loading beyond yield until failure
occurred.

Beam Test Results

Although there are some inconsistencies in the results of the beam

tests in Table I, it is possible to draw a number of general
observations and conclusions. The most striking observation is
that with only one exception the adhesive bonded structures were
superior to the equivalent spot-welded and riveted beams in all
test performance parameters and in some cases the stiffnesses and
ultimate strength properties were more than doubled.

It is also clear that there is little correlation between the

performance of a structure and the performance of a small lap shear
strength test piece. In the case of bending (flexure) tests the
shear force on the joint interface is relatively low and in all
tests failure of the beam occurred by buckling of the inner webs.

Figure 2. Arrangement for flexure tests (dimensions in mm).

ADHESIVE-BONDED THIN-GAUGE SHEET METAL STRUCTURES 631

Table 1. Summary of Results on Beam Tests.

BEAM FLEXURAL TORSIONAL ULTIMATE

MATERIAL JOINING METHOD STIFFNESS STIFFNESS FLEXURAL
kN/mm kNm/rad STRENGTH
kN

Rivet 1.8 2.5 3.4

Spot Weld 1.9 9.5 3.6

2 Part, Cold-Cured
0.9mm Epoxy Adhesive 2.9 11.6 7.7
Mild
Steel 1 Component, Hpt-
Cured Epoxy Adhesive 4.2 8.2 6.9

Toughened Acrylic
Adhesive 2.85 11.0 5.0

Riv~t 1.6 2.6 7.0

Spot Weld 1.9 5.3 6.4

1.2mm 2 Part, Cold-Cured

AI. Epoxy Adhesive 2.8 6.7 8.2
Alloy
(2.5% 1 Component, Hot-
Mg) Cured Epoxy Adhesive 3.1 11.1 8.5

If the section of the web is considered as a simple strut it will

be realised that the buckling, or crippling load is a function of
eccentricity of load application and rigidity of end-constraint.

This,in turn,will depend on the bend radius which is introduced

during the forming of the flanges. The mild steel sheet naturally
assumes a greater bend radius during forming because of its higher
yield strength and thus in the ultimate flexural strength tests the
mild steel beams will buckle more easily.

Extending this hypothesis it may also be reasoned that in all the

adhesive bonded beams the strut end constraints are more rigid
because the adhesive support extends up to, and beyond, the centre
line of the web as it forms a fillet in the corner of the box-
section. Compared with the more remote attachment of a spot-weld
or rivet this additional support may explain greater flexural
strength of bonded beams.
632 A. BEEVERS AND A.C.P. KHO

Theoretical Analysis of Flexural Stiffness

The deflection, 0, of a beam under three-point bending with the load

at mid-span is given by the expression

° WL 3
48EI
where W = applied load
L = length of beam between supports
E = modulus of elasticity of beam material
I second moment of area of beam section about
neutral axis
The flexural stiffness is then
48EI
(1)
L3

In an idealized beam model it may be assumed that the flange/plate

connections are homogeneous and continuous and the second moment of
area I, of the section can be calculated on this basis. At the
other extreme, if there were no joint between the flanges and plates
the only shear flow from the top sheet into the webs would be by
friction. The section would then approximate to a double channel
configuration with no top and bottom sheets and the value of I can
be similarly calculated. It might be expected that the stiffness
of a fabricated beam would be between these two theoretical
conditions.

Taking L = 7l0mm as used in experimental tests, = 2l0GN/m 2

Es t ee 1
and EAl = 70GN/m 2 , the values of I and flexural stiffness, wiG,
calculated from equation 1 are given in Table II. These theoretical
values of stiffness are compared with experimental test results in
figure 3.
It can be seen from the bar charts in figure 3, that the bonded
aluminium beams give values of flexural stiffness which are close
to the idealized model of the homogeneous beam. The spot-welded
and riveted beams in both steel and aluminium have stiffnesses
which are closer to the beam model having no flange/sheet connection.
It is evident that the discrete point attachments allow some local
elastic buckling of the top sheet in compression giving slight
interfacial separation between each joint. This would cause a
greater deflection of the welded or riveted beams compared with the
bonded structures in which the joint is continuous. In the case of
the riveted structures some interfacial slippage during flexure may
also occur due to the clearance between the rivet body and the hole,
and this may also account for the low values of torsional stiffness
obtained in the experimental tests in Table 1.
ADHESIVE-BONDED THIN-GAUGE SHEET METAL STRUCTURES 633

Table II. Theoretical Stiffnesses of Box-Section Beams.

Theoretical Sheet I Theoretical Stiffness

beam model thick- kN/mm
ness mm 4
Aluminium Steel
mm xl0 6

Homogeneous 0.9 0.227 2.15 6.39

Continuous
Connection 1.2 0.303 2.84 8.53

No connect- 0.9 0.078 0.72 2.14

ion
1.2 0.112 0.94 2.83

STIFFNESS RATIO 1.0

Theoretical beam
continuous, homogeneous joint
hot-cured, one-part epoxy
I
adhesive I
Toughened acrylic
adhesive
I
Spot Weld I
Rivet 1
Theor. beam
No joint
I 0.9mm Mild steel

STIFFNESS RATIO
Theoretical beam
continuous, homogeneous joint
Hot-cured, one-part epoxy
adhesive
Toughened acrylic
I
adhesive J
Spot weld
J
Rivet

Theor. beam
I
No joint I 1. 2mm AI. allo y
Figure 3. Comparison of beam stiffnesses with theoretical models.
634 A. BEEVERS AND A.C.P. KHO

Figure 4. Beam section compression test piece (dimensions in mm).

Table III. Results of Beam Section Compression Tests.

SHEET THICK- FLANGE ULT. COMPo STRENGTH kN

MATERIAL NESS BEND
mm RADIUS ADHESIVE RIVET SPOT
mm BOND WELD

2.5 1.03 0.51 0.51

0.9 1.5 1.18 0.53 0.73
1.0 1.47 0.72 0.80
A1.Alloy
(2~%Mg)
2.5 3.12 1.09 1.12
1.2 1.5 2.87 1.19 1.49
1.0 2.80 1. 63 1.55

3.0 1.89 0.54 0.59

Mild 0.9 2.5 1.94 0.58 0.76
Steel 2.0 2.94 1.11 1. 25
ADHESIVE-BONDED THIN-GAUGE SHEET METAL STRUCTURES 635

In all the tests on the mild steel beams the ratio between the
stiffness of the experimental beams and the idealized models are
relatively lower than for aluminium beams. This may be partly
attributed to the localised deflections occurring at the flange
bend radii which are greater in the mild steel beams.

BEAM SECTION COMPRESSION TESTS

In order to verify the suggested explanation for the ultimate

flexural strength characteristics a further series of tests was
carried out on specimens which were, in effect, a short length of
beam section as shown in Figure 4. Bending of the flange was
carried out so that a range of flange bend radii was obtained and
the strips were then joined using the same methods as in the
fabrication of beams. The adhesive used was an aluminium-filled,
one-component, hot-curing, epoxy resin.

The sections were subjected to compressive loading between two

parallel plates and the ultimate strengths were recorded. The
results of these tests are given in Table Ill.

In all the tests failure occurred by the inward collapse of the side
plates. With the riveted and spot welded sections the smaller bend
radii gave higher strengths than those with the larger radii. These
observations support the inference that the flange bend radius has
a significant effect on the buckling strength of the structure.
With the adhesive bonded structure, the buckling strength is not so
dependant on the flange bend radius. This again supports the earlier
observations that the adhesive fillet provides additional support
at the corner radius of the formed flange thus giving greater
resistance to buckling. It should also be noted that all the
adhesive bonded specimens give considerably higher strengths than
the equivalent rivet and spot-welded structures.

Finite Element Analysis of Section Compression Tests

The finite element stress analysis package progrannne, "Pafec 75"

was used to determine the distribution of stress in a box-section
structure under a compressive load. A comparison of the maximum
stresses sustained in the side plates of structures with different
flange radii was investigated.

Since the box-section structure was axisynnnetric about both the 'x'
and J y ' axis, only a quarter of the structure was considered in this
analysis. The eight-noded isoparametric element {code number 36210
in Pafec} was used to divide up the structure and the nodal diagram
for a typical section is shown in Figure 5 •
636 A. BEEVERS AND A.C.P. KHO

20
21
22
23
2~

Figure 5. Nodal diagram of typical section for finite element

analysis <r/t = 1).

Analysis was made of section models without adhesive and with

adhesives of varying values of elastic moduli. The calculated
stresses resulting from applied compressive loads were used to
provide a measure of buckling resistance; the higher the maximum
stress in any model, the lower would be its resistance to buckling.
The analytical results are shown in Table IV, in which the critical
dimensional features of the finite element model are expressed as a
ratio between the flange bend radius and the sheet thickness, rIte
In the ideal case the bend radius would be zero, i.e. r/t = 0 and
the tabulated values of buckling resistance are given relative to
this ideal flange condition.

Table IV. Buckling Resistance of Sections from Finite Element

Analysis.
Without With
r/t Adhesive
Adhesive

0 1 -
0.4 0.314 0.418

1.0 0.285 0.37

1.6 0.225 0.3

ADHESIVE-BONDED THIN-GAUGE SHEET METAL STRUCTURES 637

The finite element analysis of sections without adhesive shows a

substantial loss of compressive strength as the value of r/t
increases. In practice r/t>l. It is also seen that the bonded
section models give greater buckling resistance, and although their
compressive strengths are still dependant on the r/t ratio it is
evident that the adhesive fillet has contributed to the improved
performance.

In comparing these values with the results of the compressive tests

in Table III, it might be suggested that the riveted and spot-
welded sections are equivalent to the finite element models without
adhesive. Similar trends are observed in the experimental values
but the adhesive-bonded sections are proportionately better than
predicted from the finite element analysis.

CONCLUSIONS

The flexural and torsional stiffness and strength properties of

bonded thin-sheet box-section beams are higher than for similar
beams formed by riveting or spot welding.

The improved ultimate flexural strength properties of bonded beams

are partially due to the supporting effect of the adhesive fillet
at the flange bend radius.

ACKNOWLEDGEMENTS
Support of this work by Permabond Adhesives Ltd. is gratefully
acknowledged.

REFERENCES

1. J. D. Minford and E. M. Vader, "Adhesive Bonding Aluminium Body

Sheet", Society of Automotive Engineers Paper No. 740078 (1974).
2. G. E. Nordmark, "Fatigue Performance of Aluminium Joints for
Automotive Applications", Society of Automotive Engineers Paper
No. 780328 (1978).
3. S. J. Thompson, "Investigation Into the Impact Properties of
Adhesive Bonded Joints", Oxford Polytechnic postgraduate thesis
(1979) •
4. W. A. Lees, Int. J. Adhesion Adhesives, 1, 5 (1981).
5. G. P. Anderson, S. J. Bennett and K. L. DeVries, "Analysis and
Testing of Adhesive Bonds", Academic Press, New York, 1977.
6. A. J. Kinloch, Editor, "Developments in Adhesives - 2", Applied
Science Publishers, London, 1981.
CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES

S. Mall,* W. S. Johnson, and R. A. Everett, Jr.**

NASA Langley Research Center

Hampton, Virginia 23665

To analyze the fatigue behavior of a simple

composite-to-composite bonded joint, a combined experi-
mental and analytical study of the cracked-lap-shear
specimen subjected to constant-amplitude cyclic loading
was undertaken. Two bonded systems were studied:
T300/5208 graphite/epoxy adherends bonded with adhe-
sives EC 3445 and with FM-300. For each bonded system,
two specimen geometries were tested: (1) a strap
adherend of 16 plies bonded to a lap adherend of
8 plies, and (2) a strap adherend of 8 plies bonded to
a lap adherend of 16 plies. In all specimens tested,
the fatigue failure was in the form of cyclic debonding
with some 0 0 fiber pull-off from the strap adherend.
The debond always grew in the region of adhesive that
had the highest mode I (peel) loading and that region
was close to the adhesive-strap interface. Furthermore,
the measured cyclic debond growth rates correlated well
with total strain energy release rates GT as well as
with its components GI (peel) and GIl (shear) for
the mixed-mode loading in the present study.

*University of Maine, on leave of absence at NASA Langley Research

Center.
**Structures Laboratory, U.S. Army Research and Technology Labora-
tories (AVRADCOM), NASA Langley Research Center.

639
640 S. MALL ET AL.

INTRODUCTION

To increase performance and fuel economy, aerospace industries

have been turning more and more to the use of advanced composites
in both commercial and military aircraft. These materials offer
excellent strength-to-weight and stiffness-to-weight ratios. But
their efficient application requires more sophisticated joining
methods than used in metallic structures. Because composites are
severely weakened by fastener holes, their weight advantage may be
lessened when mechanically fastened joints are used. Adhesive
bonding, on the other hand, provides a desirable alternative to
mechanical fastening because of the following potential advantages:
(1) higher joint efficiency index (relative strength/weight of the
joint region), (2) lower part count, (3) no strength degradation
of basic laminate due to fastener holes, (4) less expensive and
simpler fabrication techniques, (5) lower maintenance costs, and
(6) potential corrosion problems avoided.

Design methods for adhesively bonded composites require cri-

teria to predict both strength and durability. Although analytical
and experimental work has been reported on the static strength of
bonded composites,l-3 very little information is available on their
fatigue behavior. Several possible fatigue failure modes exist for
bonded composites: cyclic debonding (i.e., progressive separation
of the adhesive bond under cyclic load), interlaminar damage
(delamination), adherend fatigue, or a combination of these. There-
fore, life predictions require a basic understanding of the mechan-
ics associated with each failure mode.

Many of the results obtained in cyclic debonding studies of

composite-to-metal joints 4 and metal-to-metal joints 5 may be appli-
cable to the present case of composite-to-composite joints. In the
study by Roderick et al.,4 the fracture mechanics concept of strain
energy release rate was used to model the cyclic failure of bonded
composite-to-metal joints. This is similar to the approach in
metals where fatigue-crack-propagation rate is correlated with the
strain energy release rate. The total strain energy release rate,
GT, associated with the cyclic failure of an adhesive bond can be
resolved into three components GI , GIl' and GIll associated
with three debonding modes: I (opening), II (sliding), and III
(tearing), respectively. However, in most practical applications,
only GI and GIl' due to peel and shear stresses, respectively,
exist near the debond front during cyclic.loading. Using metal-to-
metal joints, Brussat et al. 5 developed the cracked-lap-shear speci-
men to study the effect of mixed-mode (G I and GIl) loading on
adhesive joints. Romanko and Knauss 6 have extended these fracture
mechanics concepts to investigate the failure of adhesive joints
under various environmental conditions involving temperature,
moisture, etc. Everett 7 recently showed that the strain energy
CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES 641

release rate GI associated with peel stress had a significant

effect on cyclic debonding.

The objective of the present study was to analyze the fatigue

behavior of simple composite-to-composite bonded joints subjected
to constant-amplitude cyclic loading. For this purpose, graphite/
epoxy (T300/5208) cracked-lap-shear specimens were tested using
EC 3445 and FM-300 adhesives. This investigation focused on the
correlation of the measured cyclic debond rates with strain energy
release rates calculated using a finite element analysis of the
cracked-lap-shear specimen. Failure modes of this simple joint
were also analyzed to investigate the relative influence of GI
and GIl'

NOMENCLATURE

a length of debond, mm
da
debond growth rate, mm/cycle
dN
B width of specimen, mm

c,n curve-fit parameters

E Young's modulus of adhesive, GPa

Young's moduli of composite, GPa

G shear modulus of adhesive, GPa

shear modulus of composite, GPa

mode I strain energy release rate, J/m 2

mode II strain energy release rate, J/m 2

mode III strain energy release rate, J/m 2

total strain energy release rate (= GI + GIl)' J/m 2

N number of cycles

P applied load, kN

r residual from least-square curve fit

Poisson's ratio of adhesive

Poisson's ratios of composite

642 S. MALL ET AL.

~---------- 2~m m ------------

STRAP ADHEREND

~------------ ~lm m --------------~

Figure 1. Cracked-lap-shear specimen.

EXPERIMENTS

Test Specimen

The cracked-lap-shear specimen, shown in Figure 1, was employed

in the present study because it represents a simple structural joint
subjected to in-plane loading. Both shear and peel stresses are
present in the bond line of this joint. The magnitude of each com-
ponent of this mixed-mode loading can be modified by changing the
relative thicknesses of strap and lap adherends. 5 ,8 For the present
study, the strap and lap adherends had 8 or 16 plies. These adher-
ends had a quasi-isotropic layup and were typical of those currently
employed in an Army program to build an all-composite helicopter
airframe which is almost entirely adhesively bonded. 9

Two bonded systems were studied--graphite/epoxy (T300/5208)

adherends bonded with EC 3445 adhesive and with FM-300 adhesive.
The EC 3445 adhesive is a thermosetting paste adhesive with a cure
temperature of l2l oC. Specimens with this adhesive were fabricated
using conventional secondary bonding procedure. However, specimens
with FM-300 adhesive were fabricated by a co-curing procedure whereby
adherends were cured and bonded simultaneously. The FM-300 is a
modified epoxy adhesive supported with a carrier cloth with a cure
temperature of l77 oC. These adhesives, as well as the concepts of
CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES 643

Table 1. Adhesive Material Properties.

Modulus, GPa Poisson's Ratio

Adhesives
E G \J

EC 3445 1.81 0.65 0.4

FM-300 2.32 0.83 0.4

secondary and co-cure bonding, are also being employed in the 9

bonded all-composite helicopter airframe mentioned previously.
The bonding process followed the manufacturer's recommended proce-
dures for each adhesive. The nominal adhesive thickness was 0.10
and 0.25 mm for the EC 3445 and FM-300., respectively.

The Young's modulus of FM-300 adhesive was calculated from the

shear modulus provided by the manufacturer, assuming the adhesive
to be an isotropic material. Poisson's ratio was assumed to be 0.4
for both adhesives which is a typical value for adhesives. The
EC 3445 adhesive is the paste version of the AF-55 adhesive film.
Therefore, the Young's modulus of EC 3445 was calculated from the
shear modulus of AF-55,10 assuming the adhesive to be an isotropic
material. The material properties of both adhesives are given in
Table I.

The composite adherends consisted of quasi-isotropic lay-ups

of [0 0 /45 0 /-45 0 /90 0 ]8 and [0 0 /45 0 /-45 0 /90 0 ]28' The material proper-
ties of graphite/epoxy, presented in Table II, were obtained from
reference 11. For each bonded system, two types of specimen were
tested: (1) thin lap adherend of 8 plies bonded to thick strap
adherend of 16 plies, and (2) thick lap adherend of 16 plies bonded
to thin strap adherend of 8 plies. This arrangement provided four
sets of specimens. Initially, the lengths of strap and lap ad her-
ends were 381 and 254 mm, respectively (a total of 127 mm was for
grip support on both ends). However, specimens with a thick co-
cured lap adherend were modified due to a pinched-off edge obtained
during fabrication. This pinched-off edge, as discussed in refer-
ence 12, caused a nonuniform thickness of lap adherend and adhesive
near the end of the overlap. To avoid problems due to this non-
uniform thickness, the pinched-off ends were removed by machining.
In the resulting modified specimen, lengths of strap and lap adher-
ends were 305 and 203 mm, respectively.

Testing Procedure

All specimens were tested in a closed-loop hydraulic test

machine at a frequency of 10 Hz. In all tests, constant amplitude
644 S. MALL ET AL.

Table II. Graphite/Epoxya Adherend Material Properties.

Modulus,b GPa Poisson's Ratio b

El E2 G12 \112 \123

131.0 13.0 6.4 0.34 0.35

aT300/5208, fiber volume fraction is 0.63.

bThe subscripts 1, 2, and 3 correspond to the longi-
tudinal, transverse, and thickness directions,
respectively, of a unidirectional ply.

cyclic loads were applied at a stress ratio of 0.1. The debond

initiated at the lap end. Subsequent cyclic debonding was moni-
tored throughout each test.

MEASUREMENT TECHNIQUES
t9
140
o PHOTOELASTI C
o COMPLIANCE
tf>
120 0 EC 3445
t:.
c=:
OX- RAY
~O I
100
0 TH ICK STRAP
DEBONO 80
LENGTH,
8
mm ~
60

:mss +ST~SS
~
40 0

20
@~ 144 MPa 185 MPa

OS

0 .3 .6 .9 1.2 x 106
CYCLES

Figure 2. Comparison of different techniques to measure debond

growth.
CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES 645

Three techniques to measure the cyclic debond growth were

evaluated. For each technique, the location of the debond front
was measured periodically to calculate debond growth rates. The
first method used a sheet of photoelastic material bonded to the
lap adherend of the specimen, as discussed in reference 4. Iso-
chromatic fringes developed at the debond front as a result of the
high strain gradient in that vicinity when subjected to load. These
isochromatic fringes were observed through a polarizer and were used
to locate the debond front. The second method involved locating
the debond front with an X-ray technique using a dye penetrant,
zinc iodine. The third method involved measuring the compliance of
the specimen and then calculating the debond length using a crack
length compliance formula. The compliance of the specimen was
measured with two displacement transducers attached on opposite
sides of the specimen. All three methods provided good agreement
for debond growth measurement as shown in Figure 2. In the present
study the photoelastic technique was selected; an automated measure-
ment system photographed the isochromatic fringes at predetermined
intervals.

150
STRESS
246 MPa~
I
125
I FM-300

I
213_ F? I
DEBONO 100 TH I CK STRAP
LENGTH. 193_
mm 75
172 __ /
50

25 ~/
o 2 4 6
CYCLES

Figure 3. Typical variation of debond length with fatigue cycles

at different stress levels.
646 s. MALL ET AL.
The strain energy release rates (G T , GI , and GIl) are usually
uniform in the cracked-lap-shear specimen for a significant debond
growth region. 8 The debond data were measured over this region.
This region is discussed in the subsequent section on finite-element
analyses. Tests were conducted at two or more constant amplitude
stress levels to obtain several values of debond growth rate (da/dN)
from each specimen. At each stress level, the debond was measured
as it grew over about 3 to 4 cm to ensure an accurate estimate of
the debond growth rate over that region. Figure 3 shows a typical
debond data for different stress levels. In all cases debonding
was initially nonlinear, but became linear after the debond pro-
gressed a short distance (about 10 mm). For each stress level, the
debond data were fitted with a straight line using regression
analysis.

LAP
STRAP

Figure 4. Debonded surfaces of cracked-lap-shear specimen (EC 3445,

thick strap).
CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES 647

ADHESIVE

Mi SSING 0" FIBERS

0" FIBERS FROM STRAP

STRAP LAP

Figure 5. Debonded surfaces of cracked-Iap-shear specimen (FM-300,

thin strap).

Debond Surfaces

As previously mentioned, the possible failure modes in a com-

posite bonded joint are cyclic debonding, delamination, adherend
fatigue, or a combination of these. All specimens in the present
study failed primarily by cyclic debonding of the adhesive.
Although some adhesive remained on both the strap and lap adherends,
significantly more adhesive remained on the lap adherend. Also,
there was some 0 0 fiber pull-off from the strap adherend in most
specimens. Typical debonded surfaces with these failure details are
shown in Figures 4 and 5 for both systems. On close examination of
the debonded surface, the following conclusions regarding the cyclic
debond can be drawn. The debond was basically of a cohesive nature,
i.e., failure was within the adhesive with some 0 0 fiber pull-off
from the strap. Further, the debond was always closer to the strap
than to the lap. A possible explanation for this debond character-
istic is discussed in the next section.
648 S. MALL ET AL.

FINITE-ELEMENT ANALYSIS

Previous studies of fatigue damage mechanisms in adhesively

bonded joints have shown that the strain energy release rate,
defined from fracture mechanics principles, may be useful for cor-
relating cyclic dgbond growth rate. 4 ,5 Therefore, all four sets of
specimgns were analyzed with a finite element program, called
GAMNAS to calculate strain energy release rates. This two-
dimensional analysis accounted for the geometric nonlinearity asso-
ciated with the large rotations in the unsymmetric cracked-lap-shear
specimen. The importance of a geometric nonlinear analysis for the
cracked-lap-shear specimen has been discussed in detail in refer-
ence 8.

A typical finite-element mesh, shown in Figure 6, consisted of

about 1600 isoparametric four-node elements and had about 3000
degrees of freedom. Each ply of composite was modeled as one layer
in the finite-element model, except for the ply at the adhesive
interface which was modeled in two or three layers. A multipoint
constraint was applied to prevent rotation of the loaded end of the

-r-fD • •
FINITE ELEMENT MESH IY COORDINATES ARE MAGNIFIED 20XI

ADHESIVE
T H rS

4 - LAYER MODEL 8 - LAYER NDDEL 12 - LAYER NDDEL

MESH AT DEBONO TIP

Figure 6. Finite element model.

CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES 649

model (i.e., all the axial displacement along the ends are equal) to
simulate actual grip loading of the specimen. A double-cracked-lap-
shear specimen with isotropic adherends was analyzed using a quasi-
three-dimensional analysis like that in reference 13 to show that
the present two-dimensional analysis should be based on the plane
strain condition. The strain energy release rates GT • Gr , and
Grr in the analysis were computed for the maximum load in the
fatigue cycle using a virtual crack closure technique. 14 The
details of this procedure are given in reference 8. The calculation
of strain energy rates in the present analysis depends on three
parameters: (1) debond location, (2) load level, and (3) debond
length. These parameters will be discussed next.

As mentioned previously, the debond grew near the interface of

the adhesive and strap adherend in all tests. rt was very difficult
to measure the exact location of the debond, but in general the
debond grew within one-fourth of the thickness of adhesive closest
to interface. The effect of the location of the debond within the
adhesive in the cracked-lap-shear specimen was investigated using
GAMNAS. 8 Figure 7 shows the variation of calculated GT, Gr, and
Grr with the debond location within the thickness of the EC 3445
adhesive. These results were calculated by modeling the adhesive

~I - - 4 lAYER MO DEL
THICK STRAP - - 8 lAYER MODEL
----12 lAYER MODEL
LAP
ADHEREN[I
.10

4, 8, 12
ADHES IVE lAYER
TH I CKNES S, .05 RESULTS
mm

o
STRAP 23 24 3 4 5 6 17 18 19 20 x 10- 3
ADHEREND NORMALIZED STRAIN ENERGY RELEASE RATE GB/P

Figure 7. Variation of strain energy release rate with location of

debond within EC 3445 adhesive.
650 S. MALL ET AL.

c=: FM-3oo
1200
THICK STRAP

1000

800
G,
2
J/m 600

400

200

o 200 400 600 800 1000

APPLIED LOAD/UNIT WIDTH, kN/m

Figure 8. Variation of strain energy release rates with applied

load.

with 4, 8, and 12 layers of elements, as shown in Figure 6. Fig-

ure 7 clearly shows that Gr remains constant for all locations of
the debond, while Gr has its maximum value near the adhesive-strap
interface and Grr has its maximum value near the adhesive-lap
interface. The debond always initiated and grew in the region of
highest Gr (near the adhesive-strap interface). This indicates
that Gr has the greater influence on the debond location in adhe-
sive joint. This is consistent with the observations that adhesives
are inherently weaker under peel loading than under shear load-
ing. l ,2,15 Additionally, these results show that an accurate eval-
uation of GT can be achieved by a four layer model, while accurate
evaluation of GI and GIl require a more refined model. To
CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES 651

50
EC 3445

40 < ........} GT

30
---------------.. . . <} Gn
G.
r===' 1-
J/m2 THICK STRAP
20
~I
THIN STRAP
10

o 25 50 75 100 125 150

DEBONO LENGTH. mm

Figure 9. Variation of strain energy release rates with debond

length.

analyze the experimental debond growth rates, the debond location

for all subsequent calculations was selected by engineering judge-
ment to be at one-sixth of the adhesive thickness away from the
adhesive-strap interface. Also, the l2-layer model was used in
these calculations.

Figure 8 shows the typical variation of strain energy release

rates GI and GIl with the applied load on the cracked-lap-shear
specimen obtained from geometric nonlinear analyses. The GI and
GIl in nonlinear analysis were found to be functions of the square
of applied load within one percent.

All four sets of specimens were then analyzed to determine the

variation of ~, GI, and GIl with the debond length. Figure 9
shows the typical dependence of GT, GI, and GIl on the debond
length for both types of specimen with EC 3445 adhesive. For
specimen with thick strap, GT, Gr , and GIl were constant up to
140 rom of debond length. Similar behavior was found for co-cured
specimens with thick straps. For specimens with thin straps,
652 S. MALL ET AL.

Table III. Strain Energy Release Rate.

Strain Energy Release Ratea (J/m 2 )

Specimen Details
for the Applied Stress of 82.0 MPa

Strap
Adhesive GI GIl GIIG II
Type

EC 3445 Thick 9.75 31.08 0.31

(Secondary
bonding) Thin 8.23 33.20 0.25

FM-300 Thick 11.21 29.60 0.38

(Co-cure
bonding) Thin 10.66 32.40 0.33

aCalculated with debond location at one-sixth of adhesive

thickness from the strap-adhesive interface.

Gr, GI, and GIl were constant up to debond lengths of 115 rom and
65 rom in secondary and co-cure bonded systems, respectively. The
constant values of strain energy release rates for all four sets of
specimens are provided in Table III for a specified stress level.

RESULTS AND DISCUSSION

An attempt was made to determine if one of the components of

strain energy release rate (GT , GI' or GIl) dominates the cyclic
debonding. The measureddebond growth rates were, therefore, cor-
related with each of the calculated strain energy release rates GI,
GIl' and GT' These correlations are shown in Figures 10 and 11
for EC 3445 and FM-300, respectively. If one component of strain
energy release rate had a dominant influence, it would correlate
significantly better than the others when comparing the debond data
from specimens with different GI-to-GII ratios.

An equation of the form

da = c(G)n (1)
dN
was fitted to the data in Figures 10 and 11 by using a least-squares
regression analysis. The values of c and n, as well as the sum
of errors, ~r2, are shown in the figures. For each adhesive, the
values of ~r2 are about the same for the GI' GIl, and GT' How-
ever, the ~r2 term is lowest for GT, indicating that GT pro-
vided a somewhat better correlation than either GI or GIl' This
suggests that debond growth rate is a function of the combined
CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES 653

o THICK STRAP l!. THI STRAP

°C · 3.51E-ll C •4.69E-14
n · 4.06 n • 4.36
2
~} • 2.52
o
[r • 2.90

da
d ' 10- 4
mm!cycle

10- 5
C -1.81H4
n · 4.34
~} • 2.28

10 100 1000 10 1000 10

Gr
Figure 10. Relation between strain energy release rates and debond
growth rate for EC 3445 adhesive.

o THICK STRAP l!. THIN STRAP

C •6.32E-13 C • 7.14E-15
n • 4.54 n · 4.53
r/ . 1.14 [r 2 • 0.98
da 10- 4
dN'
mm!cycle
C •l.52E-l5
10- 5
n · 4.55
z:i ' 0.92

~____-L____~II~ ______~I____~I ~I ______~____~I

10 1000 10 100 1000 10 100 1000
Gil Gr , J/m 2

Figure 11. Relation between strain energy release rates and debond
growth rate for FM-300 adhesive.
654 S. MALL ET AL.

effects of GI and GIl' However, the GI-to-GII ratios for the

test specimens were all within the rather narrow range of 0.25 to
0.3S. As a result, it is not surprising that Gr, GIl, and GT
all were reasonably successful in correlating the data in Figures 10
and 11. Furthermore, each figure shows that data from both specimen
geometries are within an acceptable scatter band (similar to that
observed in fatigue crack propagation in metals).16 This indicates
that specimen geometry did not influence the relationship between
the debond growth rate and strain energy release rate. The rela-
tions between da/dN and Gr, for both adhesives, are compared in
Figure 12. This figure shows that the debond rate with FM-300 is
about 40 percent of that for EC 3445 for the same applied load and
debond length.

Because of the log-log scale in Figure 12, the curves relating

da/dN and G have slopes equal to the n term in Equation (1).
The values of n found in this investigation ranged from 4 to 4.5.
This is quite high compared to typical values of n for fatigue
crack growth in aluminum and steel alloys that range from 1.5
to 3. 16 These steep slopes mean that a small change in applied
load would cause a large change in debond growth rate. Thus, the
debond propagation in adhesive joints is more sensitive to errors
in design loads than are typical cracks in metallic structures.
Because of these steep slopes, it may be difficult to design bonded
joints for finite life. Minor design alterations or small analysis
errors could cause a much shorter life than the design value. A
viable alternative would involve an infinite-life approach. For
this purpose, the no-growth threshold, Gth' (based on GT for dis-
cussion purpose here) may be an important material property for
bonded systems. If a 10- 6 mm/cycle rate is arbitrarily assumed to
be the no-growth threshold, the curves in Figure 12 show that Gth
values for EC 3445 and FM-300 are 60.8 J/m 2 and 90.5 J/m 2 , respec-
tively. These Gth values are equivalent to applied stresses
(based on nominal area of strap) of 100 MPa and 121 MPa for EC 3445
and FM-300, respectively.

The threshold strain energy release rate, Gth, appears to

depend on the adhesive and not on the specimen geometry within the
range of this study. SAs a result, a Gth value and an analytical
method such as GAMNAS could be employed during design to determine
the maximum cyclic loads to obtain an infinite fatigue life for
bonded structural components. An initial debond must be assumed to
exist in order to calculate strain energy release rate. 'The size
of the initial debond can be estimated from the anticipated manu-
facturing defects on NDr limitations. The maximum design load can
then be determined to ensure that the applied GT in the bondline
is below Gth for the given adhesive. In the present study the
values of Gth for EC 3445 and FM-300 were obtained under labora-
tory ambient conditions. However, as with other adhesive proper-
ties, it is anticipated that Gth will be influenced by the
CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES 655

~
dN'
mml cycles -4
10

ASSUMED THRESHOLD
10 -6 DE~~N~ _G~9Y~!~ ~~T
,'
I" I
II
10- 7 '--_ _---'_ _ _..;..' "-I'_ __ -:""""'

1 10 100 1000
GT , J/m 2

Figure 12. Comparison of debond growth rate between two adhesives.

environmental conditions. Thus, experimental data as presented in

Figures 10 and 11 should be reproduced for the environment and
cyclic frequency of interest.

CONCLUSIONS

A combined experimental and analytical study of the cracked-

lap-shear specimen subjected to constant-amplitude cyclic loading
656 s. MALL ET AL.
was undertaken to analyze the fatigue behavior of simple composite-
to-composite bonded joints. Two bonded systems were studied--
graphite/epoxy adherends bonded with EC 3445 adhesive in a secon-
dary bonding procedure and with FM-300 adhesive in a co-cure bonding
procedure. With each bonded system, two specimen types were tested:
(1) strap adherend of 16 plies bonded to lap adherend of 8 plies,
and (2) strap adherend of 8 plies bonded to lap adherend of 16
plies. A finite-element analysis was conducted to calculate the
strain energy release rates: GI due to peel stress, GIl due to
shear stress, and GT due to both stresses. The present study led
to the following conclusions:

1. Cracked-lap-shear specimens were found to provide consistent

debond growth data. The cyclic debond growth rates were repro-
ducible from one specimen to another within a scatter band com-
parable to that for crack growth in metals.

2. Cracked-lap-shear specimens failed predominately by cyclic

debonding of adhesive, accompanied by some 00 fiber pull-off
from the strap adherend.

3. The debond always grew in the region of the adhesive that had
the highest mode I loading (peel stress or GI)' This indicates
that GI had a stronger influence than GIl on the debond
location.

4. Debond growth rates correlated very well with GI , GIl' and

GT' However, the specimen geometries tested in this study did
not produce a wide enough range of mixed mode loading condi-
tions to clearly establish the relative influences of strain
energy release rates GI, GIl, and GT on debond growth rate.

5. The slopes of the debond growth rate curves are considerably

higher than for crack growth rates in metals. This difference
makes adhesive debond growth life much more sensitive to
possible errors in design loads. Therefore, a safe-life (i.e.,
no debond initiation or growth) design criterion may be more
acceptable. This safe-life approach would emphasize the
threshold debond growth region.

ACKNOWLEDGMENTS

The first author is grateful for the support from the Army
Structures Laboratory (RTL-AVRADCOM) through the In-house Labora-
tory Independent Research (ILIR) Program during the course of this
study. -
CYCLIC DEBONDING OF ADHESIVELY BONDED COMPOSITES 657

REFERENCES

1. F. L. Matthews, P. F. Kilty, and E. W. Godwin, "A Review of

the Strength of Joints in Fibre-Reinforced Plastics, Part 2--
Adhesively Bonded JOints," Composites, 13 (I), 29 (1982).
2. L. J. Hart-Smith, "Analysis and Design of Advanced Composite
Bonded Joints," NASA CR-22l8, National Aeronautics and Space
Administration, Washington, DC (1974).
3. w. J. Renton and J. R. Vinson, "The Analysis and Design of
Composite Material Bonded Joints Under Static and Fatigue
Loadings," AFOSR-TR-72-l627, Air Force Office of Scientific
Research, Washington, DC (1973).
4. G. L. Roderick, R. A. Everett, Jr., and J. H. Crews, Jr., in
"Fatigue of Composite Materials," ASTM STP 569, pp. 295-306,
American Society for Testing and Materials, Philadelphia, PA,
1975.
5. T. R. Brussat, S. T. Chiu, and S. Mostovoy, "Fracture Mechan-
ics for Structural Adhesive Bonds," AFML-TR-163, Air Force
Materials Laboratory, Wright-Patterson AFB, OH (1977).
6. J. Romanko and W. G. Knauss, "Fatigue Behavior of Adhesively
Bonded Joints," Vol. I, AFWAL-TR-80-4037, Air Force Wright
Aeronautical Laboratories, Wright-Patterson AFB, OH (1980).
7. R. A. Everett, Jr., "The Role of Peel Stresses in Cyclic
Debonding," NASA TM-84504, National Aeronautics and Space
Administration, Washington, DC (1982).
8. B. Dattaguru, R. A. Everett, Jr., J. D. Whitcomb, and W. S.
Johnson, "Geometrically-Nonlinear Analysis of Adhesively
Bonded Joints," presented at the 23rd AIAA/ASME/ASCE/AHS
Structures, Structural Dynamics, and Materials Conference,
New Orleans, LA, May 1982. Submitted to ASME Journal of
Engineering Materials and Technology.
9. G. R. Alsmi1ler, Jr. and W. P. Anderson, "Advanced Compos-
ites Airframe Program - Preliminary Design," USAAVRADCOM-TR-
80-D-37A, U.S. Army Aviation Research and Development Command,
St. Louis, MO (1982).
10. E. J. Hughes and J. L. Rutherford, "Evaluation of Adhesives
for Fuselage Bonding," Report No. KD-75-74, The Singer Com-
pany, Little Falls, NJ (1975).
11. K. N. Shivakumar and J. H. Crews, Jr., "Bolt Clampup Relaxa-
tion in a Graphite/Epoxy Laminate," NASA TM-83268, National
Aeronautics and Space Administration, Washington, DC (1982).
12. L. J. Hart-Smith, "Effect of Adhesive Layer Edge Thickness on
Strength of Adhesive-Bonded Joints," MDC J4675, McDonnell-
Douglas Corporation, Long Beach, CA (1981).
13. 1. S. Raju and J. H. Crews, Jr., "Interlaminar Stress Singu-
larities at a Straight Edge in Composite Laminates," Com-
puters and Structures, 14 (1-2), 21 (1981).
14. E. F. Rybicki and M. F.Kanninen, "A Finite Element Calcula-
tion of Stress Intensity Factors by a Modified Crack Closure
Integral," Engineering Fracture Mechanics, 2. (4), 931 (1977).
658 S. MALL ET AL.

15. F. A. Keime1, in "Kirk-Othmer: Encyclopedia of Chemical Tech-

nology," Vol. 1, 3rd ed., pp. 488-510, John Wiley & Sons, New
York, NY, 1978.

16. "Damage Tolerant Design Handbook," Battelle Metals and Ceramics

Information Center, Columbus, OH, 1972.
EFFECTS OF LOW CYCLE LOADING ON SHEAR STRESSED ADHESIVE BONDLINES

W. Althof
Deutsche Forschungsanstalt fur Luft- und Raum-
fahrt (DFVLR), Institut fur Strukturmechanik
D-3300 Braunschweig-Flughafen, West-Germany

Adhesively bonded thick adherend model joints were cyc-

lically loaded on squarewave and sinusoidal modes. Dur-
ind the cycling loading,the real shear strains of the
bondlines and, for the first time, the failure strains
were quantitatively measured by use of a special shear
strain extensometer. The analysis of bondline strains
of a modern structural adhesive indicates dependence
on time, shear stress and cycle frequency, and suggests
a strengthening of the adhesive in the bondline under
cyclic loads. A linear viscoelasticity can be expected
only at very low bondline shear stresses. At higher
stresses, residual strains build up to a critical level
and exceeding this level causes failure of the bondline.
Adhesive material properties obtained in the present
study provide a guide to allowable shear stresses or
strains to design cyclically loaded adhesively bonded
structures.

659
660 W.ALTHOF

INTRODUCTION

Adhesives are organic polymers and at mechanical loading the

displacements of the polymers are time-dependent as a consequence of
their viscoelastic properties. At a constant load, the displacements
increase with time and the polymer creeps. On the other hand, the
load required for a constant displacement decreases with time and
the polymer relaxes. Corresponding time-dependent displacements are
expected at cyclic loading of polymers and adhesives.

Low cyclic loading means periods of loading and unloading in

a slow sequence. The loading and unloading can continually follow
at constant or different frequencies, or after the loading and un-
loading the loads are held constant for a fixed time. Especially
at the last cycle mode the viscoelasticity of the adhesives is of
a great importance.

In the last few years the viscoelastic properties of adhesives

have been intensively investigated. 1,2 Often viscoelastic proper-
ties were determined by tests on cured neat adhesive coupons. At
DFVLR, however, adhesive shear properties are determined by meas-
uring the shear strains of short bondlines between adherends of
high stiffness. 3 It was proved that the stress-strain relationships
of such short bondlines are representative of the adhesive sections
in long bondlines. 4 This method was employed on static short time
tests and also for determining viscoelastic properties for creep
and relaxation. S

All the investigations cited established that adhesives deform

linear-viscoelastically only at low stresses, and only under this
condition the time-dependent stress-strain relationships can be de-
scribed by simple mathematical formulae. It should be noted that
the adhesive bond lines failed on exceeding a limit of shear strain. S

Adhesively bonded joints under sinusoidal cyclic loads were

already investigated in 1963 and an accumulating of residual shear
displacements was found. 6 Quantitative stress-strain-time relation-
ships were not ascertained because methods to measure the shear
displacements in bondlines of the used thin adherend joints were
imperfect at that time. Investigations on cycled thick adherend lap
joints were recently reported but again the bond line displacements
were not measured quantitatively.7 At DFVLR, however, this measure-
ment is possible and new promising investigations on cyclically
loaded joints were carried out dealing with the following:

Measurement of the time-dependent bondline shear strains

during cycling at different times.

Analysis of the time-dependent shear strains with respect to

a mathematical description.
EFFECTS OF LOW CYCLE LOADING 661

Estimation of bondline shear stresses at which strains are

negligible with respect to a long life of bondlines.
Comparison of the effects of squarewave and sinusoidal cycled
bond lines •
Estimation of failure criteria of bondlines under cyclic
loading.

!-- T MAX
V>
V>
L1.J
=
.....
V>
LMIN

ltL1Cycle tu TIME t

Figure I. Squarewave stress/strain-time functions.

EXPERIMENTS
Cycle Modes, Specimens, Equipments

The tests were carried out mostly on a squarewave load-time

function with parameters imitating simplified flight missions of an
aircraft. The specimens were loaded and unloaded during ca. 10 sec-
onds, the times under maximal loads (load periods) and under mini-
mum loads (unload periods) lasted between 0.5 and 60 minutes.
Figure 1 represents the stress-time and the strain-time functions.
Table I shows the different times of load and unload periods, tL
and tu, during squarewave cycling tests.
662 W. ALTHOF

Table I. Times of Load and Unload Periods During Squarewave Cycling.

Load Period Unload Pet;iod Abbreviation
(minutes) (minutes)

60 15 60'/15' "Boeing mission"

15 5 15'/5' "Douglas mission"
4 2 4' /2'
1 1 1' / 1'
0.5 0.5 0.5'/0.5'

Thick adherend specimens bonded by a modern structural epoxy

adhesive (Cyanamid "FM 73") were tested. The shear strain was meas-
ured by using a zero-gage extensometer developed at DFVLR. Figure 2
shows schematically the arrangement of the thick adherend specimen
(model joint) with the clipped extensometer and the bondline shear
displacement.
The selected loads caused the maximum bondline shear stresses
't max 20, 25, 30, 35, 37.5 and 40 Mpa, and minimum shear stresses
'tmin = O.OI't max ' respectively.

0=6 F
~

Figure 2. Thick adherend specimen (model joint) with clipped shear

strain extensometer, schematically.
EFFECTS OF LOW CYCLE LOADING 663

60
~
~

l-
V)
V) 40
UJ
0::
tn
0::
<{ FM73
UJ
:z:
V)

o 0.1 D.2 0.3 0.5 1.0 1,5

SHEAR STRAIN 'I

Figure 3. Basic shear stress-shear strain diagram of the tested

structural adhesive "Cyanamid FM 73".

On sinusoidal cycling the frequencies f = I Hz and 10 Hz were

used. The load amplitudes caused bondline shear stresses t max = 10,
15, 20, 25, 30, 32.5, 35 and 40 MPa, and tmin = 0.1 t max ,
respectively.
For the adherend thickness, a = 6 mm (aluminum alloy bare
2024-T351), and the lap length, L = 5 mm, the resulting shear stress
distribution in the loaded bondline is predominantly uniform.
Therefore the shear stress can be calculated from

1: = F/b.L

and the shear strain is obtained as

tany = n/t

where F = load, b = width, t = bondline thickness, n = shear dis-

placement of the bondline which is the measured relative movement
of the adherends. Figure 3 represents the basic shear stress-shear
strain diagram of the adhesive tested as recorded on a static short
time test.

At the cycling loading, the bondline shear strain was measured

and recorded continually. On sinusoidal cycles, the failure strains
664 W.ALTHOF

were obtained by use of a transient recorder. The failure strain

was defined by the maximum strain recorded at the last complete
sinusoidal strain amplitude before the bondline fails.

The specimens were loaded on square wave cycles by a mandrel-

driven tension tester. On sinusoidal cycles, a servo-hydraulic test
equipment controlled by a microprocessor developed at DFVLR was
used which also stored continually the measured strains.

RESULTS

Model Joint Squarewave Cycling

During the cycling of model joints the bondline shear strains

were recorded from the start of cycling up to the failure of the
bondline. Figure 4 shows a typical stress-strain relation during
cycling.

The loops in Figure 5 recorded at the shear stresses L = 35

and 25 Mpa during 60'/15'-, 15'/5'-, and 4'/2'- cycles characterize
the stress- and time-dependent shear strain behavior of bondlines
under low cyclic loading. The broken lines in the diagrams repres-
ent the static stress-strain curve of the adhesive tested. With the
aid of these lines it is possible to discern whether the initial
strain at the first loading is related to the elastic, the elastic-
plastic or the plastic section of the total bondline shear strain.

I-' n"t
V')
V')
L.U ,/ ;i
# .
, I
0:::
>-
V')
L./
ST RAIN tan j'

Figure 4. Shear stress-shear strain loops recorded on squarewave

cycling, schematically.
 m
"T1
"T1
m
(")
-i
C/l
o
"T1
r
o
:E
CYCLE 60'/15' CYCLE 15'/5' CYCLE 4'/2' (")
60
-<
(")
r
40 m
r
~ o
~ 20
»
o
~ Z
G)
~ OL,----~-----
w
1r60
ti
a::
(;5 40 51 A1\~--------., _--------,;:;;;;---4 -----------,
:I: - ---- ( o.J\AJ,
----215.
III 11 4. 26. 118. CYCLE fl. , 17.66. I 975. CYC LE
9, 22.
i' I 1627. C,
~l II NO FAIL. 1f III NO FAILURE
MJ If F~Pl. I ,
0,5 1.510 0:5 ; 1.5 10 0,5 1.5
SHEAR STRAIN r
Figure S. Shear stress-shear strain loops recorded at different shear stresses and different
squarewave cycle periods,

m
m
U'I
666 W. ALTHOF

The following characteristics can be deduced from Figure 5:

If in the first cycle the load already causes a plastic strain,
the bondline survives only relatively few cycles.
Shear stresses below the quasi-elastic stress limit, 1: E' cause
low strains; however, they increase continually and slowly.
The bondline creeps during the first cycles more than in the
later ones.
The recovery during the unload periods is mostly lower than
the creep during the load periods hence residual strains remain.
The residual strains build up to a critical strain at which
the bondline fails.
The shorter the periods of loading and unloading, the more are
the numbers of cycles up to failure.

Table II shows, for all the tests, the failure strains and the
numbers of cycles at bondline failure. A comparison indicates:
The failure strain corresponds to the failure strain in static
tests.
The number of cycles at failure increases with shorter cycle
periods and lower shear stresses.
At shear stresses below 1: = 20 MPa, the residual strains are
low and failures are expected at very high numbers of cycles.

Table II. Test Results of FM 73-Adhesive Bondlines, Squarewave

Cycled on Different Times of Cycle Periods and Different Shear
Stresses.
Shear Time of Load Period/Unload Period (Minutes)
Stress 60/15 15/5 4/2 1/1 .5/.5
max Nf Yf Nf Yf Nf Yf Yf
40 MPa 7 I .48 6 I .48 5 1.46 29 1.48 30 1.50
5 1.45 25 1.60
37.5 8 I .40 II I .45 62 I. 41 27 1.43 28 1.38
5 1.43 13 1.52 26 1.50
35 15 I .50 35 I .48 47 1. 36 164 1.52 72 I .42
35 1.53 56 1.40
171 1.45
30 33 1.50 61 1 .60 49 1.48 1765 n.f.= no failure
52 I. 38
25 120 n.£. 1626 n.£. 975 n.£. 1640 n.£.
20 1270 n.£.

Nf numbers of cycle at failure, Yf = shear strain at failure,

On static loading the bondlines fail at Yf = 1.50.
EFFECTS OF LOW CYCLE LOADING 667

Analysis of Bondline Strains

The stress-strain loops shown in Figure 5 assume that strains

built up at loading are reversible, and that the time-dependent
creep and recovery during the load and unload periods are of import-
ance with respect to the life time of the bonded joints. For analys-
ing the strain components additional specimens were tested at diff-
erent shear stresses using cycle periods of 30 minutes. During cycl-
ing strains in the succesive periods were recorded •

... .. ..
" "
LOADED UP TO , : 25 MPa
<t
ct:: O~----~------L-----~------~--~
l-
V)

ct::
.... o ..
«
w
o ... 0 •

I
" S
V)~
20 UNLOADED DOWN TO ,:0.25 MPa

o 10 20 30 40 75 100
NUMBER OF CYCLES N

Figure 6. Shear strain after loading up to shear stress 1: = 25 MPa

and after unloading down to 1: = 0.25 MPa vs. number of cycles.

In Figure 6 the measured strains at loading'YL' and unloading,

Yu, are compared for a stress level of 1 = 25 MPa. This diagram as
well as other . test results at lower stresses indicate that these
strain components are approximately equivalent pointing to a quasi
elastic strain behavior at loading up to the maximum stress and un-
loading down to the minimum stress.
668 W.ALTHOF

10 20 min 30

~
ii::'
~ 2
trl
cr 3 L----...,.-I:--------='::----~.
o 10 20 min 30
TIME tu

Figure 7. Creep and recovery vs. time at low shear stresses during
the load and unload periods on 30'/30'- squarewave cycles.

For different stress levels the creep strain, YC' and the re-
covery strain, Y R, at successive cycles are presented in the next
figures. Figure 7 shows these strain components for the stresses
~max = 10 and 15 MPa. The creep and the recovery are approximately
equivalent indicating a linear-viscoelasticity at these low stress-
es. The time-dependence of strains at low stresses is often better
described by a power function than by the use of a spring-dashpot-
model: 5

Yc ,R = a· t n

For L = 10 and 15 MPa stresses, the formula is applicable with

a = 0.00055, n = 0.19, and a = 0.0011, n = 0.23, resp.

The creep and the recovery at L = 20 and 25 MPa are shown in

Figure 8. It can be seen that already at 20 MPa the creep is higher
than the recovery. The difference is more remarkable at L = 25 MPa.
Moreover, during the initial load periods the creep increases, later
it decreases with increasing number of cycles. The recovery behaves
similar to the creep. However, after about 100 cycles creep and re-
covery are approximately equivalent.
EFFECTS OF LOW CYCLE LOADING 669

T=2SMPo

10 20 30
TIME tl

1-
o .::::..."""":.--------------1----- }r-20
~ ~=~~~~~====L~~~~~=
il: 10 ~ 1 BO KlO
~ 2 50 T=25
6 20
u
lJ.J
a:: 3 0 10 20 min 30
TIME tu

Figure 8. Creep and recovery vs. time at medium shear stresses

during the load and unload periods on 30'/30' squarewave cycles.

The strain behavior described is typical of polymers which

strengthen during longtime loading and then deform quasielastically.
Such a strengthening can be explained by a stretchability of the
initial amorphous arrangement of the molecules in the polymer
caused by the longtime loading.

It is evident that for a cycle-dependent strengthening of the

adhesive the time-dependence of the strains cannot be described by
simple mathematical formulae. It was often recommended to ignore
the cycles up to the finished strengthening and thereafter describe
the viscoelasticity mathematically. However, this method is doubt-
ful as the strains during the strengthening may be high in relat-
ion to the failure strain.

Model Joint Sinusoidal Cycling

As previously mentioned, for the first time, the bondline

failure strain was measured on low and fast sinusoidal cycles. As
an example, Figure 9 shows some strain cycles recorded just before
670 W. ALTHOF

165 FAILURE
mnil Tmox" 35MPo
1.15
160 'Vl 1 Hz
<'"l
~
d
1.10 0 "
~ II')
Z ~.
0
i=
u "
..;
"- w
...J Z
~ 1.05 lL
W ~
0 0:
0:
f-
(/') 0:
IJi
0:
0:
« w w
0:
L1J row
:x: ::>
~ 1.0 11.0
(/') :1:0:
::>::>
...J
:x:
(/')
Z...J ~
~~
~<
u
0.95

130 2370 2375 2380

2280 2
NUMBER

Figure 9. Shear strain vs. numbers of cycles during 1 Hz- sinus-

oidal cycling to determine the failure strain.

the bondline failed. On the left the first 10 cycles of the record-
ed last 100 cycles are to be seen, on the right the last 20 cycles
are shown with the last complete strain amplitude from which the
bondline failure strain was determined.

Figure 10 shows the envelopes of 1 Hz - shear strain ampli-

tudes recorded at different shear stresses. Although the semilog-
arithmic scale distorts the diagram, the following can be inferred
from the test results:

At low shear stresses (1)25 MPa) the bondline strains are

approximately constant up to a high number of cycles. The bondline
deforms quasi-elastically. At higher shear stresses the strains in-
crease and the characteristic of the strain behavior is similar to
EFFECTS OF LOW CYCLE LOADING 671

the static creep.5 The creep is caused by the effective mean shear
stress. During the initial period of cycling the creep rate is high,
later it is low and a short time before the bondline fails the rate
increases again.

The amplitudes of the strain are constant for a long time and
increase just before the bondline fails. With increasing stresses,
the number of cycles at failure decreases. Furthermore, the failure
strain decreases with decreasing shear stresses and increasing num-
ber of cycles. This phenomenon was observed for the first time.

During some sinusoidal tests the stress-strain loops were also

recorded, an example is presented in Figure 11. The loops recorded
at different shear stresses indicate following more important facts.

At high stresses with an initial elastic-plastic bondline

displacement, the strain does not reach its maximum value at the
maximum stress. Moreover, the loop is not closed after the unloading.

1.2
f = 1H z FA ILURE

1,0 I
I
I
I
I
0,8 /
I

z
~ 0,6
tn
0::
~ 0,4
I
Vl I
I
I
I
0,2 /
I

25 20
------------------------~-
o
10 2 10 3 10'
NUMBER OF CYCLES N

Figure 10. Envelopes of shear strain amplitudes measured on 1 Hz-

sinusoidal cycling at different shear stresses.
672 W. ALTHOF

This means the strain is not recovered. Such a time-delayed nonre-

versible strain behavior with a time lag in phase of stress and
strain is typical of nonlinear-visoelastic materials such as poly-
mers. With an increase in cycles the lag in phase disappears and
the bondline deforms quasi-elastically up to a short time before
the failure. In this short time nonreversible strains appear again.

The change of the strain behavior from nonlinear-viscoelastic

to elastic is probably due to the effect of the strengthening in
the adhesive similar to that at squarewave cyling. The longer the
bondline is cyclically loaded the more the adhesive strengthens.
After the strengthening the adhesive becomes brittle. This effect
probably causes the observed decrease of the failure strain during
sinusoidal loading at medium shear stresses and high number of cyc-
les, see Figure 10. The results of tests at the frequency of 10 Hz
seem to confirm this hypothesis, see Table III which includes all
the measured failure strains and numbers of cycles up to failure at
different shear stresses and cycle frequencies.

40 l.CYCLE 2. 3.
30

_ 20
~
.! 10

Figure 11. Shear stress-shear strain loops on sinusoidal cycling

(1 Hz) at different shear stresses.
EFFECTS OF LOW CYCLE LOADING 673

Table III. Test Results of FM-73-Adhesive Bondlines under Sinusoidal

Cycling at Different Frequencies and Different Shear Stresses.

Shear f = I Hz f=IOHz
Stress
1: max Nf Yf Yf
40 MPa 3.3 x 10 1 1.24

35 2.4 x 10 3 1. 13 2.5 x 10 4 0.86

7.2 x 10 3 1.13 2.9 x 10 4 0.89
7.3 x 10 3 1.11 2.1 x 10 4 0.87

32.5 1.2 x 10 4 1.12 1.0 x 105 0.58

1.3 x 10 4 1.06 1.5 x 105 0.89
4.0 x 10 4 1.03

30 5.0 x 10 4 0.93 1.9 x 105 0.80

2.1 x 105 0.70
3.2 x 105 0.68

25 1.0 x 105 no failure 1.6 x 10 6 0.73

20 2.8 x 105 no failure
Nf numbers of cycle at failure, Yf shear strain at failure

Post-Test Examination of Bondlines

The failure surfaces of the bondlines failed on squarewave

and sinusoidal cycling did not differ significantly from the faces
of bondlines failed after static loading or creep. The bondlines
failed in the middle of the adhesive layer. However, a "finger nail
crack front" on the end of the lap was not identified. It has been
reported that such crack front extend in the bondline more and more
on succesive cycles.

Actual Lap Joint Squarewave Cycling

The test results described were obtained on specimens with

short bondlines. It is assumed that they are representative of ad-
hesive sections in longer bondlines. To prove results concerning
shear stresses causing negligible residual strains, standardized
thin adherend lap joints were cyclically loaded on squarewave mode
with load and unload periods of I hour~ During cycling the bondline
strain was measured in the middle of the lap length in the previ-
ous manner by the use of a shear strain extensometer with meas-
uring points placed at shorter distances than in Figure 2.
674 W. ALTHOF

1.5
T,no.
tm

~
1.0 12 L
z
4:
0::
tIi t =30MPa (T"".= 34 ; t o =26)
0:: 0.5 m~
«
w
I
V)

o L..----''----J , It , I L---....J
!1O . CYCLE ) 120. CYCLE) 122. C.l

rL r
11. CYCLE)
0.06

~ 25132,17)
« Q04

0 "),11) ,---' (----

0::
l-
V)

~
':l:! 0.02 J~~O_l~ I : :
V)

o
~, __ __ ~ J
I'

2 18 19
.
! ~~_l~8 ; 8 )
I
(:- __J
20 38
.
~ ..____ ~

..
I
j

39
.t,--_..J
l
~._

L.O 42 43
TIME I lh)

Figure 12. Shear strain vs. time for a handline section in the middle
of an actual lap joint on squarewave cycling at different mean shear
stresses.
EFFECTS OF LOW CYCLE LOADING 675

The joints were loaded at the mean shear stresses '( m = 15, 20,
25, and 30 MPa. The respective maximum shear stresses at the end of
the overlap were calculated to betmax = 28,30,32, and 34 MPa; the
m~n~mum stresses in the middle of the lap to be ~O ~ 8, II, 17, and
26 MPa, respectively.4

The results are presented in Figure 12. At low mean shear

stresses the bondline deforms quasi-elastically, at medium mean
stresses the strains are negligible because the shear stresses in
the middle of the lap length are lower than the estimated limit of
linear-viscoelasticity of the adhesive tested (1= 15 MPa).

At high mean stresses and stresses in the middle, which are

higher than the mentioned limit, the bondline deforms nonreversibly
and residual strains remain.

At very high mean stresses and stresses in the middle, which

exceed considerably the critical limit, the number of cycles at
failure is low.

The example confirms the requirement that in long-time loaded

bondlines, at the bondline section with the lowest stress, stresses
or strains of the adhesive should not exceed the limit of the lin-
ear viscoelasticity. Otherwise, high nonreversible bondline strains
must, be expected.

CONCLUSIONS

Adhesively bonded thick adherend metal joints (model joints)

were low cyclically loaded on squarewave and sinusoidal modes with
different period times. During cycling the bondline shear strain as
well as the failure strain were quantitatively measured. The tests
confirm the qualitative wellknown time-dependent strain behavior of
adhesives. Also the following new information was obtained with re-
gard to engineering design of structural adhesive bonded joints.

Only at low stresses the adhesives deform linear~viscoelastic­

ally and the time-dependence of the strains can be described mathe-
matically. For the adhesive tested the stress level is about 15 MPa.
At medium stresses (1 = 15 - 25 MPa) the strains contain plastic
portions. For cycles of short periods, e.g. I Hz-sinusoidal cycles,
the time is not adequate to cause disadvantageous residual strains.
However, on squarewave cycles with long periods of constant loads
residual strains build up in the first few cycles; later the bond-
line deforms quasi-elastically. This change of the strain behavior
may be a result of the strengthening of the adhesive. At high,
stresses (t > 25 MPa), more plastic strains appear which are not
reversible on squarewave cycles. The residual strains build up to
676 W.ALTHOF

a critical strain at which the bondlines fail. The failure strain

corresponds to the failure strain at static short-time loading or
at creep. Therefore the exceeding of the strain capacity may be a
failure criterion for bondlines under squarewave cycling.

The exceeding of the strain capacity may also be a failure

criterion for bondlines under sinusoidal cycling on low and medium
frequencies. This hypothesis is a consequence of the observed shear
strain behavior at medium and high shear stresses. The constant load
amplitudes cause constant strain amplitudes, but the effective mean
load causes an increase of the total strain which is analogous to
the strain behavior at static creep tests under sustained loads. In
sight of these facts, the failure criterion cannot be a fatigue
failure caused by a progressive crack front in the bondline as ob-
served in other investigations. 7 A progressive crack front, of
course, must continually reduce the bond area. At constant load
amplitudes the stress will also continually increase and therefore
the strain amplitudes too.

The observed strain behavior is strictly valid only for short

bondlines with a homogenous shear stress distribution in the bond-
line. In bonded. structures, however, the stresses in the bondlines
are mostly shear stresses combined with normal stresses. Therefore
the adhesive strain behavior is very complex and also affected by
its dependence on time and strengthening. However, at the recom-
mended low stresses the computation methods of the structural mech-
anics are applicable, providing a linear-elasticity or a linear-
viscoelasticity of stresses and strains. Furthermore, at low stress-
es, the strain capacity of the adhesive is not fully exploited, so
that there are strain reserves in the case of local bond line stress
concentrations.
The tests described and the measuring method allow determina-
tion of adhesive material properties for shear loading required in
engineering design of bonded joints. Most probably the observed
effects of the shear strain behavior are transferable to tension
loaded joints.
Finally, the allowable shear stress requirements determined
were successfully proved to be applicable to actual bondlines by
tests of cyclically loaded thin adherend lap joints.

ACKNOWLEDGEMENT

Grateful acknowledgement is made to Messrs. G. Klinger and

G. Neumann who carefully carried out the experiments.
EFFECTS OF LOW CYCLE LOADING 677

REFERENCES

1• J. Romanko, in "AGARD-Lecture, Ser. No. 102," p. 4-1,

March 1979.
2. V. H. Kenner, W. G. Knauss, and H. Chai, Experimental Mechanics,
22, 75 (Feb. 1982).
3. if."" Althof and W. Brockmann, in "Bicentennial of Material Pro-
gress, SAMPE Symp. Proc. Vol. 2 I," p. 581, .April 1976.
4. W. A1thof, in "New Horizons - Material and Processes for the
Eighties, SAMPE NCTS 1 I," p. 309, Nov. 1979.
5. W. A1thof, J. Reinforced Plastics and Composites, I, 29 (Jan.
1982). -
6. A. Matting and G. Ulmer, Kautschuk und Guromi - Kunststoffe,
16, 334 (1963).
7. :r:- Romanko and W. G. Knauss, in "Developments in Adhesives - 2,"
A. J. Kinloch, Editor, p. 173, Applied Science Publishers,
London, 1981.
EFFECT OF SCRIM CLOTH ON ADHESIVELY BONDED JOINTS

E.C. Francis and D. Gutierrez-Lemini

United Technologies
Chemical Systems Division
Sunnyvale, California 94086

This report discusses the effect of the scrim

cloth on the fracture behavior of adhesively bonded
joints. The thick-adherend model joint specimen was
analyzed with the TEXGAP finite element computer
code. Three situations were considered: (1) uniform
adhesive without inclusion (scrim), (2) adhesive with
bonded inclusions, and (3) adhesive with debonded
inclusions. The material properties and specimen
geometry were those of a model joint previously
analyzed without inclusions by General Dynamics/
Fort Worth Division (GD/FWD).

Significant differences in the KI and KII values

were noted between the joints with bonded and unbonded
inclusions. The results gave insight into why the
scrim weakens the bond and provides a crack-growth
path. Scanning electron microscope (SEM) photographs
were obtained which show that the scrim cloth is
either poorly bonded or not chemically bonded to the
FM-73 epoxy.

679
680 E.C. FRANCIS AND D. GUTIERREZ-LEMINI

INTRODUCTION

The use of adhesives for bonding aircraft components has

increased greatly in the last two decades. In fact, the high
strength-to-weight ratios of adhesively bonded joints has led
recently to their application in primary fuselage structures l
and thus to the development of suitable criteria for evaluating
the integrity and structural durability of adhesively bonded
joints 2 .

To control the adhesive thickness in a bonded joint, a scrim

cloth is generally utilized in the adhesive. Published analyses
have neglected the presence of the scrim or mat carrier (refer-
ences 2, 3 and 4), which, as described in the following section,
may weaken a bonded joint. To investigate the structural influ-
ence of the scrim cloth on the fracture behavior of adhesively
bonded joints, we idealized the fibers of the scrim as cylindrical
Dacron inclusions located near the tip of a crack at an adhesive-
adherend interface. The results obtained are supported by experi-
ment, as described below.

Experimentally Observed Effect of the Cloth Carrier

There is evidence that the scrim or mat cloth carrier

influences the response and fracture behavior of structural adhe-
sives. The cloth in FM-73 is"most likely not bonded to the epoxy
resin, or if it is bonded, the bond is very weak, as inferred from
Figure 1. This figure shows SEM photographs of fracture surfaces
of FM-73 adhesive that were taken from slow static tests of thick-
ad her end model joints. The photographs show a smooth surface
between the cloth and the epoxy resin. One would also suspect
that the scrim or mat provides a pathway for preferential moisture
ingression. Consequently, the mat may represent a built-in defect.
The neat adhesive is, in fact, similar to a foam, even though
reported values of Poisson's ratio ranging from 0.32 to 0.40
(Reference 5) are much higher than those typical of foams (0.10 to
0.16).

Most response and fracture tests on the neat adhesive are

performed so that the scrim either reinforces or weakens the epoxy
and the fracture plane extends through the Dacron mat. In bonded-
joint fracture tests, on the other hand, the fracture plane is
parallel to the mat. It is not clear how one can meaningfully
relate these two fracture processes, or even the fracture data.
Fracture data from the neat-neat adhesive (FM-73U) may perhaps
be more relevant to fracture in a bonded joint.
EFFECT OF SCRIM CLOTH 681

M = 33 M = 300

M = 100 M = 1000

Figure 1. SEM photographs of fracture surfaces of FM-73 adhesive.

Recently, GD/FWD 2 found that fatigue life increased by a

factor of four (25,000 cycles to 100,000 cycles) and that the
aluminum adherend failed, rather than the adhesive, in tests of
thick-adherend model joint specimens bonded with FM-73U. These
results suggest again that the scrim may indeed be detrimental
to the structural durability of bonded joints.
 682 E.C. FRANCIS AND D. GUTIERREZ-LEMINI

Finite Element Analysis of the Model Joint

The thick-adherend model joint shown in Figure 2 was analyzed

by us at Chemical Systems Division. The corresponding finite
element idealization is depicted in Figures 3A and 3B. The geometry,
loading, and material properties of the structure correspond to
those used previously by the GD/FWD (see, e.g., QPR 5 of Reference
2), except for the fact that we included only one crack instead of
the two antisymmetrica11y placed cracks present in the GD/FWD model.

~
+----
0.125in.
2.225 in. - - - - + 1 -....

~r_-----------------------~
0.125 in.
1 -0.500 in.-I--+----2.225 in. ----I-
_0.
_0
• "in

-0
.0
~
0.

-
~

~.~------~~--------~/ ~------~----------------------~~
-~II
b

Crack

Adherend Adhesive Layer

Material Properties

Value Adherend Adhesive Inclusion

E. psi 10.3 x 106 131.290 450.000

u 0.33 0.32 0.32

Figure 2. Geometry, loading, and material properties used in the

analyses of the model joint.

The behavior of the model j oint with a single crack near the
adhesive-adherend interface was examined under the following
cond it ions:

1. By considering a homogeneous adhesive without inclusions

2. By placing a couple of inclusions of modulus 450,000 psi and

Poisson's ratio 0.32, bonded to the homogeneous adhesive

3. By assuming these inclusions to have a very low modulus (i.e.,

10 psi) •
EFFECT OF SCRIM CLOTH 683

-
-- - - \'
-
-
-
-
~
~ -- ~
~

- A. Finite Element Mesh ot the Adhesive llyer and

lett Adherend

Adherend

Crack

_ _---L_ _.....L_ _-'-_----'It'-_----'_ _---L_ _--'-_ _ _ Adhereno

B. Det.~ at the Finite Element Mesh Around the

Crack and lnciusions

Figure 3. Finite element model of the bonded jOint.

The region around the crack tip was modeled with the crack
element of TEXGAP-2.5, which is based on a displacement hybrid
formulation and yields the stress intensity factors directly.
General Dynamics, on the otherhand, used vectorial crack opening-
displacement data to obtain the stress intensities. The results
of these analyses are summarized in Table r.

While the model employed is a crude approximation to the

scrim cloth, significant differences were noted in the magnitude
of Kr if the scrim is unbonded or perhaps only weakly bonded.
This observation supports our contention of the importance of the
scrim and gives some insight into why the scrim may weaken the
bond and provide a crack-growth path. The discrepancy with GD/FWD
results for Krr is not known, but it may be related to the fact
that we used a singular element, whereas GD/FWD did not.
684 E.C. FRANCIS AND D. GUTIERREZ-LEMINI

Table I. Stress Intensity Factors for the Model Joint with and
Without Inclusions, and with a IO-Mil Crack

Without
Inclusion
With Unbonded
Value eso GO/FW Inclusion (eSO) Inclusion (eSO)
KI, psi 'VIii: 9.2 8.0 8.9 18.5

Kulsi vrn: 45.2 66.5 45.4 39.9

DISCUSSION

The analyses reported herein, as well as those carried out

by GD/FWD, bring out several important aspects concerning the
behavior of adhesively bonded joints:

1. The thickness of the adhesive layer is so small (only a few

mils) that the size of the yield zone may be of critical
importance. If this is the case, the stress intensity
factors may not adequately characterize the response fields
at the crack tip. Tests at GD/FWD on adhesive layers contain-
ing a scrim show whitening around crack tips on a scale that
is comparable to the thickness. thus suggesting that a stress
field exists associated with the stress intensity factor
concept used in the analyses.

2. The adhesive itself is probably not an isotropic material, as

was assumed in the analyses, since the mat carrier introduces
an in-plane reinforcement that is quite different from the
through-the-thickness reinforcement. Even without the scrim,
the material is probably orthotropic because of the thin
layer; the boundaries introduce orientation effects in the
molecules and give a directional effect to the curing
process.

3. Structural adhesives may respond in a complicated, nonlinear

viscoelastic or viscoplastic manner which depends on the
history of loading, temperature, moisture, and state of
"damage" induced by these environments. Thus, their idealiza-
tion as linear elastic materials may very well be an over-
simplified assumption.

4. The mat carrier is clearly not a set of isolated inclusions in

a matrix of neat-neat adhesive, but rather a mesh of fibers
interacting in a complicated manner.
EFFECT OF SCRIM CLOTH 685

In view of these peculiar characteristics of structural

adhesives, the results obtained so far for adhesively bonded
joints must be regarded carefully in practical design.

CONCLUSIONS

The analyses reported herein show, qualitatively, the detri-

mental effect of the scrim cloth when it is not bonded to the
surrounding adhesive. The results also show that when the scrim
cloth is strongly bonded to the adhesive, the system behaves
essentially as if the carrier were not present. Experimental
evidence indicates the existence of weak bonds between the mat
carrier and the adhesive, thus increasing the possibility of
crack propagation in the system.

REFERENCES

1. E. W. Thrall, Jr., "PABST Program Test Results", Adhesives Age,

pp. 22-23, Oct. 1979.
2. J. Romanko, K.M. Liechti and W.G. Knauss, "Integrated Metho-
dology for Adhesive Bonded Joint Life Predictions", AFWAL-TR-
82-4139, Nov. 1982*.
3. K. J. Sen and R. M. Jones, "Stresses in Double-Lap Joints
Bonded with a Viscoelastic Adhesive: Part I", AIM J.,~ ,
No. 10, 1237 (1980).
4. T. R. Brussat, S. T. Chiu and S. Mostovoy, "Fracture Mechanics
for Structural Adhesive Bonds - Final Report", Technical
Report AFML-TR-77-163, Lockheed - California Company, October
1977*.
5. J. Romanko and W. G. Knauss, "Fatigue Behavior of Adhesively
Bonded Joints", Final Technical Report, AFWAL-TR-80-4037,
April 1980*.

*Reports available from Air Force Wright Aeronautical Labora-

tory - Material Laboratory - Air Force Command - Wright
Patterson AFB, Ohio 45433.
 Part VI
Fracture Aspects
MECHANICAL MEASUREMENT OF INTERATOMIC BONDING ENERGIES AT

INTERFACES

E.H. Andrews

Materials Department
Queen Mary College
Mile End Road, London, El 4NS, England

Until recently it has not been possible to relate,

quantitatively, the mechanical strength of adhesive
joints to the surface chemistry of the system. This
situation has been changed dramatically by use of the
Generalized Theory of Fracture Mechanics (GFM) which
states that the measured work of debonding, e, per unit
area is given by

e = eo ~
.
(c, T, s)

where e is the energy to break interatomic bonds

across ~he fracture plane and ~ is a loss function
dependent on crack velocity, temperature· and strain.
Using this approach it is possible to extract the value
of 8 from direct mechanical measurements of 8. Results
are 0 presented for rubberlike and structural adhesives
and for the effects of water environments on e .
o

689
690 E.H. ANDREWS

INTRODUCTION

While it is obvious that the strength of interatomic bonds

across the interface must affect the mechanical strength of an
adhesive joint, a quantitative correlation betweenlthese parameters
has proved difficult to establish. Thus Dahlquist found a
correlation between thermodynamic work of adhesion (wA) and joint
strength in peel tests but Kaelble 2 fai!ei to do so in his studies
of an alkyl acrylate copolymer. Mittal' has reviewed the
relationship of wA to joint strength in various systems. Even when
correlations are obtained, of course, the absolute magnitude of
joint strength far exceeds wA. A second problem is that the
atomic interactions across an interface are difficult to determine
quantitatively, even measurements of wA depending heavily upon
theoretical assumptions. Thirdly, most mechanical tests of
adhesive strength determine the strength of the joint rather than
of the interface, and thus depend upon joint configuration, method
o£ loading and so on.

These problems have been largely overcome in recent studies

by the author and his co-workers at Queen Mary College and the
purpose of this paper is to review the results of these studies
and assess the progress made to date.

THEORETICAL BASIS OF THE STUDIES

The key to all the work summarized below lies in a new

formulation of fsacture mechanics called 'Generalized fracture
mechanics' (GFM) . The use of fracture mechanics in adhesive
testing is well established and has the recognized capabil ity of
eliminating the effects of joint geometry or configuration.
Essentially, fracture mechanics measures the resistance of a glue-
line or adhesive interface to the propagation of failure or
de-bonding. GFM is essentially the same in this respect, i.e. it
measures the total energy required to produce unit area of fracture
or debonding. This "failure energy" is denoted by 2 ':1 (two
identical surfaces) for cohesive fracture or e (one fracture plane)
for adhesive or interfacial failure. Where GFM goes beyond normal
fracture mechanics is in its ability to separate the total energy
requirement into two factors; thus for cohesive fracture

where ~o is the intrinsic failure energy or the energy actually

used to create unit area of new surface and ~ is the 'loss
function' dependent upon crack velocity, temperature and the
overall strain level in the specimen. In the simplest cast, 2~
corresponds to the energy needed to break unit area of interatomic
bonds across the fracture plane. The loss function has the formS
INTERATOMIC BONDING ENERGIES AT INTERFACES 691

kl
{ -----} (2)
kl - l: (3 g QV

where v is the reduced volume, kl and-g are explicit distribution

functions of input energy density in the specimen and S is the
hysteresis ratio of the solid (the fractional energy loss in a strain
cycle); S is generally dependent on crack velocity, temperature and
strain intensity and varies between zero for a perfectly elastic
material and unity for a perfectly plastic one. Clearly ~ = 1 for
perfectly elastic behaviour and (it turns out) achieves an infinite
value before (3 = 1.

The same equation can be applied to adhesive failure (see

Figure 1). If we consider an infinite block of adhesive adhering
to an infinite block of a rigid substrate, we obtain, for crack
propagation along the interface

e = eo ~ (3)

where ~ is the loss function fQr the adhesive only because the
substrate, being rigid, does not deform; e is now the energy to
break unit area of interatomic bonds acros~ the interface. If the
interface is secondarily bonded, e is identical to the thermodynamic
o

ADHESIVE
crack interface

11111111111111111111111111111111111
Figure 1. Crack at an adhesive-substrate interface.
692 E.H. ANDREWS

work of adhesion. Otherwise, 80 contains contributions from the

severance of primary bonds. If failure is not interfacial but
occurs through the adhesive or substrate, the cohesive failure
Equation (1) can be used. Indeed, for mixed-mode failure Equations
(1) and (3) can be combined mathematically to give a weighted
average failure energy.

APPLICATION OF THEORY

Taking logarithms we have,

Cohesive: log 2'1= log 2':6 + log iP (c, T, so)
. (4)
Adhesive: log 8 log 80 + log iP (~, T, so) (5)
For a given adhesive, a plot of log 2J or log 8 against any
parameter (e.g. crack velocity) on which iP depends should thus
produc~ a family of parallel curves displaced vertically by the
factor log (2::t. /8 ) where 8 is the 8 value for the n th
o on on 0

r
E
...
$ 6
\

2 ~~~--~~~--~~~~
-22 -18 - 14 -19 -6
OG (REDUCED CRACK SPEED Ims")

Figure 2. Failure energy master curves for SBR adhesive bonded to

various substrate films. From top to bottom, cohesive failure
energy for SBR; adhesive failure energy for the following sub-
strates: PET; PCTFE; Nylon 11; fluorinated ethylene-propylene
copolymer. Curves are superimposable by vertical shift.
INTERATOMIC BONDING ENERGIES AT INTERFACES 693

different substrate. This is regardless of the particular shape

of the curve, so that is is not necessary to know in advance how ~
behaves. Provided such a family of curves is obtained, they may
be superimposed by vertical shift to form a master curve, and the
vertical shifts used to determine 8 from a known value of , •
That is, provided the cohesive bond~ng term' cag ge determ~ned
or calculated by other means (as it generallyOcan, ), the inter-
facial bonding energy 80 may be precisely deduced.

ELASTOMERIC ADHESIVES

The overall failure energy 2'or 8 can be measured experiment-

ally in terms of the applied constr~ints and specimen geometry as
has been fully documented elsewhere. For a single-edge notch
specimen, for example

2,or 8 = kl cWo crit (6)

where W . is the input energy density at which the crack
o Crl.t 0
propagates and c is the crack length, whilst for a 90 peel test,

2' or 8 = plb (7)

where P is the peel force and b the breadth of the peeled strip.

for this determination. Figure 2 shows the definitive log 2'

Other test configurations (with their own formulae) may be employed
or
log 8 plot against crack velocity for an SBR adhesive bonded to a
variety of polymer film substrates. The substrates were made
effectively rigid by sticking a thin film of the substrate to a
metal base-block. For SBR, also, the Williams, Landel, Ferry
transform can be applied and the plot made against reduced velocity,
incorporating data recorded at different temperatures.

Table 1. Values of 80 and thermodynamic work of adhesion, wA' for

styrene-butadiene rubber bonded to various plastic substrates.

Substrate

Fluorinated ethylene propylene copolymer A 21.9 48.4

Polychlortrifluorethylene 74.9 62.5
Nylon 11 70.8 71.4
Polyethylene terephthalate 79.4 72.3
Fluorinated ethylene propylene copolymer B 68.5 56.8

of'0 The expected set of parallel curves is o~trJned. The value

can ~e measured in limiting fatigue tests' or else
calculated 1 and 8 derived for each substrate. Tables I and II
show the results,Ocomparing 8 with wA' the latter being determined
from contact angle measuremen~s6 For substrates where debonding
was purely adhesive, 8 0 corresponds closely with wA (see Table I),
694 E.H. ANDREWS

but where primary bonds were present '(Table II), 60 » wA and

failure is partly cohesivel2 •

Table II. Values of 6 and w for styrene-butadiene rubber bonded

to fluorinated ethylen~ propytene copolymer film etched with sodium
naphthalene to provide primary bonding sites on the substrate.

Etching time ( s) 60 /mJ m-2 wA/mJ m

-2

10 851 68.0
20 1170 70.2
60 1290 69.8
90 1620 71.1
120 1780 71.1
500 2420 72.7
1000 1990 71. 8

EFFECTS OF HATER ON EPOXY/GLASS BONDS

If 6 is determined as a function of time-of-immersion of the

bonded spegimen in water, extremely valuable information can be
obtained concerning the mechanism of attachment and the influence
of water on the joint. Such studies have been carried out on epoxy/
glass joints.

The adhesive bonding of polymeric resins to glass assumes

major importance in the glass reinforced composite materials used
increasingly in structural and aerospace application. Good bonding
is essential, not only for the structural integrity of the composite
but also to ensure adequate stress transfer from the polymer matrix
to the reinforcing fibres.

Epoxy resins are aapable of providing an excellent adhesive

bond to glass, but the adhesion deteriorates progressively in the
presence of moisture. This process is slowed, but not eliminated,
if the glass fibres are 'siz'ed' with coupling agents to promote
adhesion and durability.

A number of studies have bp-en carried out on the deterioration

of glass-to-epoxy bonds in water l3 ,14 but quantitative conclusions
are not easy to obtain because of the complex stress patterns
produced by differential thermal contraction between the resin and
the glass as the system cools from its curing temperature. Further-
more, conventional adhesion tests measure parameters which are only
indirectly related to the actual interatomic bonding energies at
the interface 8 • It is therefore difficult to monitor changes in
these bonding energies with exposure to water or with the addition
of coupling agents.
INTERATOMIC BONDING ENERGIES AT INTERFACES 695

RESIN
METAL WA SHER
PJE DISC
r--, r-~
GLASS

- ~
BRASS
BASE
BLOCK
PRESSURE

Figure 3. Test specimen for the Andrews-Stevenson fracture test

applied to epoxy-resin/glass bonds.

Many of these problems can be overcome by combining the

experimental technique of Andrews and StevensonlS with the
analytical method of GFM. The epoxy resin used throughout this
work was "Shell Epikote 828" a diglicidyl ether of bis-phenol A of
molecular mass ",370. The hardener was "Shell Epikure 114", a blend
of two cycloaliphatic amines with added benzyl al~ohol as an
accelerator. These components were mixed in the stoichiometric
mass ratio of 5: 2, cast and allowed to gel for 24 hours -at room
temperature, before post curing at l300 C for 1.5 h. Specimens
were cooled at 30 o C/h. The resulting resin-has a'glass transition
temperature at 72°C.

Three different silane coupling agents were employed

(supplied by Union Carbide Limited) as follows:

A 1120 N-S-aminoethyl-y-aminopropyl trimethoxy silane

A 187 y-glycidoxypropyl trimethoxysilane

A 1100 y~aminopropyltriethoxysilane
696 E.H. ANDREWS

The silanes were added to the epoxy resin, before casting,

in various weight proportions.

The basic specimen employed is illustrated in Figure 3 and

consists of a 2" x 2" square 'coupon' of 'Pyrex' glass on to which
is cast a squat cylindrical block of epoxy resin approximately
30 l)ll)l in radius and 8 ~ 11 l)ll)l in height.

The drilled central hole is covered by a thin circular disc

of PTFE before the resin is cast, the disc thus creating a non-
adhering region which acts as an enclosed circular crack which can
be pressurized internally through the hole. This system provides
a totally plane-strain fracture mechanics test in which, moreover,
stresses due to differential thermal contraction between the resin
and the glass cannot assist crack propagation as long as the latter
is interfacial lS • From the critical pressure for failure and the
specimen dimensions, the fracture energy 2~(cohesive} or 8
(adhesive) may be evaluated as discussed elsewhere lS • Crack velocity
was measured by high speed photography, again as detailed in the
earlier paper.

0
200
I
0
/
863 h

I~
!
e
...
\ I \X 0
ID
I XO~O
100
I C0
III
0
00 C 1175 h
h
N
I Ix~ /\ /
I 1800 h -
X X
X X-l X

50 100
CRACK VELOCITY/ms"
Figure 4. Fracture energy versus crack velocity data for epoxy
resin bonded to glass (with silane A187 at 0.1% concentration).
Specimens il)ll)lersed in water at 80 0 C for the times shown. Note that
although data shown here are not numerous, the general shape of the
curves is common to a very large number of tests using different
substrates, coupling agents and il)ll)lersion conditions.
INTERATOMIC BONDING ENERGIES AT INTERFACES 697

Results

The basic data obtained form the fracture tests take the form
shown in Figure 4 where cohesive (2~) or adhesive (8) failure energy
is plotted as ordinate against crack velocity c
as abcissa for
various times of immersion in water at BOoC and 7.B pH. Immersion
causes a steady overall decrease in failure energy without any
marked change in the velocity dependence. Resin incorporating
coupling agent suffers a much slower degradation of interfacial
strength. For example, a given system containing coupling agent
deteriorates after B60 h to about the same degree as obtained
after 40 h with unprotected resin. Data for all resin mixes can
be displayed in the same form as Figure 4. Following the super-
position procedures outlined earlier results may be obtained of the
kind displayed in Figure 5, which is a 'master curve' obtained
by vertical super-position of the logarithmic data for resin
conta~n~ng Al120 coupling agent in various proportions and
various immersion times exceeding 100 h. The shape of the master
curve does not appear to depend upon the coupling agent
employed, its concentration, the time of immersion (after the first
100 h), or the pH of the immersion medium.

2 S

0 29
J

1 S 2 9
OG ( R CSPEE I s" )

Figure 5. Master curve for failure energy formed by super-position

of data such as shown in Figure 4. Data shown are for 0.05% of
A 1120 silane in epoxy resin bonded to glass, immersed in water for
various periods of time from 500 to 2000 h at BOoC.
698 E.H. ANDREWS

16

(J

o 8.e \
\
-8.S ~2
I ~\
-\ , 99L...(1-I-~-"""""""'--:I~e8::':e.......-""--.......-1.-~
29:ee;-
OURS IMMERSION IME AT S8 DEC ,C

Figure 6. Collected data showing variation of 8 0 with time of

immersion. (1) 0.05% A187, pH = 13 (2) no silane (3) 0.05% A187
pH = 2 (4) 0.05% A187 (5) 0.05% Al120 (6) 0.1% A187 (7) 0.1% Al120
($) 1.0% AllOO.

Reaction Rate Constants

As outlined earlier, the vertical shifts required to give a

master curve give 8 provided ~ is known. Comprehensive results
for log 8 are plot~ed in Figurg 6 against time of immersion for
a range o~ coupling agents, concentrations and pH values. A linear
dependence of log 8 upon time is characteristic of a first order
chemical reaction, ~he negative slope of such a plot being what we
shall call the overall rate constant, k':

k' = [A][B) k 80 /2.303 (9 )

where [A] and [B] are respectively the concentrations of hydroly-

sable bonds and water at the interface, and kSO is the true rate
constant at 80 0 C. For comparative purposes it is sufficient to
talk in terms of k' since the water concentration should be sensibly
constant at its equilibrium uptake value (some 8% at 80 0 C).
INTERATOMIC BONDING ENERGIES AT INTERFACES 699

FIGURE 7
B.04
°
8.£13

r.:
~ D
~
181
8.£11 0,
DOD
/0/

--
'''-.-0
D...... D.....
6
pH
/
8 18 12 14

Figure 7. Hydrolysis rate constant as a funct~un of pH for (circles)

resin without silane and (squares) resin with 0.05% A1B7.

Variation of Rate Constants

Finally, we may seek information on the interfacial hydrolysis

process by examining the variation of k' with the type and
concentration of coupling agent and with pH.

Figure 7 shows k,' as a function of pH. Hydrolytic activity

is at a minimum for neutral pH, rising dramatically for both acidic
and alkaline conditions. This is in distinct contrast to the 17
results obtained for the same resin using titanium as a substrate
There, acidic conditions accelerated the hydrolysis reaction but
alkalinity retarded it. This difference constitutes strong
circumstantial evidence that the hydrolysing bonds are genuinely
interfacial. If the locus of' hydrolysis were the bulk resin, no
difference should be observed between glass and metal substrates.
700 E.H. ANDREWS

Dry Value of eo
All the data, regardless of resin composition and pH, converge
within experimental error to a single value of e at t = 0, i.e.
for the dry resin/glass bond. This point is, ofocourse, experiment-
ally inaccessible since without immersion all specimens fracture
cohesively and no information on the strength of the interface can
be deduced (except that e > 2'). The extrapolation of data for
specimens at different pHovalue~ should, of course, give a single
intercept, since all these joints are identical before immersion.
Nevertheless, the fact that they do converge to a single value of
e is strong evidence that this value can be taken as the actual
igterfacial bonding energy of the fre~hly made interface. The
point of major interest is that even specimens differing widely in
silane content also yield the same e (dry).
o
The value in question"is 7.25" Jm- 2 , some twenty-four times
larger th.<ln the expected van der Waals interaction energy for such
a system16 . This indicates that at least some of the interfacial
bonds are primary, but a precise proportion is more difficult to
specify. A figure of 25 to 50% of primary bonds would appear to
be reasonable, based on the relative dissociation energieg of
primary and secondary bonds. However, as Ahagon and GentlOpoint
out in their work on elastomeric adhesives bonded to glass, a
single primary interfacial bond may require many times the energy
for rupture of a single C-Cbond because it is attached to a length
of network chain. Taking this factor into account, the magnitude
of e (dry) can be explained if between 30% and 60% of all inter-
faci~l bonds are primary in nature. The lower figure applies if all
network chains terminating on the surface are chemically bonded to
it and if the cross-link density is the same at the interface as
in the bulk. The higher figure would apply if the network chain
effect were totally suppressed by chain immobilisation at the
interface.

The most surprising feature of the e (dry) result is that it

appears insensitive to the quantity of sil~ne added to the resin,
even though the rate constant for hydrolysis is greatly enhanced
by the coupling agent. The obvious explanation is that, with the
addition of silane, the total number of interfacial primary bonds
remains the same but that the type of bond changes. That is,
primary bonds which are established in the absence of silane
(possibly R-C-O-Si) are progressively replaced by different bonds
(possibly R-Si-O-Si) until the interface is saturated with silane-
related bonds. Under dry conditions these two types of bond would
have very similar strength, but under hydrolytic action the silane-
related bond would degrade much more slowly.
INTERATOMIC BONDING ENERGIES AT INTERFACES 701

0,004

0,003
o

\\0\
0,001
~0"-0
\..0----
-
0 - - 0-
0 -_ __ -0

0,2 O,4 O,6 0,8 1,0

CONCENTRATION OF COUPLING AGENT
Figure 8. Hydrolysis rate constant as a function of coupling agent
concentration. Circles All20, squares Al87.

The variation of k' with silane concentration is shown in

Figure 8 for silanes Al87 and All20. There is an almost linear
decrease of k' with concentration up to 0.05% w/w but at concent-
rations as low as.O.l%, saturation is beginning to occur.
Although k' continues to fall slowly, saturation is largely
achieved at concentrations of 0.5% for both silanes.

CONCLUSION

Fracture mechanics experiment~ combined with the GFM

analysis of the data, have made it possible to measure directly
the interfacial bonding energy, e , for various adhesive/substrate
systems. The role of this bonding in determining the mechanical
strength of joints is also made clear and quantitative in such
studies. Once e is known, its dependence upon such important
variables as sur~ace pre-treatments, coupling agent type and
concentration and environmental exposure can be determined. The
physical and chemical mechanisms involved may then become clear.
702 E.H. ANDREWS

REFERENCES

1. C.A. Dahlquist, in "Aspects of Adhesion", D.J. Alner, Editor,

Vol. 5, p. 183, University of London Press, 1969.
2. D.H. Kaelble, J. Adhesion, 1, 102 (1969).
3. K.L. Mittal, in "Adhesion Science and Technology", L.H. Lee,
Editor, p. 129, Plenum Press, 1975.
4. K.L. Mittal, Polym. Eng. Sci., 17, 467 (1977).
5. E.H. Andrews, J. Materials SCi.-,-9, 887 (1974).
6. E.H. Andrews and Y. Fukahori, J. Materials Sci., 13, 1307 (1977).
7. N.E. King and E.H. Andrews, J. Materials Sci., 13-,-1291 (1978).
8. E.H. Andrews and A.J. Kinloch, Proc. Roy. Soc. (Lond.) A, 332,
385 (1973). -
9. G.J. Lake and P.B. Lindley, J. Appl. Polym. Sci., ~, 1233
(1965) .
10. A. Ahagon and A.N. Gent, J. Polym. Sci., (Polym. Phys. Ed.),
13, 1903 (1975).
11. ~J. Lake and A.G. Thomas, Proc. Roy. Soc. (Lond.) A, 300, 108
(1967) .
12. E.H. Andrews and A.J. Kinloch, Proc. Roy. Soc. (Lond.) A, 332,
401 (1973).
13. J. Comyn, D.M. Brewis, R.J. Shalash and J.L. Tegg, in
"Adhesion 3", K.W. Allen, Editor, P. 13, Applied Science
Publishers, Barking, Essex, 1979.
14. J.P. Sargent and K.H.G. Ashbee, J. Adhesion, 11, 175 (1980).
15. E.H. Andrews and A. Stevenson, J. Materials Sci., 11, 1680
(1978) .
16. E.H. Andrews and N.E. King, J. Materials Sci., 11, 2004 (1976).
17. E.H. Andrews and A. Stevenson, J. Adhesion, l~,-Y7 (1980).
REVIEW OF CONTINUUM MECHANICS FACTORS

IN ADHESIVE FRACTURE

M. L. Williams

School of Engineering
University of Pittsburgh
Pittsburgh, Pennsylvania 15261

The possibility of treating adhesive fracture as

an engineering analysis problem in continuum mechan-
ics has been emphasized in previous reviews, with ap-
plications to designs which utilize bonded surfaces and /
or various composite materials. Cost-effective use of
this analysis ability will ultimately depend, of course,
upon the accuracy of fracture and structural life pre-
diction, as well as the availability of methods to con-
duct non-destructive examination.

This review summarizes some of the principal

implications from continuum mechanics which influence
the character of adhesive fracture, including some new
experimental data for the rod pull-out problem. The
intent is to furnish an overview of major mechanics
factors to those chemists whose interest lies in making
direct contributions to the engineering design of adhe-
sively bonded structures.

703
704 M.L. WILLIAMS

INTRODUCTION

Continuum mechanics has been used to analyze various ad-

hesive bonding configurations since at least the first part of this
century. As the use of bonded components became more widely
recognized and reliable as a design option, the subject of adhesion
has correspondingly attracted increased attention from both scien-
tists and engineers. Efforts by the former have tended to focus
upon a basic understanding of the chemical structure of the mating
components such as reflected in the nature of the interfacial forces
and the relative contact or wetting angles. On the other hand ef-
forts of engineers have gravitated toward producing acceptably
accurate assessments of various aspects of structural integrity,
including the development of representative and reliable test
methods for quality control.

Since the author presented his first continuum mechanics

review to the American Chemical Society approximately ten years
ago, 1 followed by one with And2rson tel the International
Congress on Fracture in 1977, an excellent review, dealing with
both the scientific and engineering ap~roaches, has recently
(1980-82) been presented by Kinloch. In addition the monograph by
Anderson, Bennett and DeVries 4 has provided considerable ampli-
fication, including many important aspects of testing methods.
With these recent references readily available, it is considered
sufficient for the present purposes to call attention to this review
literature, reiterate major {continuum mechanics} points, and
provide supplementation in a relevant manner.

Because of this ready availability of the recent reviews on

the subject of continuum mechanics and adhesive fracture, and
particularly the full exposition presented so recently by Kinloch, 3
the present coverage will be directed to chemists with the intent
of transmitting an insight into the type of conclusions which might
reasonably be expected from co-workers in mechanics. Conse-
quently, and despite its importance, there will be little said of the
importance of a Griffith energy balance type approach to cohesive
or adhesive fracture in elastic, plastic, or viscoelastic materials.

Instead, after some initial remarks relating to the types of

configurations presently used in the community, the first part will
be directed toward the utility of continuum mechanics in interpret-
ing fracture through the mechanical properties of the materials
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 705

involved, the effect of local or global geometry of the part, and the
importance of loading orientation(s) with respect to the line of frac-
ture. The second part will be devoted to a discussion of identifying
potential adhesive fracture locations primarily through a readily
definable knowledge of singular stresses. The third part will sup-
plement the Kinloch review by discussing certain aspects of the rod
or fiber pull-out problem of interest in composite materials, civil
engineering, and automobile tire design, with special emphasis on
some of our continuing experiments and numerical analysis of the
partial debonding and adhesive crack growth from, along, around
and into the glass-rubber interface of the embedded rod. The con-
cluding section deals with some aspects of adhesive fracture which
appear to the author to be a collective responsibility among chem-
ists, physicists, and mechanicists if a more rapid and enhanced
understanding of the phenomenon is to be attained.

CONFIG URA TIONS

Generally speaking the various geometries discussed in the

literature seem to fall into two general categories: practical de-
sign components and engineering data specimens. Neither of
course is completely exclusive. Very practical single and double
shear lap joints have been used as test specimens to assess the
import of n-ormal stress in the adhesive perpendicular to the angle
of pull; on the other hand, the pressurized blister test configura-
tion used to measure the fundamental specific adhesive fracture
energy ()$'a) also yields stress and deformation formulas useful in
analyzing the (blister) debonding of paint and thin films or the
welding efficiency for explosively bonded steel plates. I (Figure 1.)

The principal adhesive configurations are as follows: stress

analyses for all of them are available to varying degrees of accura-
cy. The butt-end joint, including its tubular relative called the
"napkin ring" butt joint, had its early history in connection with
welded joints in metals and, in rubber, with bridge bearings and
shock mounts. The group at the British Rubber Producers Associ-
ation, stimulated by the early work of Rivlin, Thomas, Andrews,
Gent and co-workers, analyzed the effective modulus in tension
and compression of thin rubber disks bonded to rigid mountings.
This configuration was subsequently adopted in the American
rocket industry to study triaxial tension failure and flaw genera-
tion in the rubber liner which was bonded to the rocket motor case.
706 M.L. WILLIAMS

",-,MM _V'DH"""
Figure I-A. Single lap shear lat.

Figure 1- B. Double lap shear lesl.

Y
3 rM.+D£BOHl>£D
LEH011l

Figure 1- C. T.p«ltest.

ADRESrvt:

Figure 1- D. Double c&fttilever Figure 1- E. TorsioNJ1 sh~ar sucss spt'cimDt.

cleavare speciaen

Figure 1. Several typical adhesive test specirnens 2 ,4.

CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 707

ou . . .

T ..TI
CII

- T-!-
T.... I. UII:II
z

I
APPL IED LOAD

...
~

1-- - - 1."01 ~
1 11'- co........ •
Figure I-F. ... puu,"", r--....--- ..............
.... t COftfi ......U ...

Filure 1-0. C- ICI.

IOLlT'KANI· .. EMUAIQ:

'.I - '.ID! .

1r- I. ' D! .

r~~~~-..............~......-...J
• U. FlT'1' life

~ pus.suu

r uXlB U TU1UlfO

Figure I-H. Blisl~' I~SI SpUim~fI.

Figure 1 (continued)
708 M.L. WILLIAMS

The stresses in this "poker-chip" speciInen and its variations, in-

cluding consideration of the stress singularities at points of the
interface, have been used to examine the propensity for nominally
normal stress debonding. (Substantial portions of the limited dis-
tribution Messner work has now been reproduced. 4) The split dou-
ble cantilever beam (DCB), adopted from the Obreimoff test for
the mica fracture energy and Gilman's test for the splitting forces
between sheets of mica, and the pres surized or point-loaded bUs-
ter test used by Williams and Jones and by Malyshev and Salganik,5
also measure primarily the normal stress or opening mode of
fracture (Mode I), but contain important shear components.

On the other hand, single and double shear lap joints with and
without scarffed ends deal primarily with the shear mode of frac-
ture (Mode II) as described in the Kinloch review. The conical plug
pull-out test, originally introduced by Anderson 9 in conjunction with
dental research dealing with gold fillings, has the virtue of varying
the proportion of the Mode I (normal stress) and Mode II (shear
stress) contribution to fracture. It may also be used for torsion
tests (Mode III) by twisting the plug. In one of its limit forms - -the
straight pull out- -it represents the very practical problem faced
by automobile tire manufacturers who wire-reinforce the tread and
wall. Alternately, and usually in shorter lengths, but not neces-
sarily much different length-to-diameter ratios, the pull-out or
"tooth-pick" problem is the elemental problem in fiber-reinforced
materials. Our experimental and analytical work, growing out of
the initial experim.ents of W. B. Jones and Masahisa Takashi, was
reported earlier 2 and has been extended by Atkinson, B~tz,
Smelser, and Avila 6 to include definitive experimental
data and limited parametric variation of rod geometry and debond
length and location making use of finite element analysis. None of
our analyses, nor any known to us, is yet complete enough to as-
sociate the reinforcement of a single fiber in an infinite medium
with that of randomly oriented and distributed multi-fibers.

The DCB and blister specimens were originally introduced

to measure Mode I specific adhesive fracture energy (O'a), which
is the continuum mechanics adhesive analog of the specific cohe-
sive fracture energy (<rc) commonly denoted in metals as "fracture
toughness." The latter specimen has the virtue that, being circu-
lar, it has no "sides" to become environmentally contaminated,
although the test environment can be changed at the crack front by
changing the pressurizing m.edium.. As remarked earlier,
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 709

"membrane" thickness blister tests with combinations of pressure

and temperature have certain applications to thin film adhesion.
Further extensions of the basic blister configuration have been
used to assess the basic influence of adhesive layer thickness and,
by a double sided blister, the stiffening effects of different ad-
herends.

Final mention should include various partial debonding anal-

yses of cylindrical and spherical inclusions, and the related cases
of elliptical rod and ellipsoidal geometries, which are important
in granular composite media such as cement-aggregate concrete,
or in the potentially important, fairly new sulfur-aggregate con-
struction material. (Interestingly enough, one of the major draw-
backs is the apparent lack of adherence of the smoother, larger
particles which suggests that present continuum models should in-
clude interfacial layers of finite thickness, as introduced into the
blister test, and from the chemical side the introduction of appro-
priate surfactants. *) In more generality, granular composite
material adhesives technology may well apply to models of pow-
dered metals, and even the increasingly interesting metal matrix
composites.

CONTINUUM MECHANICS

A mechanics analysis can be expected to yield information

on the influence of mechanical properties of the materials involved,
geometric configuration of the design, and the type of loading.

The first- -mechanical properties - -is reflected mainly in the

stress-strain law and includes the effects of time and temperature,
separately or together. Separately one may assess thermal strain
through a combination of the coefficient of thermal expansion (0<: )
and the temperature change (AT), or the magnitude of creep defor-
mation at essentially constant temperature. Together, time (rate)
and temperature are important for many adhesives and appear in
the temperature-reduced time, WLF 7 shift factor (aT). Discov-
ered experimentally, it comes about through a semi-empirical
correlation between viscosity and free volume of viscoelastic
materials incorporating another material property, the glass
temperature (T g). The fundamental material descriptors of stiff-
ness (modulus of elasticity, E or G) and degree of incompressi-
bility (~v lv, Poisson's ratio,) are typical quantities that form

*Private communication Dr. D. Saylak, Texas A&M Univ. ,

1982.
710 M.L. WILLIAMS

the matrix of material constants which for most materials relate

the six components of stress (three shear and three normal) to
similar components of strain.

The second- - geometric configuration- -has already been im-

plied in the reference to the categories of geometries discussed in
the literature, e. g., design components such as single and double
shear lap jOints and specimens used to gather engineering data on
the properties of the materials involved. For example, the over-
all distinction between single and double shear specimens in terms
of the presence or absence of a moment imposed upon the joint is
commonly accepted. Other perhaps less obvious geometric changes
however can also be assessed. One of the more important geo-
metric effects pertinent to adhesion is the thickness (h) of the ad-
hesive interlayer. 4,8 Another is the presence or absence of fillets
at sharp corners, usually interpreted by stress analysts in terms
of a stress concentration factor, and by chemists, in terms of a
wetting angle between the adhesive and the adherend. Many of the
geometric considerations, however, relate mainly to the arrange-
ment of material, say for low weight of the assembly. As such
they are of major concern to the stress analyst who frequently
finds it necessary to employ rather sophisticated and powerful
numerical techniques, such as the "finite-element method, " to
deal with these design configurations, although such methods are
occasionally used even in analyzing engineering data specimens.

The third- -type of loading- -deals with the manner in which

loads are oriented and applied to the entire design or the individual
component. It may be either mechanical or thermal in nature,
imposed in a time dependent way (shock, steady, repeated or
oscillatory fatigue), and from one or more directions. For our
present purposes it is sufficient to think in terms of a simple
specimen subjected to steady mechanical loads imposed from one,
two, or three directions at right angles to each other, along with one,
two, or three sets of shear forces. Specializing to simple geo-
metries subject to adhesive fracture, the mechanics analyst re-
fers to a crack normal to the primary direction of applied loading
as subjected to "Mode I" (normal stress only) loading. Similarly
if shear loads are applied parallel to the crack, it is subjected to
Mode II (shear) or Mode III (shear) loading.. Figure 2 illustrates
the three modes for cohesive fracture. As will be seen later,
whereas for an isotropic homogeneous cracked specimen, as in
Figure 3a, the mode descriptions are unambiguous, the same
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 711

Figure 2. The basic modes of Crack Surface Displacements.

a a
tff, , fttt,

_
~
r'
c==::a-~oLe-

2a I--
x, u -
~2a~
r'
c:=::::a ~~- x. u

t t ta t t t t ta t t
(a) (b)

BOUDdary CODdl tlolUl: Ix I :. • BOUDdary CondltlollS: Ix I :. •

(1)" .. (:1:,0) - 0 (1) ... (x.o) - 0
(2) ,.(x,o) - G(au( .. ,o)/ay + a y h:,o)/03"J - ° (2) u( .. ,o) - 0
or becau.e of (1). equlvaleDtly
(2a) au( .. ,o) - °
iJy
Pllf\U"e 3a. Cohesive Fracture Fll(\lre 3 b. Adhesive Practure

a
cr
- V.!
"a
EYe

Figure 3. Comparison of essential boundary conditions for

cohesive and adhesive fracture. The difference
is slight and concerns only whether the lateral
displacement u(x,o), or its normal derivative,
au(x,o)/ay, is prescribed. Both sets of 2 bound-
ary conditions lead to singular stresses •
712 M.L. WILLIAMS

can not be said for a bi-material case, say cracked along the bond
line. If a simple one material center-cracked specimen (Figure
3a) is loaded perpendicular to the crack it will be subject solely to
the opening mode (Mode I), i. e., there is no shear stress along
the line of crack prolongation. If, on the other hand, the material
is different in the half-planes above and below the crack (Figure
3b), a load applied in the same direction as before will still create
an "opening mode" but there will be shear along the line of crack
prolongation. Hence the common intuitive usage, which implies
for bi-material adhesive specimens that an "opening mode" pro-
duces only a local normal stress (and no shear stress) at the crack
tip, or, similarly, a "sliding mode" produces only a local shear
stress (and no normal stress), is incorrect. As will be shown
later, the intuitive conclusion is, however, substantially correct
most of the time for the usual material combinations. Neverthe-
less, it is important to be aware of the true situation so that im-
proper extrapolations will not be made.

Indeed a preliminary experimental study was conducted by

Anderson 9 ,4 using a cone pull-out test in which he could vary the
proportion of Mode I: Mode II: Mode III. The initial evidence sug-
gests that a different amount of new fracture area (A S) is gener-
ated in each of the three modes of loading. This experimental
observation is in contrast to the usual assumption in the analysis
using projected new fracture area for each mode, but allowing for
different specific adhesive fracture energy (?t-a , in-Ibs /in 2 ) for
each mode. Hence this preliminary conclusion suggests that the
physical fracture process could be incorrectly modelled by the
continuum mechanics analyst and a reassessment in conjunction
with chemists and physicists might be very worthwhile.

SINGULAR STRESSES IN MULTI-MEDIN:'

In mathematics it is commonly accepted that a knowledge of

singular solutions is essential for representing complete solutions
in the large. In this sense there is an important parallel in
fracture, particularly to the extent that conclusions based upon a
linearly elastic material and small deformations may represent
guidelines for the actual behavior. For this reason it is important

':'The examples given are confined essentially to two materials,

although in principle the approach can readily be extended.
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 713

to em.phasize the existence of m.athematical singular solutions in

both cohesive and adhesive fracture analysis, as well as sim.ulta-
neously recognizing their lim.itations. Clearly the infinite stresses
predicted m.athematically at certain sharp corners and at crack
tips in an interface can not exist physically- - but the stress m.agni-
tudes m.ay be near enough to it to cause practical trouble!

First for cohesive fracture, e. g., Figure 2 (I) - Mode I. As-

sum.ing the crack is in a thin sheet of unit thickness, the material
m.athem.atically occupies a region 2 iI' about the crack tip and the
crack which has two unloaded (free) faces encloses zero space
between them.. Starting from. a more general case of a V -shaped
crack having two free sides and an opening angle W between them.,
the character of the stress variation in the vicinity of the crack in
terms of the radial distance (r) from. its tip can be obtained 10 as'~

cr .-./t A (~) - 1 f (S) r:J.... = 21T' - w (1)

where A (cJ,.) has been extracted from Reference 10 and presented

in Figure 4. For the special case of a crack,W+O or" = 21'(,
A(2tr) is found to be 1/2. In such a case the exponent of the radius
becom.es negative, and

(2 )

and the stress becom.es unbounded at the origin with its character-
istic inverse square root singularity.

Equation (1) reveals another interesting feature for thin

sheets. At the particular angle,ol':', at which A (0(*) = 1, there will
be ~ singular or infinite stress at the origin. Reading from. the
Figure 4 for free-free edges, one finds A':' = 1T. (For smaller
included angles, the stress will be even less severe at the origin. )
More generally, depending upon the boundary conditions on the
faces, the stress at any notch tip will remain finite whenever an
opening angle exists for which A (c<.) ~ 1.

For a case of interest to adhesion, consider a rigid adherend

upon which an adhesive material having an (included) wetting angle
bt) is laid down, or a scarffed joint with an (included) bond angle

*Note that the A used here 10, say AW' is different from that
used in Reference 4, say A.A' such that AA = 1 - AW.
714 M.L. WILLIAMS

to the base plate (ol). In this situation the Il1aterial would have an
included angle (<:1) and rigid-free boundary conditions. FroIl1 Fig-
ure 4, one finds A(cJ..*) = 1 at cJ... * = 62 0 +. Hence one concludes that if
the included angle A*~ 55 0 , + there will be no high stress at the
V-tip. Note that the Il10re the "adhesive" does not "wet" the base,
the larger c/o.. is, and the stronger the stress singularity and the
higher the stress at the V -tip, consistent with experience.

The above results were deduced using the data for thin
sheets (plane stress). 10 A similar curve can be obtained for
claIl1ped-free boundary conditions and plane strain, i. e., infinitely
thick sheet on an infinite rigid base, and yields 45 degrees as the
largest (included) angle without inviting large stresses at the tip
of an incoIl1pressible (v = 1/2) adhesive. i. e., sm 2,Ad..= 1 - A,2 sin2.:L .

Other pertinent deductions can be quickly made. For ex-

ample as will be used later, consider the cross-section through
the geoIl1etry of the circular rod pull-out test (Figure 5). Assum-
ing the rod to be rigid cOIl1pared to the incompressible rubber
Il1atrix in which it is bonded, one can easily identify two Il10re
"corner probleIl1s." Between the free surface and the rigid (fixed)
side along the rod, the Il1atrix Il1aterial occupies an included angle
of 90 degrees. Is the stress (Il1athematically) infinite at the free
surface eorner? And then if it is, how strong is the singularity
cOIl1pared to the other possible singular location at the corner of
the rod tip? Here, with the rod tip flat, the matrix material is
rigid (fixed) along the bottoIl1 of the rod face and its side, thus sub-
tending an included angle of 270 degrees. AssuIl1ing for the mo-
ment the cross section of Figure 5 represents indeed a thin flat
sheet (plane stress) instead of the actual axially symIl1etric geOFll-
etry it really is, one reads froIl1 Figure 4

A(tr/2) = 0.76 =) CT-IL -0. 24

clamped-
free
-0.65
).,(31\/2) = 0.35 = > (I"""";r,
claIl1ped-
clamped

Thus there are stress singularities at both locations and the one at

+The curve in Figure 4 prepared for V = O. 3 shows oi~ = 62 0 •

A siIl1ilar curve prepared for'" = 0.5 would actually give 55 0 . (Ref. 10).
 6.0
(")
CASE NO. BOUNDARY CONDITION o
\ I FREE - FREE Z
--l
2 CLAMPED - CLAMPED Z
C
5 .0 \ 3 CLAMPED- FREE C
s:
s:
m
(")
\ :r
\ \ »
z
4 .0 ,,_,Ro>'-I P 0( _r
o
(J)
""T1
...::: »
.,
\ 1\ (")
--l
!r o
3 .0 \ \ CD :II
:: (J)
::::>
:;;: Z
z
1\ I~ »
o
:;;: :r
m
(J)
2.0 \ \ ~~
~ <:
~® m
........... ""T1
:II
\ "'- »
(")
--l
"'r-- r-- r--- ~
1. 0
" C
:II
~ "-...... ............." ~ m
~ I--
r -®
-
o 40 0 80° 120 0 160 0 200 0 240 0 280 0 320 0 36 0 0
VERTEX ANGLE- C(

Figure 4. Variation of mlnlmum real part of the stress singularity exponent ~

wit~ opentBg angle for various boundary conditions: plane stress 01
(~ - 0.3) .
716 M.L. WILLIAMS

tt

..;.
I
, ~,

~-+--....I'

Figure 5. Profile of hypothesized unbonds

ofaxisymmetr~c body in rod
pull-out test •
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 717

the corner of the rod tip is stronger. (Actually the behavior at the
interface of the rod is better associated with plane strain, 11)
although the conclusions are qualitatively similar. This situation,
in the limit, can be understood by thinking of the rod as having a
very large radius so that the material approaches an infinite thick-
ness as the rod radius becomes infinite. The cross -section in
Figure 5 would thus be the same for either plane stress or plane
strain, but Figure 4 for the critical eigenvalue J\. (~) is determined
for plane strain from an equation containing Pois son I s ratio which
is slightly different from the one for plane stress. IO

As a further example, while pursuing this limit analogy of

axial symmetry to plan\? strain, return to a previous plane strain
conclusion that for a 45 degree included angle between rigid and
free faces there would be no stress singularity. This result would
suggest that a cylindrical butt-end adhesive joint, or the "poker-
chip" test of incompressible rubber as used in our early Caltech
work 1Z and analyzed numerically by Messner to have high stresses 4
at the periphery, would have finite or negligible stresses there if
45 degree indentation fillets were incorporated instead of the 90
degree (straight sided) configuration we used. Such a design would
give increased assurance and reliability of more regularly obtain-
ing the desired hydrostatic tensile condition at the center of the
specimen.

As many bi-material configurations do not involve the con-

venient rigid-incompressible material combination, a general for-
mulation was set-up to determine the character of stress singular-
ity to be expected at the crack tip along the interface between two
materials with different properties. 13 While originally prepared
to investigate layered geophysical fault lines, the analysis is
easily adapted to adhesive fracture. The most intriguing discovery
was a stress singularity whose order was still the characteristic
inverse square root type, but it was modulated by an oscillating
trigonometric function such that ins'tead of Equation (Z), one obta ined
behavior such as
I
- - - sin (b Info) (3)
v;:;:-
Also the shear stress was of the same character and did not vanish
unless the two materials were identical, which was reflected in the
material dependency of b = b (El, EZ' V 1, ~Z), i. e., b = 0 if
El = EZ = E and V 1 = -0 Z = V' thus recovering the result Equation
(Z) for cohesive fracture.
718 M.L. WILLIAMS

This result, namely that both normal and shear stress co-
existed along a bi-material interface near the crack tip, became
clear physically after a bit of reflection, but produced the afore-
mentioned result that a Mode I opening mode would have not only
the expected normal stress across the crack, but also an accom-
panying shear stress. (By extension to Mode II, the converse is
evident.) Because of the existing tendency for the uninitiated to
speak of "shear" or "normal stress debond" adhesive tests, it is
important to make some assessment of the relative contributions
of each in, say, nominally Mode I or II loadings. Fortunately this
can be done through a relatively simple, but mathematically so-
phisticated, example reproduced 4 in Appendix A. 'The principal
result presented in this appendix permits the interested reader
to determine the exact percentage contents of shear to
normal stress in Mode I, i. e. (1"'/a-)r, and normal to shear stress
in Mode II, i. e., (O"/I')II' for the desired material combination

(4 )

where '3 is material dependent and f is distance from the tip.

The oscillatory character of the predicted stress singularity

should also be understood. From physical grounds, there must be
a shear stres s along the bond line if the materials above and below
it are different (Figure 3b). Their lateral contraction is different;
for them to maintain continuity they must be displaced relative to
each other, which requires the shear stress. But why oscillatory?
The net shear stress, say on each side of the (symmetric speci-
men) centerline must vanish, so the shear stress must reverse
direction, which it does, in between each node of the trignometric
cycle. Finally it was shown by Malyshev and Salganik 5 and
Erdogan 14 that the oscillatory character was confined very, very
closely to the crack tip, well within any local distortion due to,
say, plasticity, so that the basic solution still represented the
interface stresses quite well.

As a final comment on the utility of critical stress singular-

ity evaluation in bi-material combinations at various included
angle combinations, the tabulation completed by Hein and Erdogan 15
should be particularly noted. Some of the cases are quoted by
Anderson et al 4 , and one of them is of special interest: plane
stress or plane strain of two quarter planes bonded along a com-
mon edge with the other two faces unloaded and continuous. In
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 719

general there is a weak singular stress where the bond line inter-
sects the free surface and varies from r -0.22 to r -0.00
(plane stress) as the modulus ratio of the two materials varies
from several orders of magnitude to unity, respectively.

THE ROD FIBER PULL-OUT PROBLEM

The rod or fiber pull-out problem has stimulated consider-

able and long standing interest, perhaps from an early civil en-
gineering need to analyze the stability of pilings by friction and
length, or the strengthening of concrete by reinforcing bars.
Another source of early interest was in aeronautics due to the
drive toward low weight design which led to integral or mechan-
ical attached stiffeners to thin sheet. Thus the transfer of load
between the strong, stiff member to the surrounding material be-
came of considerable practical importance. Inasmuch as many of
the previous analyses have a direct bearing upon the stress analy-
sis of a rod or fiber wholly or partially embedded in a surrounding
softer or harder matrix, and furthermore, current composites ma-
terials development is properly concerned with maintaining the in-
ternal interfacial integrity, it is appropriate to supplement the
Kinloch review in this regard. A selected limited bibliography
biased toward recent contributions has also been prepared and is
appended.

Because of the relative scarcity of precise experimental data

for the pull-out problem in an adhesive debonding mode, and stimu-
lated by the somewhat unusual crack growth behavior in the related
situation of thermally driven debonding of rubbery rocket fuel from
the surrounding motor case, the major emphasis was placed upon
experiments to dis·cover the general nature and progression of the
failure. In the rocket problem the fuel essentially filled a rigid
cylindrical case when positioned and cured vertically. The top
surface was free of external loading, flat, and bonded to the case
with very little fillet. In view of the foregoing remarks, the mate-
rial occupied a space subtending 90 degrees with its faces a.bout the
vertex being rigid and free. When the temperature changed in a
way to exert a thermal stress across the bond line, there was a
tendency for fracture to initiate at the 90 degree corner and quickly
proceed to propagate along the case interface. What was surprising
however was the fact that the crack surface only proceeded a short
way, i. e., much less than a case radius, before coming to a stop.
720 M.L. WILLIAMS

This stable propagation is rather unusual, though observed early

by Benbow in the indented punch problem. 16 In that case the
broadening conical "mushroom-ring" collar-like fracture was not
catastropic or unstable, but required increasing load to extend it.
Similarly in the bonded rocket fuel, the crack front only proceeded
along the interface while there was sufficient strain energy being
released to drive it, as subsequently confirmed by stress analysis. I7

It developed that the fracture surface in a rod pull-out test

behaved in somewhat the same manner. In Figure 5, for example,
there are several possible fracture modes shown, depending upon
the as sumed point of initiation, but it was not a priori clear from
analysis as to the origin or propagation of the failure. After con-
siderable exploratory work by W. B. Jones and G. P. Anderson
while at Utah, and Dr. Masahisa Takashi from Aoyama University,
while a visiting professor at Pittsburgh, a satisfactory experiment-
al procedure evolved. It was then refined and standardized at
Pittsburgh by Dr. Eric Betz of the University of Newcastle,
Australia. His carefully accomplished experiments led to the data
reported earlier. 2 It was subsequently analyzed analytically and
numerically by Atkinson, Avila, Betz and Smelser. These results
have just been published 6 which reinforce the visual phenomenologi-
cal description. The suspected high (singular) stresses at the free
surface and the rod tip corners were confirmed and, after conduct-
ing the requisite theoretical energy balance, seemed to show a
general consistency between the theory and experiments. For the
size of rod used in the experhnents, the initial fracture generally
occurred at the center of the flat tip of the rod rather like the
origin of a penny-shape crack between the glass rod end and the
rubber matrix. It then propagated across the flat end and turned
round the corner and crawled up the rod until- -like the rocket
motor behavior- -stopping upon encountering a compressive normal
stress state across the interface.

The next step was a second initiation, this time at the corner
by the surface, after which the fracture surface then propagated
downward to join the previous, stopped crack and complete the de-
bonding. This behavior did not always occur, especially when the
specimen was precracked without load to see if the initial crack
location could be forced toward a different origin.

Figures 6-7 show the applied force and displacement history

for the tip and surface debonds, respectively, as associated with
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 721

the debonding progression along the interface. For our present

purposes it is sufficient to state that a numerical analysis was
carried out from which the change in strain energy with crack
position (strain energy release rate) could be accurately calculated
and correlated with the experimental data. For example the com-
pressive stress field at the lower end of the rod was found, and the
relative energy release rates were such that the order of crack in-
itiation and growth were properly predicted. The maximum dis-
placements for the two cases (Figures 6-7) have been consolidated
on one curve for comparative purposes (Figure 8).

The practical conclusion from this work is that we now have

sufficient confidence in the analytical work, which leaned heavily
upon the use of the characteristic stress singularities, to proceed
with a parametric variation of the length/diameter ratio of the rod,
adjustments in the tip configuration, and the degree of partial em-
bedment in the matrix. Aside from producing a measure of the
matrix stiffening by the rod, we presume it will now be possible to
evaluate the critical loading at which internal deterioration of the
interface affects the load carrying ability of the combination.

CONCLUSION

While there are many potential areas for further work in ad-
hesive fracture from the point of view of continuum mechanics
alone, I would like to reiterate my earlier 1 remarks relating to
two areas where collaboration with chemists and physicists appears
especially profitable.

The first area, referred to earlier in this work, is to estab-

list any consistency, if indeed it is present, in a fundamental
specific adhesive fracture energy (~a) independent of the loading
mode. Three fracture modes have been defined (Figure 2). As
found by Anderson 9 ,4 and reported and discussed by Kinloch, 3 the
amount of new surface area (ASk) created seems to be sensitive to
the loading mode (k = I, II, III) imposed. The analytical treatment
assumes that the total surface energy (r) created is split into
three separate specific fracture energies (~a ya(k» by mode, each
multiplied by the projected new fracture area which of course does
not recognize the existence of different height/depth of hills/valleys.
 722 M.L. WILLIAMS

w
u
a::
o....

0"

_ _ _....!t

0"

A
/
/
~
/ F
/
/

o ~------ ____________ ~ ________________ ~~ ______ ~

K
DISPLACEMENT

Figure 6. Typical force-displacement diagram for tip de-

bond 6 .
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 723

w
o
c:::
a
u.

A~'P

/
/,
/
:
.:/
'/

1'/
'

/
/
./
./
\0'
./
-\..-
Oil

/
/ 1./
~/
./: ,..
,.. ,..

/ ./. ,..
,/ :,..'"
.
/
/,/ ..- ..-'
/' /' /

/
/
/
./ '"'",.. /
/

/ '/,..""
/
/ / ""
/
/.//
'1;'/
a ~ ______________________L -_ _ _ _ _ _ _ _ _ _ ~

E
DISPLACEMENT

Figure 7. Typical force-gisplacement diagram for

surface debond ,
 -..J
I\J
~ZOE c .;:.
¥ ' cr
a
I
60
I Al

50
I x

I
40
I
30
I
I
20
I
10 ---~-- -- --
t ~- Surface debond
~ A , I S; TIp location (tip debondl
0.9 0.8 0. 7 0.6 0. 5 0.4 0.3 0.2 O. I
o
0.0 O. 1 O. Z 0. 3 0.4 0.5 0. 6 0. 7 0. 8 0.9 1.0
X/(S + Rol ; Tip location (.urlace debond) s:
r
Figure 8. Approximate critical displacement for tip and surface debonds ~
deduced from Reference 6, following Reference 17, using r
r
}>
ocr = [ 1/C20E) ] 1/2[ Ya/y a CE=500,ocr=O.Ol) Jl/2 s:
en
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 725

'0
Molecular Modified power law parameters
characte ris tic s Symbol Eg Ee n Tg

Cross-link density
Chain s tiffne s s
Monomeric friction
coefficient
Solubility parameter
Molecular weight
Heterogeneity index
Molecular weight be-
tween entanglements
Degree of crystallinity A
Volume fraction of filler
Volume fraction of
plasticizer
Figure 9. Basic Interaction Matrix arrangement showing
chemical structure and mechanical properties
parameters 18 •
726 M.L. WILLIAMS

Hence the assumed change in energy to create new fracture area is

AP = [<r a (I) + ~ a (II) + ita (III)] A S, whereas there may be physico-
chemical grounds to boldly postulate that there is one unique specif-
ic adhesive (or cohesive) fracture energy ((fa (0» which is perhaps
identifiable with a chemical structure parameter, e. g., bond
strength independent of fracture mode, to give instead
r=
/l. t' a (0) [A SI + A SII + ASIII].

The second area of collaboration is the convenience as well

as significance in developing an Interaction Matrix 18 to provide an
explicit guide to each other's terminology and phenomenology. (Fig-
ure 9). The column elements could be mechanical descriptors such
as rubbery modulus (Ee), glassy modulus (E g ), etc., representing
the relaxation modulus through the modified power law

(5)

The row elements could then be the chemical descriptors such as

cross link density (Ve)' molecular weight (Mw), etc. The object is to
obtain quantitative mechanical-chemical associations for all row~
column intersections in the matrix. The simplest one of course,
from rubber elasticity theory, is Ee = 3Ve k T; the complicated
ones may need to be empirical. In any event a joint effort to de-
velop such a scheme even in terms of only first order correlation
would seem to benefit communication between the fields - -and
eventually technical progress.

ACKNOWLEDGEMENT

Major portions of this work were completed under a grant

from the USAF Office of Aerospace Research whose support is
gratefully acknowledged. The author also is indebted to M. C.
Williams, Assistant Editor of the International Journal of Fracture,
for providing the selected rod pull-out bibliography.

REFERENCES

1. M. L. Williams, J. Adhesion, i, 307 (1972).

2. M. L. Williams and G. P. Anderson, in "Fracture 1977," 4th
Inti. Congr. Fract., !, 643 (1977), University of Waterloo
Press, 1978.
3. A. J. Kinloch, J. Matl. Sci., Part I, .!i, 2141 (1980); Part II
12, 6 17 (1982).
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 727

4. G. P. Anderson, S. J. Bennett, and K. L. DeVries, "Analysis

and Testing of Adhesive Bonds," AcadeITlic Press, New York,
1977 •
5. B. M. Malyshev and R. L. Salganik, Int. J. Fract. Mech., .!.,
114 (1965).
6. C. Atkinson, J. Avila, E. Betz, and R. E. Smelser, J. Mech.
Phys. Solids, lQ., 97 (1982).
7. M. L. WilliaITls, R. F. Landel, and J. D. Ferry, J. Am.
CheITl. Soc., 77, 3701 (1955).
8. M. L. Williams, J. Appl. Polym. Sci., .!.!, 1121 (1970).
9. G. P. Anderson, "Applied Adhesive Fracture Mechanics, "
Ph. D. Dissertation, Univ. of Utah, Salt Lake City, 1973.
10. M. L. WilliaITls, J. Appl. Mech., .!.2., 526 (1952); 20, 590
(1953).
11. A. R. Zak, J. Appl. Mech., 2.!., 150 (1964).
12. M. L. Williams, Int. J. Fract. Mech., .!.' 292 (1965).
13. M. L. WilliaITls, Bull. SeisITloi. Soc. Am., 49, 199 (1959).
14. F. Erdogan, J. Appl. Mech., E, 403 (1965).
15. V. L. Hein and F. Erdogan, Int. J. Fract., 1, 317 (1971).
16. J. J. Benbow, Proc. Phys. Soc., B75, 697 (1960).
17. M. L. WilliaITls, in "Proc. 4th ICRPG Working Group, " San
Diego, Appl. Phys. Lab. ,Johns Hopkins Univ. ,NoveITlber 1965;
abr. in Proc. 5th U. S. Nat. Congr. Appl. Mech., 460 (1966).
18. M. L. WilliaITls and F. N. Kelley, Rubber Chern. Technol.,
42,1175 (1969).

SELECTED BIBLIOGRAPHY - ROD PULL OUT PROBLEM

Ad aITl s , D. F., see Wolrath, D. E.

Allen, G., see WilliaITls, T.
Allred, R. E., and 1973 J. Mater. Sci. 8, 245
Schuster, D. M.
Archangelssha, 1. N., and 1976 J. Mater. Sci. 11, 356
Mileiko, S. T.
Arridge, R. G. G., see
Takaku, A.
ASTM 1971 ASTM, D2138-67, Book
ASTM Stand. Part 28, 913
ASTM 1971 ASTM, D2630-70, Book
ASTM Stand. Part 28, 1009
Atkins, A. G. and Mai, Y. W. 1976 J. Mater. Sci. 11., 2297
Atkins, A. G., see
Marston, T. U.
728 M.L. WILLIAMS

Atkinson, C. 1972 Int. J. Eng. Sci. 10, 45

Aucouturier, M., see
Rebout, D.
Aveston, J., and 1976 J. Mater. Sci. !..!:' 1877
Si1lwood, J. M.
Avila, J. 1980 M. S. Thesis. Analytical and
numerical study of the fiber
pull-out problem. Univer-
sity of Pittsburgh, Pittsburgh,
PA
Bader, M. G., Bailey, J. E., 1973 J. Phys. D §.' 572
and Bell, 1.
Bailey, J. E., see Bader, M. G.
Barry, P. W. 1978 J. Mater. Sci. 13, 2177
Batson, G. 1976 Mat. Sci. Eng. ~, 53
Beaumont, P. W. R., and 1972 J. Mater. Sci. 7.., 1265
Harris, B.
Beaumont, P. W. R., and 1977 J. Mater. Sci. 12, 1853
Plimpton, B.
Beaumont, P. W. R., see
Kirk, J. N.
Beaumont, P. W. R., see
Harris, B.
Becker, E., and Brisbane, J. 1965 Report No. S76. Application
of the finite element method
to stress and analysis of
solid propellant rocket grain.
Rohm & Haas Co., Huntsville,
AL
Bel'chenko, G. L., and 1978 Met. Sci. & Heat Treat. 20,
Gubenko, S. L. 551
Bell, I., see Bader, M. G.
Betz, E. 1978 Report No. SETEC-ME 79-
34. Experimental studies of
the fiber pull-out problem.
University of Pittsburgh,
Pittsburgh, PA
Blumentritt, B. F., Vu, B. T. , 1974 Polymer Eng. Sci. ~, 633
and Cooper, S. L.
Blumentritt, B. F., Vu, B. T. , 1975 Composites, §.' p. 105
and Cooper, S. S.
Blumentritt, B. F., see
Eagles, D. B.
Bowden, P. B. 1970 J. Mater. Sci. ~ 517
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 729

Bowling, J., and 1979 J. Mater. Sci. 14, 431

Groves, G. W.
Bowling, J., and 1979 J. Mater. Sci. 14, 443
Groves, G. W.
Brisbane, J., see Becker, E.
Broutman, L. J., see
Mallick, P. K.
Broutman, L. J., see
Gaggar, S.
Brown, C. B. 1966 J. Franklin Inst. 282, 271
Capiati, N. J., and 1975 J. Mater. Sci. -.!Q, 671
Porte r, R. S.
Compton, J., see
Lessig, E. T.
Cooper, G.A., and 1977 Advances in Research on the
Piggott, M. R. Strength and Fracture of Mate-
rials,.!., 557, Pergamon Press
Cooper, S. L., see
Eagles, D. B.
Cooper, S. L., see
Blumentritt, B. F.
Cottrell, A. H. 1964 Proc. Roy. Soc. A282, 2
de Ferran, E. M., see
Harris, B.
Eagles, D. B., Blumentritt, 1976 J. Appl. Polymer Sci. 20,
B. F., and Cooper, S. L. 435
Ebert, L. J., see Kim, H. C.
Ellis, J. H., see Loveles s, H. S.
Erdogan, F., and 1974 Int. J. Solids Struct. 10, 785
Pacella, A. H.
Felbeck, D. K., see
Marston, T. U.
Field, J. E., see Gorham, D. A.
Fielding-Russell, G. S., see
Nicholson, D. W.
Fielding-Russell, G. S. , 1979 ASTM STP 694, p. 153
Nicholson, D. W., and
Livingston, D. I.
Finello, D., see Marcus, H. L.
Gagger, S., and 1976 Flaw Growth and Fracture,
Broutman, L. J. ASTM, 10th Nat. Symp. on
Fracture Mechanics, p. 310
Gent, A. N., see
Nicholson, D. W.
730 M.L. WILLIAMS

Gershon, B., and Marom, G. 1975 J. Mater. Sci. .!.Q., 1549

Ghesquiere, A. 1976 J. Appl. Polymer Sci. 20,
891
Goodier, J.N.,and Hsu, C.S. 1956 J. Appl. Mech. 76, 271
Gorham, D. A., and 1976 J. Phys. D 9, 1529
Field, J. E.
Groves, G. W., see
Bowling, J.
Groves, G.W., see
Morton, J.
Gubenko, S. L., see
Bel'chenko, G. L.
Hadjis, N., and Piggott, M. R. 1977 J. Mater. Sci. 12, 358
Harris, B., see
Beaumont, P. W. R.
Harris, B., Beaumont, 1971 J. Mater. Sci. 6, 238
P. W. R., and
de Ferran, E. M.
Harris, B., Morley, J., and 1975 J. Mater. Sci. ~ 2050
Phillips, D. C.
Harris, B., see
McGuire, M. A.
Hastings, G. W. 1978 Composites..2, 193
Henriksen, M., and 1979 SAMPE Qtly • .!Q, 15
Thornton, H. R.
Hoover, W. R. 1977 J. Compos. Mater. 11, 17
Hsu, C. S., see Goodier, J. N.
Hull, D., see Jones, M. L. C.
Huntsberger, J. R. 1978 Adhesives Age f.!., 23
Ishikawa, T., Koyama, K., 1977 J. Compos. Mater. !..b 332
and Kobayashi, S.
Jones, M. L. C., and Hull, D. 1979 J. Mater. Sci. 14, 165
Kanninen, M., see Rybicki, E.
Kaufman, M. S., see
Williams, T.
Keer, L. M., see Luk, V. K.
Kelly, A., and Zweben, C. 1976 J. Mater. Sci. g, 582
Kendall, K. 1975 J. Mater. Sci. .!Q., 1011
Kim, H. C., and Ebert, L. J. 1978 J. Compos. Mater. g, 139
Kirk, J. N., Monro, M., and 1978 J. Mater. Sci. !l, 2197
Beaumont, P. W. R.
Kobayashi, S., see
Ishikawa, T.
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 731

Koyama, K., see Ishikawa, T.

Lawrence, P. 1972 J. Mater. Sci. 7, 1
Lessign, E. T., and 1946 Rubber Chem. Technol. !.2,
Compton, J. 223
Livingston, D. I., see
Nicholson, D. W.
Livingston, D. I., see
Fielding-Russell, G. S.
Loveless, H. S., and 1977 J. Test. Eval. 5, 369
Ellis, J. H.
Luk, V. K., and Keer, L. M. 1979 Int. J. Solids Struct. i. 587
Mai, Y. W., see Atkins, A. G.
Majumdar, A. J. 1974 Cement Concrete Res. i, 247
Majurn.dar, A. J., see
Stucke, M. S.
Mallick, P. K., and 1974 J. Mater. Sci. 2, 1420
Broutman, L. J.
Marcus, H. L., and 1978 Analytic Methods for Study-
Finello, D. ing Fiber /Matrix Interface.
Texas University at Austin,
Mat. Sci. Lab., Cont.
N00014-78-C-0094
Marom, G., see Gershon, B.
Marston, T. U., Atkins, ./1.. G., 1974 J. Mater. Sci. :l..! 447
and Felbeck, D. K.
Maslov, B. V. 1978 Sov. Appl. Mechs • .!f., 95
McGuire, M. A., and 1974 J. Phys. D 7, 1788
Harris, B.
Mileiko, S. T., see
Archangelssha, I. N.
Millman, R. S., and 1975 J. Phys. D ~ 1065
Morley, J. G.
Millman, R. S., and 1976 Mater. Sci. Eng. ll, 1
Morley, J. G.
Mindlin, R. D. 1936 Physics 7...., 195
Miyase, A., and 1977 J. Compos. Mater • .!.!., 33
Piekarski, K.
Morley, J., see Harris, B.
Morley, J. G. 1976 Physics Reports 27, 245
Morley, J. G., see
Millman, R. S.
Morton, J., and Groves, G. W. 1974 J. Mater. Sci. 9, 1436
Morton, J., and Groves, G. W. 1975 J. Mater. Sci . .!.£, 170
732 M.L. WILLIAMS

Morton, J., and Groves, G. W. 1976 J. Mater. Sci. 11, 617

Muki, R., and Sternberg, E. 1969 Int. J. Solids Stroct. 5, 587
Muki, R., and Sternberg, E. 1970 Int. J. Solids Struct. f, 69
Munro, M., see Kirk, J. N.
Nicholson, D. W., Livingston, 1979 Tire Sci. Technol. 2,., 114
D. I., and Fielding-
Russell, G. S.
Nicholson, D. W., Livingston, 1978 Tire Sci. 'Technol. 2,., 71
D. I., Fielding-Russell,
G. S., and Gent, A. N.
O'Brien, T. K., and 1977 J. Test. Eval. 2., 384
Reifsnider, K. L.
Pacella, A. H., see
Erdogan, F.
Phillips, D. C., see Harris, B.
Phillips, D. C. 1974 J. Mater. Sci. 2..- 1847
Piekarski, K., see
Miyase, A.
Piggott, M. R., see
Cooper, G. A.
Piggott, M. R. 1970 J. Mater. Sci. i. 669
Pig gott, M. R. 1974 J. Mater. Sci. 2..- 494
Piggott, M. R., see Hadjis, N.
Pinchin, D. J. 1976 J. Mater. Sci. g, 1578
Pinchin, D. J., and 1974 J. Mater. Sci. 2., 300
Woodhams, R. T.
Plimpton, B., see
Beaumont, P. W. R.
Porter, R. S., see
Capiati, N. J.
Rebout, D., Stohr, J. F., 1978 J. Mater. Sci. ~, 2333
and Aucouturier, M.
Reifsnider, K. L., see
0' Brien, T. K.
Rybicki, E., and 1977 Hybrid and Select Metal
Kanninen, M. Matrix Composites. AIAA
Journal, 53
Schuster, D. M., see
Allred, R. E.
Short, D., see
Summerscales, J.
Sillwood, J. M., see
Aveston, J.
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 733

Skolnik, L. 1974 Rubber Chem. Technol.

47, 434
Sternberg, E., see Muki, R.
Stohr, J. F., see Rebout, D.
Stucke, M. S., and 1976 J. Mater. Sci. !l. 1019
Majumdar, A. J.
Summerscales, J., and 1978 Composites 2..- 157
Short, D.
Takaku, A., and 1973 J. Phys. D 6, 2038
Arridge, R. G. C.
Thornton, H. R., see
Henriksen, M.
Tuler, F. R., see
Wagner, H. E.
Vu, B. T., see
Blumentritt, B. F.
Wagner, H. D., and 1979 J. Mater. Sci. 14, 500
Tuler, F. R.
Weetsman, J. 1977 J. Compos. Mater. 1J.., 378
Williams, T., Allen, G., 1973 J. Mater. Sci. 8, 1765
and Kaufman, M. S.
Wolrath, D. E., and 1976 J. Compos. Mater. .!Q., 44
Adams, D. F.
Woodhams, R. T., see
Pinchin, D. J.
Yamaki, J. 1976 J. Phys. D 2..,. 115
Zak, A. R. 1964 J. Appl. Mech. ~ 150
Zweben, C., see Kelly, A.
734 M.L. WILLIAMS

APPENDIX A

The Stress Distribution along a Bi-material Interface

of Finite Length

When the character of the stress singularity near the crack

tip in a bi-material body was determined, 13 the region over which
the major oscillations extended was indeterminate because there
was no need for a characteristic length parameter in the analysis.
This part of the singularity behavior was studied by Malyshev and
Salganik 5 and Erdogan 14 who showed that the violent part of the
oscillations was confined to a very small region near the crack tip.
This feature, and the co-existence of shear and normal stresses
along the interface, can be illustrated by using a simple example.

The Muskhelishvili complex function approach was chosen by

Anderson et a1 4 to illustrate the stresses for the thin plate geometry
subjected to a combined shear (P) and normal load (Q) as shown in
the insert.

( p
,
y P2 , P2

I s+

L' L L'
-a a x

~ )
-
I

S
PI' ~'t

•
p

Two semi-infinite crack geometry and loads analyzed.

CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 735

The parameters which reflect the different material properties of

the upper (S+) and lower (S-) regions are defined as

ot = (PI/oUl)
+ (1/.l~2)
(P2~2) + (1~1)

3 = (1 /21l') In cL

Pk = 3 - 4~k : plane strain

=3 --Uk : plane stress

1 +Vk

in which..«lk is the shear modulus in the material (k = 1,2) and.,)k

is the appropriate Poisson's ratio.

Without elaborating upon the entire solution, 4 the shear

stress (txy) and normal stress (O"y) on the bond line near the crack
tip can be found, in the vicinity (y = 0, x = a - of the bond line, 3)
to be

C1"'y(f, 0) = 2T(V2af
-1 Y!!'
(1 + d.. ) exp (11' ~) ~ ~ rr ~ Eo ~
LQ cos C;,ln 2a + P sln(~ln Z;~(A-l)
.
(f, 0) = (1 +ct- 1 )exp (1r'~)~ cos(jlnt)-Q sin(Sln ~~ (A-2)
i'xy 21r~2at
For a Mode I loading, i. e., P = 0,

f.
= - tan (3 In 2a ) (A-3)
()" Y I

For a Mode II loading, i. e., Q = 0,

~
tTv
= tan (~ln-)
E (A-4)
i . ;, 2a
xy II

The first of these, (A- 3), represents the contribution of shear

stress for Mode I loading, which in the case of identical materials
(CI{.= 1,.3 = 0) would be zero. Similarly, the second, (A-4), gives
the normal to shear stress ratio for Mode II. Note that
(-i /(1"') = ((1'- /f ) = tan (~ln :) (A-5)
xy y I (P = 0) y xy rr(Q = 0) a
736 M.L. WILLIAMS

A maximum of this function, (A- 5), occurs at

1
Z1r In olIn I Z a ::: -+ 1T Iz

at which value the "negligible" component becomes infinitely large!

The question arises: For what material properties or location can
this situation occur? For example, if one of the plane strain media
has an infinite shear modulus, say..l.(Z-+oo, then~ ::: 3 - 4V1 , or PI : :

Now if the Poisson's ratio of the material bonded to it is very near

ly incompressible, saY,)l ::: 0.49, then
E rrZ 11'Z
In - ::: - ~ - -- ~ - Z50
Za In 1.04 0.04
which would be predicted atEIZa.-Jexp (- Z50) which is too small a
distance for that magnitude to hold much significance in continuum
mechanics. Or, alternatively, selecting a small distance from the
crack tip, say f la = O. 1,

tan[z~ln(3 -41\) In (EIZa~ = tan[O~~ (- 3. OO~ -0. OZ

and the shear stress is a reasonably small proportion of the nor-
mal stress that close to the crack. (Note that if the ITlaterial is
ideally incompressible, then -VI = liZ and the tangent function is
identically zero. )

The important point to note is that a simple calculation of

(A-5) applicable to specific materials of immediate interest pro-
vides a guideline as to the relative importance of the shear-normal
interaction in the vicinity of a crack tip along the interface.

The following table shows the variation with distance for a

plane strain case in which E 1 /Ez
= 10, -VI = 0.3, V = 0.4. z
cia tan[3ln (EIZa)]

10- 1 O. lZ
10- Z O.ZI
10- 3 0.30
10- 4 0.40
CONTINUUM MECHANICS FACTORS IN ADHESIVE FRACTURE 737

which for this particular case would as a practical matter probably

correspond to a maximum shear stress which was of the order of
15 percent of the normal stress.
FRACTURE ENERGETICS OF ADHESIVE JOINTS

Lieng-Huang Lee

Webster Research Center

Xerox Corporation
Webster, New York 14580

In this review, we relate various fracture ener-

getics theories to the types of polymers commonly used
in the adhesive joints. The energy-balance concept of
Griffith's fracture criterion is emphasized. The same
concept led to the development of various methods of
determining the fracture energies (or the failure ener-
gies) of both rigid and flexible adhesive joints.

739
740 L.-H. LEE

I. INTRODUCTION
Fracture mechanics has been applied successfully! during the
last decade to assessing the strengths of adhesive joints, either
by the energy-balance or by the stress-intensity factor approach.
The former uses fracture energy, Gc , as the fracture criterion;
the latter uses fracture toughness, Kc. Though Gc and KC are re-
lated for linear elastic systems through the Irwin's treatment 2 ,
Gc is still generally preferred in expressing the adhesive strenth.
One of the major difficulties in obtaining Kc directly and accurate-
ly is that significant differences 3 may be observed in the stress
field from those estimated for a homogeneous material outside a
very short distance in front of the crack tip.

Many papers in the literature emphasize the energy-balance

approach derived from the Griffith's criterion;4 however, they are
sometimes contradictory. Thus, one of the purposes of this paper
is to define the application of each approach with respect to the
type of polymer, e.g., brittle, ductile, or viscoelastic. Sub-
sequently, we intend to show how the fracture energetics of adhesive
joints are derived from the cohesive fracture properties of the
three types of polymers.

Fracture energetics of brittle, glassy po1ymers S have been

discussed in terms of effective fracture surface energy, r, fracture
energy, Gc , and fracture toughness, Kc. Fracture energetics of
ductile and viscoelastic polymers have been treated in a separate
paper. 6 For ductile polymers, the critical J-integra1 7 (or ductile
fracture energy), J c ' is a better measure than Gc • For viscoelastic
polymers, T, the tear energy,S is frequently determined in terms of
Gc . In fact, the actual equivalence for Gc should be Tc, the criti-
cal tear energy (or e1astomeric fracture energy). In the recent
generalized theory of fracture mechanics,26 the failure energy T is
claimed to embrace Gc , Tc and J c •
In practice, it is difficult to classify an adhesive joint as
brittle, ductile, or viscoelastic. For technologists, it is
easier to differentiate rigid from flexible joints. Thus, in the
following discussion, we shall treat adhesive joints according to
the latter classification but with the three types of polymers in
mind.

n. FRACTURE ENERGETICS OF RIGID ADHESIVE JOINTS

1. Measurement of Fracture Energy

Early work by ~p1ing, Mostovoy and Patrick 9 involved the

application of !inear ~lastic fracture mechanics (LEFM) for the
determination of crack growth of rigid adhesive joints. Originally,
the double-cantilever beam was used to measure fracture energies,
FRACTURE ENERGETICS OF ADHESIVE JOINTS 741

Glc (Mode I) and Gllc (Mode II). Since then, the tapered double-
cantilever beam has been more widely used. 10 Assuming that an
adhesive behaves linearly, its total fracture energy (Mode I) can
be expressed as:

p2
2B
(ac)
aa p ,
(1)

where p is the wedge load; B (originally b) is the thickness of the

specimen; C is the compliance, and (aCI aa) p can be determined from
a specific specimen geometry.

Bascom's resu1ts 11 on fracture energies of an epoxy adhesive

are shown in Table 1. The calculated values of rand Kc are also
included for comparison. The fracture energy of the bulk is slight-
ly higher than that of the resin in the joint when Mode I is con-
cerned. However, they are nearly equal when the mixed mode of
fracture is taken into account.

2. Intrinsic Fracture Energy

When the crack speed is very low, or the polymer is truly

linear-elastic, the fracture energy, Ge, should approach the in-
trinsic fracture energy, (Gc)o. Under normal conditions, Gc con-
sists chiefly of the work of plastic deformation for each surface,
Wd:

(2 )

However, at the threshold condition, Wd+O, and Gc+(Gc)o' so,

(G) = 2y = Wa . (3 )
cos

In other words, under the limiting conditions, the fracture energy

approaches the thermodynamic work of adhesion, especially for
truly viscoelastic solids. For plastically deforming solids, some
residual plastic work must be done to sever intermolecular bonds
and create surface. This often takes the form of craze when homo-
geneous plastics flow is no longer present. Crazing energies are
much greater than Wa.
Under normal conditions, the adhesive fracture energy, (G ) ad.,
is affected by the adhesive composition, crack speed, temperat~re,
and environmental variations. One of the major factors influencing
(Gc)ad is reinforcement with elastomers.

3. Toughening Mechanism

Reinforcement of a brittle resin (or adhesive) with an elas-

tomer greatly increases the fracture energy of the adhesive. This
 ......
I\.)
"""

Table I. Effective Fracture Surface Energy, Fracture Energy, and Fracture Toughness of
Epoxy Polymer and Adhesive*.
Modulus r (Calc) Gc Kc (Calc)
EpoxX Resin
(HHPA-OGEBA) (10 3MN/M 2 ) (1010dyne/cm 2 ) (J/M2) (10Serg/cm 2 ) (J/M2) (10Serg/cm 2 ) (108N/M 3,k) (10 10 dyne/cm ¥'2)

Bulk 3.86 3.86 68.0 0.68 136.0 1. 36 5.3 5.3

Adhesive 3.86 3.86 58.0 0.58 116.0 1.16 4.5 4.5

Mode I

Mixed Modes 3.86 3.86 70.0 0.70 140.0 1.40 5.4 ** 5.4
(I and II)

* Ref: Modulus and fracture energy data were obtained from the paper by W.O. Bascom,
C.O. Timmons and R.L. Jones, J. Mat. Sci., 10, 1037 (1975).

~ The calculation for KI,IIC was assumed to be that for KIC .

r
±
r
m
m
FRACTURE ENERGETICS OF ADHESIVE JOINTS 743

THERMOSETS THERMOPLASTICS
10·
~

N
E
.....u
.,
Q
..
....0>
PABST - EPOXY FILM (CYANAMID FM·73)
NITRILE RUBBER REINFORCED POLYIMIDE
It: (PLASTILOCK 6551) POLYSlJLFONE FILM (UDEL P11OO)
0
NITRILE RUBBER REINFORCED POLYIMIDE
N
::E lOa FILM (PLASTILOCK 650)
..... POLYIMIDE (HUGHES HR602) FILM
.."

I.IJ
- POLYIMIDE (DU PONT NR 06S6X) FILM
POLYAMIDE·IMIDE FILM (TOR LON 4000 T)
It: {NITRILE RUBBER REINFORCEDPOLYIMIDE
....
::l
«
POL VlMIDE (FM.34·f8) FILM
C·22·J>HTHALOCVANINE'" POLYSlJLFONE MELT (UDEL Pl100)
a:.
I.IJ
a.. CYANAMID FM·300K (FILM)
::E
....
I.IJ
10 2
C·10 J>HTHALOCVANINE
{POLYIMIOE.HEXCEL HX976
::E EPOXY RESIN (TGMDA·DDS)
0
0 C·36 J>HTHALOCYANINE
It:
....
«
C·6 PHTHALOCYANINE

.x;)
...u
10'
DATA FROM A PAPER BY R.Y. TING AND R.L. COTTINGTON, ADHESIVES AGE,~, JUNE, 1981.

Figure 1. Fracture energies of thermosets and ther~oplastics.

(Data taken from a paper by R.Y. Ting and R.L. Cottington,
Adhesives Age, 24, 35, June, 1981).
744 L.-H. LEE

is illustrated by Bascom's results on the reinforcement of an epoxy

resin with a CTBN nitrile rubber (Table II).

Recently, Ting and Cottington 13 published a series of Gc values

for high performance adhesives (Fig. 1). In Fig. 1, we note that
the reinforcement of polyimide with an elastomer also significantly
increases the fracture energy.

Bascom, et al., theorize 11 that the mechanism of reinforcement

is the increase of the plastic zone size at the crack tip with the
presence of the rubber particles. Since most polymers are more or
less viscoelastic, a quantitative analysis of the plastic deforma-
tion zone can involve the time-temperature dependent viscoelastic
properties. The results, if attainable, would be too complex for
illustration. Thus, an elastic-plastic model has been used with
the assumption that within the deformation zone there is a wedge
of material at yield stress, °, but at the failure strain, Ef.
Surrounding this wedge, the material is at the yield stress, 0y'
and strain, E outside the elastic-plastic boundary which envelopes
a volume withY a diameter of 2ry (Fig. 2).

Figure 2. Crack opening displacement model of an elastic-plastic

material. W.D. Bascom et al., J. Appl. Polym. Sci., 19,
2545 (1975). -

From Griffith's equation, Bascom 11 derived the relation be-

tween GIc and (ry)c:

(4)
 "'T1
:D
»
(')
-i
C
:D
m
m
Z
m
Table II. Effective Fracture Surface Energies and Fracture Energies of Elastomer- :D
G')
m
modified Epoxy Resins. -i
n
en
r (Calc) ~
Adhesive Resin Cleavage (I) Cleavage +Shear(I,II) Cleavage (I) Cleavage +Shear('r,II) o"'T1
(J/M2) (10 Serg/cm') (J/M2) (10 Serg/cm 2 ) (J/M2) (10 'erg/cm 2 ) (J/M2) (10 5erg/cm2) »
o
Unmodified Epoxy I
58.0 0.58 68.0 0.68 116.0 1.16 136.0 1.36 m
(HHPA-DGEBA) en
<:
m
CTBN-Epoxy L
(Piperidine-DGEBA)
o
10% Elastomer 1750 Z
17.5 55.0 0.55 3500 35.0 110.0 1.1 -i
30% Elastomer 1100 11.0 55.0 0.55 2200 22.0 110.0 1.1
en

Commercial
Elastomer-Epoxy 1150 11.5 435 4.35 2300 23.0 870 8.7
Nylon-Epoxy 3050 30.5 375 3.75 6100 61.0 750 7.5

Ref: W.D. Bascom, R.L. Cottington and C.O. Timmons, Appl. Poly. Sym. 32, 165 (1977).

-...J
~
C11
746 L.-H. LEE

Strictly speaking, neither GIc nor (r )c is a true material constant

because both of them are time~tempera~ure dependent. Furthermore,
GIc is also affected by the thickness of the specimen. It should
also be emphasized that Griffith eqUation is based on linear elastic
assumptions which are not valid for a material which yields.
Equations such as (4) and (5), therefore, are approximations for
small amounts of plastic deformation and are invalid if significant
plastic zones are formed.

4. Prediction of Long-Term Stability

Since GIc depends on many variables, it cannot be used as a

single parameter for predicting the long-term stability of an ad-
hesive joint. The plastic zone size ry discussed in the preceding
paragraph should be considered an important parameter for predict-
ing long-term stability.

Under plane strain condition, at fracture, the plastic zone

size is: -

1 [ GIc
E (t)
]
, (5)
6 1r (1-V 2 ) E 2
Y
where E(t) is time-dependent modulus. According to Eq. (5), (ry)c
is a function of GIe/E (t). Thus, (ry)c can only become stable
or constant when GIc/E(t) is constant. Therefore, (r )c .is a
failure criterion for the long-term stability of a st~ctural
adhesive. This criterion has been verified by Williams et
al.,IS,16 for polymethyl methacrylate and polycarbonate. A value
of (ry)c for an epoxy adhesive has been calculated to be 16 ~m.

The latter study is based on a Dugdale line-plastic zone

model. 17 Based on this model, the crack opening displacement (COD)
at fracture, 0c' may be related to GIc and E(t) as follows:

0c = 1 [G IC ] (6)
(1_V2)Ey E(t)

For an epoxy resin, 0c has been calculated to be 4.5 ~m.

5. The Emerging Acceptance of the J-Integral

The importance of the plastic deformation at the crack-tip

could signify the gradual acceptance of using Rice's J-Integra1 7
for determining fracture energies of ductile polymers as well as
adhesives. The J-integra1 7 can be determined experimentally
according to the following equation:
FRACTURE ENERGETICS OF ADHESIVE JOINTS 747

a - W>~B

Figure 3. Tensile specimen for the J-integral

2Up 2Up
J (7)
B{w-a) Bb

where Up is the total potential energy; a is the crack length at

the notch edge; B is thickness of the test piece, w is the width
and b is the ligand length (Fig. 3).

At the crack initiation point, J takes the critical value,

J c , which is called the ductile fracture energy. Its value is:

J = 2 (Up) max
(8)
C Bb

where {Up)max is the maximum potential energy corresponding to the

initiation of crack growth. In general, J c > Gc for ductile poly-
mers because J c more inclusively accounts for the plasticity
component than does Gc • 1S ' A series of J 9 values for polymers has
been determined by Ferguson et al.l~ (F~g. 4). It is likely that
J c values for more adhesives 20 will be made available in the future.

III. FRACTURE ENERGETICS OF FLEXIBLE ADHESIVE JOINTS

Fracture energetics of elastomers have been developed by

Rivlin and Thomas. S Their work was originally derived from
Griffith's fracture criterion. 4 Thus, the fracture mechanics
748 L.-H.LEE

THERMOPLASTICS
10'
POLYVINYLCHLORIDE

POLYCARBONATE

..E
POLYETHYLENE TERPHTHALATE

....
u
10· RUBBER·REINFORCED POLYSTYRENE
~
..g• (MEDIUM IMPAcn

.:
'E
....
E
"
on
0
..
~ 10· POLYMETHYL METHACRYLATE

.,....2
.,u- POLYSTYRENE

10Z~------------------------~

Figure 4. J c values for thermoplastics. (Data taken from a paper

by R.J. Ferguson, G.P. Marshall and J.G. Williams,
Polymer 14, 451 (1973».

of flexible adhesive joints is basically similar to that of rigid

adhesive joints.

In general, the adhesive failure energy, a, described by Gent

and Kinloch 21 is essentially equivalent to the cohesive fracture
energy, Gc or Tc' discussed previously. 5,6
The measurements of the adhesive failure energy can be carried
out in the three forms of Eqs (9), (10), and (11).21

Simple extension (Fig. 5) :

ad
(Gc) =a = kaUb, (9)

Pure shear (Fig. 6) :

(G ) ad
c
= a = hoU b , (10)

and

Peeling (Fig. 6):

(Gc>ad = a = P'/B0' (11)

where Ub is the strain energy density at break (or strain energy

per unit volume); ho is the unstrained height; Bo (originally to>
is the unstrained thickness, and P' is the peel force per unit
FRACTURE ENERGETICS OF ADHESIVE JOINTS 749

MYLAR COATED
STEEL BLOCK

Figure 5. Sketch of simple extension testpiece. (A.N. Gent and

A.J. Kinloch, J. Polym. Sci., A-2, ~, 659 (1971».

width of the adhesive layer. For simple extension, only one side
of the crack plane, i.e., the elastomer side, is considered; for
the cohesive fracture, two sides are used for the summation.

SKETCH OF PURE SHEAR TESTPIECE. SKETCH OF PEEL TESTPIECE.

Figure 6. Sketches of shear and peel test pieces. (A.N. Gent and
A.J. Kinloch, J. Polym. Sci., A-2, ~, 659 (1971».

For flexible joints, the viscoelastic response of the adhesive

is manifested by the dependence of the failure energy upon crack
speed and temperature as indicated by the WLF superposition princi-
ple. 21

For simplicity, the adhesive failure energy, 6, has been as-

sumed 22 ,23 to consist chiefly of the viscoelastic dissipation energy
~ for each surface and the intrinsic failure energy 60 :
750 L.-H. LEE

(12)

Thus, when the crack speed is very low or the temperature is high,
~ + 0, and 8 + 80 ,

The generalized theory of fracture mechanics shows theoretical-

ly that 8 can also be related to 80 by a loss function ~:22-26

8 = 8 <p (a, t, E: ), (13)

o 0

where a is the crack speed; t is the temperature, and Eo is the

strain. For cross-linked elastomers, in the absence of a chemical
reaction at the interface, 80 approaches 2ys' which is the thermo-
dynamic work of adhesion, Wa' Under limiting conditions, the ad-
hesive failure energy (or adhesive strenth) is equal to the thermo-
dynamic work of adhesion. The data obtained by Andrews and King 25
further illustrates the validity of the relationship indicated in
Eq. (13).

IV. CONCLUSIONS

Adhesive fracture energies for both rigid and flexible ad-

hesive joints were discussed separately. For rigid joints, frac-
ture energy Gc is a useful quantity for comparison. However, for
predicting long-term stability, we should also consider the plastic
zone size, r y • For ductile polymers (or adhesives) it is more ap-
propriate to measure the ductile fracture energy, J c '

For flexible adhesive joints, the failure energy 8 can be

measured by three methods: simple tension, pure shear, and peel.
The failure energy can be related to the intrinsic failure energy
80 by a loss function <p. At a very low speed, or at high tempera-
ture, <p + 1, and 8 + 80 , I f there is no chemical bonding at the
interface, the failure energy equals the thermodynamic work of ad-
hesion, Wa' In other words, for an adhesive joint, when the visco-
elastic dissipation vanishes, the strength relies solely on the
thermodynamic work of adhesion. At the limiting stage, the theory
that governs is no longer fracture energetics but surface ener-
getics.
FRACTURE ENERGETICS OF ADHESIVE JOINTS 751

NOMENCLATURE

a 1/2 crack length (in the center), or full crack

length (at the edge), (em)

crack speed (rom/sec)

b ligand length (em)

B thickness (cm)

c compliance (erg-cm 3 )

E elastic modulus (dyn-cm- 2 )

-2
fracture energy (erg-cm )
-2
adhesive fracture energy (erg-em )
-2
J ductile fracture energy (erg-em )
c
-3/2
K fracture toughness (dyn-cm )
c
-2
p applied wedge load (erg-cm )

pi = peel force (dyn)

r radius of the plastic zone (em)

-2
T tear energy (erg-em )
-2
T elastameric fracture energy (erg-em )
c
t temperature (OC)
-2
total potential energy (erg-cm )
-2
strain energy per unit volume (erg-cm )
-2
work of deformation for each surface (erg-em )

w width (em)

surface free energy of solid per unit area (erg-cm- 2 )

-2
r effective fracture surface energ~ (erg-cm )

o = crack opening displacement (em)

752 L.-H. LEE

E strain
o
Poisson's ratio

(J applied stress (dyn-cm- 2 )

viscoelastic dissipative work for each surface

(erg-cm- 2 )

loss function

REFERENCES
1. A.J. Kinloch, J. Materials Sci., 17, 617 (1982).
2. G. R. Irwin, "Fracture" in "Handbuch der Physik," Vol. VI,
pp. 551-590, Springer-Verlag, Berlin, 1958.
3. S.S. Wang, J.F. Mandel and F.J. McGarry, Int. J. Fract., 14,
39 (1978).
4. A.A. Griffith, Phil. Trans. ROy. Soc. A22l, 163 (1920).
5. L.H. Lee, in Physicochemical Aspects of Polymer Surfaces,"p.523
K.L. Mitta1, Editor, Vol. 1, Plenum press, New York (1983).
6. L.H. Lee, in "Adhesive Chemistry - Developments and Trends,"
L.H. Lee, Editor, Vol. 1, Plenum press, New York (1984).
7. J.R. Rice, J. Appl. Mech. 35, 379 (1968).
8. R.S. Rivlin and A.G. ThomaS; J. Po1ym. Sci., 10, 291 (1953).
9. E.J. Ripling, S. Mostovoy and R.L. Patrick, Mat. Res. Stds.,
4, 129 (1964).
10. S. Mostovoy and E.J. Ripling, J. App1. Po1ym. Sci., 10, 1351
(1966).
11. W.D. Bascom, C.O. Timmons and R.L. Jones, J. Materials Sci.,
10, 1037 (1975).
12. ~D. Bascom, R.L. Cottington and C.O. Timmons, Appl. Polym.
Sym., 32, 165 (1977).
13. R.H. Ting and R.L. Cottington, Adhesives Age, 24, 35, June
198!.
14. R.A. Gledhill and A.J. Kinloch, Polym. Eng. Sci., 19, 82
(1979).
15. G.P. Marshall, L.H. Coutts, and J.G. Williams, J. Materials
Sci., 9, 1409 (1974).
16. M. Marvin and J.G. Williams, J. Materials Sci., 10, 1883 (1975).
17. D.S. Dugdale, J. Mech. Phys. Solids, 8, 100 (1960).
18. J .D.G. Sumpter and C.E. Turner, "Cracks and Fracture," ASTM
STP60l, pp. 3-18, American Society for Testing and Materials
(1976) •
19. R.J. Ferguson, G.P. Marshall, and J.G. Williams, Polymer !!,
451 (1973).
20. S.S. Wang, D.L. Hunston, and A.J. Kinloch, to be published.
21. A.N. Gent and A.J. Kinloch, J. Po1ym. Sci., A-2, ~, 659 (1971).
FRACTURE ENERGETICS OF ADHESIVE JOINTS 753

22. E.H. Andrews and A.J. Kinloch, Proc. Roy. Soc. London A332,
385 (1973).
23. E.H. Andrews and A.J. Kinloch, Proc. Roy. Soc. London A332,
401 (1973) •
24. E.H. Andrews, these proceedings, pp. 689-702.
25. E.H. Andrews and N.E. King, J. Materials Sci. , 11, 2004 (1976) •
26. E.H. Andrews, J. Materials Sci. , ~, 887 (1974) •
FRACTURE OF COMPOS1TE-ADHES1VE-COMPOS1TE SYSTEMS

E. J. Ripling, J. S. Santner and P. B. Crosley

Materials Research Laboratory, Inc.

One Science Road
Glenwood, 1L 60425

This program was undertaken to initiate the devel-

opment of a test method for testing adhesive joints in
metal-adhesive-composite systems. The uniform double
cantilever beam (UDCB) and the width tapered beam (WTB)
specimen geometries were evaluated for measuring Mode I
fracture toughness in these systems. The WTB specimen
is the preferred geometry in spite of the fact that it
is more costly to machine than the UDCB specimen.

The use of loading tabs attached to thin sheets of

composites proved to be experimentally unsatisfactory.
Consequently, a new system was developed to load thin
sheets of adherends. This system allows for the direct
measurement of displacement along the load line.

In well made joints separation occurred between

the plies rather than in the adhesive.

755
756 E.J. RIPLING ET AL.

INTRODUCTION

Adhesive joints fracture by the propagation of crack-like de-

fects, such as unbonded areas, that occur in the glue-line. Frac-
tures of this kind have been studied for more than a dozen years
at the Materials Research Laboratory, primarily by Mostovoy and
Ripling. These studies described the behavior of metal-adhesive-
metal (MAM) systems under a number of loading conditions including
contilJuously increasing loads, alternating load and sustained
loads. The behavior of a large number of adhesives was catalogue2
as a function of joint geometry, test temperature, and environment,
and it was demonstrated that laboratory test data could be u~ed to
predict the fracturing characteristics of complex structures. An
ASTM Test Method for measuring the fracture mechanics parameters of
adhesive bonds was developed on the basis of these studies.

Progressive separations of the type found in MAM systems are

also expected in systems in which one or both of the adherends is
a composite. Hence, the study of joint fractures has been extend-
ed to include composite adherends. One major reason for carrying
out this study was to determine whether or not the test methods
developed for metal adherend systems could be extended to include
composites, either as composite-adhesive-composite (CAC) or as
metal-adhesive-composite (MAC) hybrid systems.

Metal and composite adherends are expected to act differently,

and, in general, metal adherend systems are probably easier to
classify than composite adherend systems. In most adhesive tests,
cracking is produced byelastically bending the adherends so that
the elastic modulus of the adherend must be known to measure the
toughness of the adhesive. The elastic modulus is essentially
constant for a single type of homogeneous metals, such as aluminum
alloys. For composites it is not only a function of the reinforce-
ment and matrix species, but also of the lay-up pattern. Second,
the apparent bending modulus of metals is only dependent on the
beam span-to-height ratio when it is small, while in composites the
apparent bending modulus is constant only when the ratio is very
large. Finally, and most important, the through-the-thickness
toughness of composites is much lower than it is for metals. Hence,
for metal adherends, the crack that is initiated in the adhesive is
not able to wander ~nto the adherends; this is not the case for
composite adherends. These differences place restrictions on
specimens designed to measure the fracture mechanics parameters in
composite-adherend systems.
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 757

TEST MATERIALS AND SPECIMEN ANALYSIS

Test Materials
Two composites were used in this study: both were graphite/
epoxy sheets, nominally 0.125 inch (3 mm) thick, having the same
number of c,ross-plies. The ply lay-up, the number of fiber orienta-
tions, and total number of plies are given in Table I. Composite
A was fabricated to have the 0° plies near the sheet center, where
bending stresses, are at a minimum, while for composite B the 0°
plies were near the surface, where bending stresses are greatest.
The two different lay-up patterns were selected so that the two
composites would have different flexure moduli.

Table I. Design of Composite Adherends

Total No. of
Desig- No. of Cross
nation Plies Plies Lay-up Sequence

A 21 4 [+45/-45/0/90/0/-45/+45]3
B 21 4 [0/+45/-45/903/-453 /+45/03 ]
The adhesive AF-163-2k (0.06 oz/ft2) utilized in this program
is manufactured by the Minnesota Mining and Manufacturing Company.
It is a 250F (120C) curing adhesive reported to have excellent water
resistance. This lower temperature curing adhesive was selected
rather than the 350F (175C) curing ones normally used with compos-
ites so that the aluminum adherends could be used without signifi-
cantly altering their yield strength. The adhesive is generically
a modified epoxy.

Specimen Type and Analysis

4
As was the case for the MAM specimens in Method D3433-75 ,
double cantilever beam specimens were selected for this program.
Both the uniform height double cantilever beam (UDCB) specimen and
the width tapered beam (WTB) specimen were used. Preliminary ex-
periments were conducted on the UDCB specimen. Its geometry is
Simpler than the WTB specimen geometry and allows the efficient
utilization of composite material. The WTB specimen, unlike the
UDCB, has the advantage that j is independent of crack length, and
hence its measurement only requires that load be monitored. Un-
fortunately, it is somewhat more expensive than a UDCB specimen,
but its advantage in stress corrosion cracking and fatigue are so
great that this additional expense is justified. Both specimen
types can be produced by "wide-area" bonding.
758 E.J. RIPLING ET AL.

Analysis of DCB Specimens

5-8
Both the -UDCB and WTB specimens have been analyzed earlier •
However. in using composite adherends it is advantageous to select
the adherend thickness so that neither the rotation correction
given in Reference 6 nor the large displacement expression given in
(7) are needed. Hence. the simplified expression for the
crack extension force. ". as given below. is adequate.

For any specimen or structure the compliance, C, and load. p.

are related toft by the expression

/J. = p 2C'/(2b) (2-1)

Where C' = dC/da. the rate of change of compliance with crack length.
and b is the specimen thickness at the crack front position. The
compliance at the loading point is the displacement per unit load.
Le., C = vIp Where V is the load point displacement. The present
concern is the relation between P and V for a crack extending at
constant /1-. The desired relation is obtained by substituting the
quantity P = vIc for P in Equation (2-1). This gives

P = (2bCft/C')'(1/V) (2-2)

For a DCB specimen analyzed as a pair of built-in beams Which de-

form by bending, the compliance change is given by

2a 2 ( 2-3)
C' = EI
Where a is the crack length measured from the load line, E is Young's
modulus, and I is the moment of inertia of the beam cross section at
the crack front position.* For rectangular cross sections
(2-4)

The compliance can be obtained by carrying out. the integration

c=c +1 act da (2-5)

o 0

Where C can be formally identified as the compliance corresponding

to a = 0 0 . The integration will depend on the variation with crack

* This result is derived in Appendix A.

FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 759

length of the beam cross section and of Young's modulus.* The

simplest case is the uniform DeB specimen where ~ and ~ are inde-
pendent of ~. and the integration gives

(2-6)

assuming that C makes a negligible contribution to the compliance.

Substituting Eq8ations (2-4) and (2-6) into Equation (2-2) gives the
load-displacement relationship for the uniform DCB specimen:

p = (2ab/1/3)· (1/V) (2-7)

A second type of specimen used on this program was the width-

tapered-beam (WTB) specimen in which the thickness. b. increases
linearly with the crack length. a; i.e •• b = a/k where k is constant.
Carrying out the integration indication in Equation (2-5) gives

(2-8)

Since beam theory is not exact. measured compliance values may dif-
fer from Equation (2-8). The procedure then is to fit the experi-
mental compliance data to an equation of the form
2
C=B+a.a (2-9)

In Equation (2-9) B can be identified with C and a. is simply a

parameter fitted to experimental compliance aata. If the beam
model were3 precisely applicable. a. would have a value corresponding
to 12k/(Eh ); in fact. it might be somewhat different. The com-
pliance derivative. C'. obtained from Equation (2-9) is

C' = 20. a = 20. kb (2-10)

and substitution of this value into Equation (2-1) gives

jJ. = p2 a.k (2-11)

To the extent that a single value of a. provides a good fit to the

compliance date over a range of crack lengths. the relation between
A and P is independent of crack length in that range.

* In the composite materials tested on this program the

apparent elastic modulus in bending depends on beam
length. This situation is addressed in Appendix B.
760 E.J. RIPLING ET AL.

TEST RESULTS

Three-Point Bend Measurement of Tensile Modulus

for Composites A and B

One difficulty encountered in testing DCB specimens with com-

posite adherends, is that the apparent elastic modulus of compos-
ites loaded in bending can depend on thg adherend's span-to-depth
ratio. The results of a recent survey suggest that a span-to-
depth ratio of at least 30/1 may be necessary to avoid this effect.
As the design and analysis of DCB specimens must take this modulus
variation into account, measurements were made on the materials
used in this program. Three-point bend s~6cimen tests were con-
ducted according to the ASTM Standard D790 • Two series of tests
were run to determine Whether the modulus was affected by specimen

3
10 20 30 40 50 60
SID ="
Figure 1. Apparent modulus as a function of span-to-depth
ratio, u.
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 761

width, b. In the first series, specimens were 0.5 inch (12.7 mm)
wide, and in the second, the specimen width was increased to 1.0
inch (25.4 mm). Figure 1 is a plot of the apparent tensile modulus
measured as a function of the span-to-depth (SID) ratio for the
specimens with two different widths. The modulus is seen to be
independent of width for both composites tested. As expected,
placing the 0° plies near. the surface of composite B significantly
increased its stiffness over composite A, where the 0° plies were
located more uniformly throughout the thickness. In addition, the
apparent modulus reached a plateau with 90 percent of the saturated
value, E , near the SID ratio of 30/1. Composite A reached its 90
percent 8f the plateau value at a smaller SID ratio (20/1).

In order to model these effects in the UDCB and WTB specimen

geometries, the apparent modulus must be expressed as a function of
the crack length, i.e., E = E(a). Several empirical fitting func-
tions having the general shape of Figure 1 were investigated. The
best fit was obtained with:

E=E (l-u/u), u=S/D (3-1)

o 0

which has a correlation coefficient of better than 0.98 for both

composites. In order to relate the crack length, a, in a double
cantilever beam to the span in a three-point bend test, the follow-
ing identifications are made:
S So s S
o
"2 - a, 2 - a o' u = D' and U o (3-2)
D
so that,
u D u a
a = ~ and --2. ___ --2. (3-3)
o 2' u a

Based on the observations for two extremes of composite struc-

ture and the correlations between a three-point bend specimen and a
double cantilever beam specimen, the requirements for initial crack
length were established. A span-to-depth ratio of 30/1 for a
three-point beam is equivalent to a ratio of 15/1 for a cantilever
beam. Since the composites used in the program were 0.125 inch
(3 mm) thick, the initial crack length, a, should be almost two
inches (50 mm) long. The initial blunt-starter flaw was made
shorter than this to allow room for it to jump to form a natural
crack on loading. While data is collected for alh < IS, it is not
considered valid until it extends beyond this region. In design-
ing the test specimens for this program, the starter cracks were
made to equal about one-half the required value of a/h. In this
way, both the specimen size and magnitude of displacements that had
to be measured were reduced to manageable sizes. The critical
value of alh = 15 is marked on the test records (Figures 3, 4, 12
and 13).
762 E.J. RIPLING ET AL.

UDCB Specimen

Specimen shape. Uniform double cantilever beam specimens with

the geometry in Figure 2 were fabricated for testing. Individual
oversized specimen blanks. approximately 7/8 x 4 inch (22 x 100 mm).
were bonded together following the manufacturer's recommended pro-
cedure. with the exception of pressure. This value was decreased
by half to minimize excessive adhesive run-out experienced during
gluing trials with unidirectionally reinforced composites. The
purpose of these trials was to establish the characteristic failure
mode with composite adherends. and determine the feasibility of
measuring adhesive toughness with the UDCB specimen.

Test results of UDCB specimen. Typical load-deflection test

records from UDCB specimens with composite A and composite B adher-
ends are shown in Figures 3 and 4. respectively. Both figures
show non-linearity associated with the initial loading. In addi-
tion. whenever the tests were interrupted to optically measure the
crack length. the load decreased by 7-10 percent. This might be
due to plastic flow at the crack tip. or slow. subcritical crack
extension. No arrest marks or similar features were observed at

w<D
(6 . 3) ----=;-+---l

0.1880
(4 .80)
t:'H!fiIIB. " AD H ES IVE

3.2~0 J
I (826)

~
I I
I I
b I I
I I I
I
-.! 0 1 ~
NOTES <D W· 3 . 000 (76 . 2)
(g) h· O. 12~ (3 . 2)
@ b· O. 7~0 119.0)

Figure 2. UDCB specimen geometry with W/h 24. All dimensions

are inches (mm).
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 763

V, mm
2 3 4 5 6
40

UDCB 150
COMPOS ITE A
30
( SP. NO 9)

{j, TEST INTERRUPTED TO

100 z
Q..-
0.- READ CRACK LENG T H

/:J . I ~
In
(175 N/m) 50

- - -,

50 100 150 200

v, x 10 - 3 i nch

Figure 3. Typical load-deflection curve for similar composite

adherends whose failure mode is 100% adhesive failure, i.e., separ-
ation at the adhesive-composite interface.

v, mm
3 6 7 8

G · 1.0 in (25 . 4 mm ) uoce 500

CO M POS I TE 8
FA IL URE
100 ( SP. NO . 7)
NEAR
COMPOSITE -
IN I TlATI ON OF
_ 50 '1. CO M POSITE 4 00
ADHES I VE COMPOSIH I N TER - PLY
INTER - PLY FA ILURE
eo \ \ INTERFACE FA I LURE

\ 'l"'h . 7. 55 Iblo . 1132 1 N/ml

" ', I
300
D
>:
.:
60 . , 1 60 , . (40. 6",m l 0."

" " 200

40 o · 2 . 20 I n ( 55 . 9 mm)

V1 - · 3 9 2 In
(9.96mm)
t:. 10 0
20 TEST INH R -
RUPTED TO
...
0.111:
-:>
M EASURE CRAC~
LE N GTH
;, .,"'oJ-
If
50 100 150 200 2 50 300 35 0

Figure 4. Experimental load-deflection curve for similar compos-

ite adherends whose failure mode gradually changes.
764 E.J. RIPLING ET AL.

those locations using optical fractography up to SOX magnifica-

tion. Upon continuing the test, the load monotonically in-
creased until the pre-interrupted loads were reached. Further
work is needed to understand how sensitive ~ is to loading rates.

The broken line in Figure 3 is the P-V curve for~ = 1.0 lb/in
(174 N/m) calculated from Equation (2-7). The theoretical curve
beyond the initial non-linearity closely matches the experimental
line, if the interrupted portion of the test is neglected. The
low value of jt is due to the fact that this specimen failed by
adhesive failure, i.e., separation at the adhesive-composite inter-
face.

Following the same procedure for the curve in Fig. 4, which

is a specimen with B composite adherend, such good agreement was
not found. This is due to the fact that the fracture mode changed
after the test was stopped to measure the crack length. Failure
near the adhesive-composite interface was the primary failure mode
until the crack grew to 1.35 inches (34.3 rom). At this point the
test was interrupted to measure the crack length. After this in-
terruption, the failure mode changed to include interply separation
along one edge of the crack front. The test was interrupted again
at a crack length of 1.83 inches (46.5 rom). Approximately half of
the crack front failed by interply separation, and the remainder of
the crack front continued to fail near the composite-adhesive inter-
face when the crack length reached two inches. Since interrupting
the test to measure the crack length is associated with a change in
the fracture mode, this procedure appeared to be a questionable
practice. Figure 5 shows the macro fracture surfaces of the spe-
cimens, which are summarized in Table II.

WTB Specimen

Evaluating the fracture toughness using the UDCB specimen re-

quires the simultaneous measurement of the load and crack length.
As noted in the previous section, interrupting a UDCB specimen test
to measure the crack length may cause transients to occur. This
raises questions of interpreting the test record. The geometry
of the WTB specimen is designed so that only knowledge of the load
is required to measure ft. ..

Loading thin adherend panels. Loading thin (0.125 inch, i.e.,

3.2 rom) adherends with externally attached tabs such as shown in
Figure 2, was not a satisfactory experimental arrangement. Separ-
ately attaching tabs is an expensive, time consuming process, in
which frequent tab failure was experienced. In addition, while
several procedures for assuring pin hole alignment were tried, none
were completely satisfactory. Therefore, a new approach for load-
ing thin adherend specimens, illustrated in Figure 6, was devised
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 765

Figure 5a. Fractographs of UDCB specimens with matching

composite A adherends.

Figure 5b. Fractographs of UDCB specimens with matching

composite B adherends.
766 E.J. RIPLING ET AL.

Table II. UDCB Specimen Test Results

Bond Line
Spec. Adher- Thickness
No. end mils (mm) Failure Mode*
4 B 9(0.23) CF(a/h<24)/BF(00)[on edges for
(a/h<24) ]
6 B 6(0.15) AF(a/h<24)
7 B 4(0.10) CF(a/h<16)/BF(00)[a/h>16]
5 A 5(0.13 ) AF(a/h~16)/BF(00)[a/h>16]
8 A 5(0.13) AF(a/h<24)/BF(00) [7.2<a/h<9.1]
9 A 6(0.15) AF(a/h<24)
* Failure modes in adhesive tests.

Failure
Mode Description

AF Adhesive failure: separation at the adhesive-

adherend interface.
Beam failure: failure of the adherend.
(x = 90°: fracture of the beam arm perpendicular
( to the bond line.
( .
(x = fracture of the beam arm parallel to
the bond line (interply separation
for composite adherends).
CF Cohesive failure in the center of the bond line.

which could be applied to either a UDCB specimen or a WTB specimen.

However, in the present program, this approach was used only on WTB
specimens with the geometry illustrated in Figure 7.

The unique feature of this loading arrangement was the use of

the same member for applying the load and measuring displacement.
In this way no special effort is required to be certain the dis-
placements being measured are those from the load line. The de-
tails of this loading arrangement are shown in Appendix C.
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 767

Figure 6. Loading concept for thin adherend specimens •

...

0 . 2500
(6 3)

- 3500
(SS . 9)

NOTES . CD W· 3 .000 (76.2)

(%I h ·0. 125 (3 . 2)
B • 1.000 (25 . 4)

Figure 7. WTB specimen geometry. All dimensions are inches (mm).

768 E.J. RIPLING ET AL.

To prevent the load rod from coming into contact with the
access hole in the top adherend as the specimen is opened, its nose
was tapered. A simple calculation (see Figure 8) of geometrically
wedging the loading rod in the top adherend for the wors§ case spe-
cimen (WTB specimen with an adherend modulus of 5.2 x 10 psi, i.e.,
35.8 GPa) showed that the projected diameter, d, decreases by 95
percent, While the rotation of the adherends ca8ses an offset, e,
of approximately 0.17 inches (4.4 mm) at a displacement of 1.0 inch
(25 mm). The loading rod geometry shown in Figure C-2 would
physically wedge at this displacement. Wedging effects were not
noticed during this program for displacements less than about 0.5
inches (13 mm). Since the WTB specimen equilibrium load is well
established well before this displacement, the wedging problem was
of little practical importance in the current study. However, the
potential problem can be completely eliminated by replacing the hole
in the top adherend with a slot. The slot geometry would also al-
low easier loading of the specimen into the test fixture.

Compliance measurement of WTB specimens. Aluminum alloy

2024-T73 was used as the adherend for the compliance measurement.
Six specimens were prepared with initial crack lengths ranging

T
V

---d
0---...

t
V

WEDGING· PROJECnON + OFFSET

= (do -<\» + e/2
= do(l-COse)+ 1/2VTAN e, e· TAN-I(~a)
Figure 8. Geometrical wedging of the loading rod during the open-
ing of a cantilever beam specimen.
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 769

from 3/8 inch to 2~ inch (9.5 mm to 57 mm) following the procedure

outlined in Appendix D. Each specimen was loaded, removed from
the test fixture and re-loaded six times. The reported compli-
ance is the average of these trials. A standard deviation of five
percent was observed at each crack length tested. As the WTB spe-
cimen compliance is proportional to the crack length squared [see
Equation (2-8)], the results were plotted on this scale in Figure 9.
A linear least squares regression was used to determine the coeffi-
cients in an equation of the form:
2
C=6+a.a (2-9)

A correlation coefficient of greater than 0.99 statistically con-

firms what is obvious from the plot. It is noteworthy that the
intercept, 6,_!s nearlY_fero, and that the fitted slope
a. = 1.74 x 3 10 (in-lb) is identical to the theoretical value,
a. = 12k/Eh •

WTB test results and fractography. Test results for the WTB
specimen composite adherends are summarized in Table III. Fracto-
graphs of the first four specimens are shown in Figure 10. These

14
WTB COMPLIANCE CURVE

2024 T 73 AI ADHERENDS
12
k-3

B = 1.0 in (25 mm)

.c
..... 10

u
c:
,..,
I 8
~
u~

6
C - (1.74 E-3)02_(0.40 E-3)

CORR. COEF. • 0.9953

4 -

2 3 4 5 6 7 8
0 2 , inches 2

Figure 9. WTB specimen [k = 3, B = 1.0 inch (25 mm)] compliance.

770 E.J. RIPLING ET AL.

Table III. Composite-Adhesive-Composite Test Results for WTB

Specimens.

Outer
Ply Primary
Orien- Spec. E P IJ: Failure
tation No. 106 psi (GPa) lbs, (N) lb/irt (N/m) Mode

45° Al 5.3 (36.6) 11 (49) 0.5 (88) Adhesive

45° A2 5.3 (36.6) 10 (44) 0.4 (70) Adhesive
0° B1 12.3 (84.7) 31 (138) 1.6 (280) Mixed*
0° B2 12.3 (84.7) * Mixed*
45° A3 5.3 (36.6) 20 (89) 1.6 (280) Interply
45° A4 5.3 (36.6) 20 (89) 1.6 (280) Interply
0° B3 12.3 (84.7) 23 (103) 0.9 (158) Interply
0° B4 12.3 (84.7) 29 (129) 1.4 (245) Interply

* Initially adhesive failures transforms into interlaminar ply

failure. Specimen No. B1 is primarily adhesive failure, while
specimen No. B2 has a larger fraction of its surface with inter-
laminar separation. There was no region on the load displace-
ment curve where the load remained constant.

four specimens were made with adherends that were cleaned, but not
roughened. As was the case for the UDCB specimens, specimens made
from composite A showed a greater tendency to separate at the adhe-
sive-composite interface than the B composite.

Subsequently, the composite specimen surfaces were roughened with

400 grit SiC paper to help prevent adhesive failure. Figure II,
which contains the fractographs for this series of tests, shows that
the failure mode changed from adhesive failure to wholly interply
separation. The fracture in all of these specimens initiated be-
tween plies even though an initial unbonded area in the adhesive
layer was made with "Teflon" tape. It must be noted that all these
bonds were thinner than recommended, however. Interestingly, the
load-deflection curves remained remarkably flat, even when the frac-
ture plane changed through the adherend thickness from one interply
layer to another, as illustrated in Figure 12 for specimen A-4.
Clearly, in this example, the plies in question are not oriented
to bear a significant portion of the load in the beam. However,
this same phenomenon persists for composite B, where the interply
separation removes the 0° fiber from one adherend and effectively
adds them to the other (see Figure 13).
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 771

Figure 10. Fractographs of WTB specimens with similar composite

adherends prepared without surface roughening prior to bonding.
Prefix of specimen number is composite type.

Figure 11. Fractographs of WTB specimens with similar composite

adherends prepared with surface roughening prior to bonding.
Prefix of specimen number is composite type.
 -..J
-..J
N

V, mm
2 3 4 5 6 7 8 9 10 II
40

WTB
COMPOSITE A 150
(SP. NO . A4)
30

.a
'" 100
z
0.- 20 0.-

50
10

50 100 150 200 250 300 350 400 450

v, x 10- 3 inches
m
t..
Figure 12. Typical experimental load-deflection curve for similar
:0
composite adherends Whose failure mode is 100 percent interply :;;
beam failure. r
Z
G)
m
-I
»
r
 "TI
::0
»
()
-I
C
mm ::0
V. m
2 3 . 5 6 7 8 o"TI
40 ()
o
~
J 150
)j = l,q Ib / in.
o"en
30 =i
m
:i>
o
I
m
. / h = 15 en
~ 20 r ro~ <
m
n
o
~

10 I- /
/ ! COM;:'~TE .1 50
"o
en
(SP. NO . B4J =i
m
en
-<
en
-I
m
50 100 150 200 250 300 350 ~
en
v, ]I 10- 3 inches

Figure 13. Typical experimental load-deflection curve for similar

composite adherends whose failure mode is 100 percent interply
beam failure.

......
......
w
774 E.J. RIPLING ET AL.

It might also be noted that the "- values fit within the same
scatter band when alh was more than or less than 15. Obviously.
the scatter in toughness overshadowed the variation in modulus.

Separation between the plies near the afthesive was also the
fracturing behavior found by Han and Koutsky for adhesives ustig
a glass fiber reinforced polyester composite. Bascom. et a1.
have pointed out that the interp1y toughness of composites is always
far lower than the bulk toughness of the matrix resin. They sug-
gest that this behavior occurs because the interp1y layer is a
very thin joint which does not allow the formation ofa full defor-
mation zone at the crack tip. If this is the case. the fact that
these joints were thinner than recommended did not contribute to
the fracture morphology.

CONCLUSIONS AND RECOMMENDATIONS

Uniform and width tapered double cantilever beam specimens

were used to measure the fracture toughness of adhesive joints made
with composite adherends. Both types of test specimen were de-
signed to take into account the fact that the bending elastic modu-
lus of composites is a function of span-to-depth ratio. A novel
method for loading thin-adherend DeB specimens was introduced in
which the load-point displacement measuring device was also the
loading member.

For well made joints. the crack appeared to initiate and prop-
agate between the plies near the adhesive joint for monotonically
increasing loads even though a starter flaw was placed in the ad-
hesive. This behavior was also reported eor adherends made of
glass fiber reinforced polyester composite. It has been suggest-
ed that a resin matrix in a composite will always have a lower value
of toughness than its bulk value because the interp1y layer is a
thin joint that does not all~y for the development of a full defor-
mation zone at the crack tip •

Additional Mode I opening load tests will be conducted with

other than monotonically increasing loads to determine whether or
not the interplies always have lower crack propagation resistance
than the adhesive layer under opening mode conditions. Further-
more. joint configurations that do not allow interply separation
will also be examined.
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 775

ACKNOWLEDGMENT

The study reported in this program was conducted by the Materi-

als Research Laboratory. Inc •• of Glenwood. Illinois. and was spon-
sored by the National Aeronautics and Space Administration through
the Lewis Research Center under Contract No. NAS3-21824.

REFERENCES

1a. E. J. Rip1ing. R. T. Corten and S. Mostovoy. J. Adhesion. ~.

107 (1971). (Also published in SAMPE. 2. 259 (1970).
lb. S. Mostovoy. C. R. Bersch and E. J. Rip11ng. J. Adhesion. ~.
125 (1971). (Also published in SAMPE. 2. 273 (1970).
1c. E. J. Rip1ing. S. Mostovoy and C. Bersch~ J. Adhesion. 3, 145,
(1971). (Also published in SAMPE, 2, 287 (1970). -
2. S. Mostovoy and E. J. Rip1ing in "Adhesion Science and Tech-
no10gy," L. R. Lee, Editor, 9B, 513, Plenum Press, New York,
(1975). -
3. T. R. Brussat and S. Mostovoy, "Fracture Mechanics for Struc-
tural Adhesive Bonds, " Final Report for Contract No. F33615-
75-C-5224 (15 June 1975 - 15 July 1977), Technical Report No.
AFML-TR-77-163, Air Force Materials Laboratory, Wright-Patter-
son AFB, OR 45433.
4. "Standard Recommended Practice for Fracture Strength in Cleav-
age of Adhesives in Bonded Joints," ASTM Standard D3433-75,
1979 Annual Book of ASTM Standards, Part 22: Wood; Adhesives,
American Society for Testing and Materials, Philadelphia, PA.
5. S. Mostovoy, P. B. Crosley and E. J. Rip1ing, J. Materials,
2, 661, (1967).
6. T. R. Brussat, S. T. Chiu and S. Mostovoy, "Fracture Mechanics
for Structural Adhesive Bonds," Tenth Interim Report, Contract
No. F33615-75-C-S224, Air Force Materials Laboratory, Wright-
Patterson AFB, OR 45433.
7. D. F. Devi~t, R. A. Scbapery and W. L. Bradley, J. Composite
Materials, 14, 270 (1980).
8. K. S. Ran and J. Koutsky, J. Composite Materials, 11, 371,
(1980) •
9. C. Zweben, W. S. Smith and M. W. Wardle, "Composite Materials:
Testing and Design (Fifth Conference)," ASTM STP 674, S. W.
Tsai, editor, 228, American Society for Testing and Materials,
Philadelphia, PA, (1979).
10. "Standard Test Method for Flexural Properties of Plastics and
Electrical Insulating Material, " ASTM Standard D790-71 (Re-
approved 1978), 1979 Annual Book of ASTM Standards. Part 35:
Plastics - General Test Methods, Nomenclature. American Soci-
ety for Testing and Materials, Philadelphia, PA.
11. W. D. Bascom, J. L. Bitner, R. J. Moulton and A. R. Siebert,
Composites, !l,9 (1980).
776 E.J. RIPLING ET AL.

APPENDIX A

Demonstration that jz Depends on Beam

Cross Section at the Crack Tip

Double cantilever beam (DCB) specimens can be analyzed as a

pair of opposed built-in beams. Consider the bending of the beam
shown in Figure A-l. The curvature is given by

y" =~
EI
= Px
EI
(A-l)
the slope by.

'=/aPxdx (A-2)
y x EI
and the deflection by.

=1 al a~dx dx (A-3)
y x x EI
where.

M = moment (= Px)
E = Young's modulus
I = moment of inertia
A condition of zero slope at x .. a has been incorporated in the
limits of integration. In general. EI may be considered to be a
function of x.

The compliance of a DCB specimen composed of two such beams is

taken as

(A-4)
where
Yo =y evaluated at x .. 0

Let F be a func tion such tha t

F(a) - F(x) = 1 xa ~ dx (A-5)

or.
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 777

dF 2x (A-6)
dx = EI
The compliance can then be written
C = aF(a) -! oa F(x) dx (A-7)

and, upon differentiation,

dC
da = aF' (a) (A-8)
or

(A-9)

The noteworthy feature of this result is that, within the limits of

beam theory, dcl da, which is the crucial factor in determining A,
depends only on the beam dimensions (specifically, on EI) evaluated
at the crack front.
A more complete description takes into account shear deforma-
tion of the beam. In this case

dC = 2a 2 + 6 (1 + v) (A-I0)
da EI EA

where the second term represents the shear contribution to dC/da,

and where

v = Poisson's ratio
A = beam cross-section area
Again, all the quantities are evaluated at the 2crack front position.
For a rectangular beam of height ~, A .. 12 I/h and
2 2
dC .. 2a [1 + (1 + v) (~) ] (A-ll)
da EI 4 a
Thus, so long as hla < 1/8 and V ... 1/3, the contribution of a shear
is less than 1/2 percent.
778 E.J. RIPLING ET AL.

(a)

(b)
L x --------I

Figure AI. Nomenclature used to describe the DCB specimen, (a),

as a pair of opposed built-in beams (b).

APPENDIX B

Effect of Crack Length Dependent Elastic

Modulus on DCB Specimen Behavior

For a DCB specimen, the load-displacement relation for a crack

propagating at constant It is given by Equation (2-2) as
PV = (2b,b)· (C/ C ' ) (B-1)

For a specimen in which the apparent tensile modulus, E, depends on

crack length, the expression for C is unchanged as long as E is in-
terpreted as the apparent modulus. Because the apparent modulus
depends on a, the expression for C' (derivative of C with respect
to~) is different.

UDCB Specimen. For the UDCB specimen

8a3 (B-2)
C = Ebh3

as given by Equation (2-6), but evaluation of C' gives

FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 779

2
e' = de = 24a [1 _ aE'] (B-3)
da EBh3 3E
Where E is the apparent modulus, and E' = dE/da. Substituting for
e and e' in Equation (B-1) gives

PV = (2b/l)' ( al 3) • cj> (B-4)

Where
cj> = [1 - aE'/(3E)]-1 (B-S)

The factor cj> can be recognized as a correction to the expression,

Equation (2-7) obtained for constant E. Substituting the experi-
mentally fitted modulus expression, Equation (3-1) into Equation
(B-S) gives

3 (a - a )
o
cj> = 3a - 4a (B-6)
o
WTB Specimen. For the WTB specimen the experimental compli-
ance was adequately represented by

12 ka 2
C = --::-- (B-7)
Eh3

Differentiation with respect to ~ gives

e' = 24ka (1 _ aE') (B-8)

Eh3 2E

The relationship between P and ~ as given in Equation (2-1) is

(B-9)
Substituting Equation (B-8) for e' gives
2 Eh 3 aE' -1
P = (2b,h)· 24ka • (1 - 2E ) (B-10)

Noting that k = alb gives

p2 =PEh3 (1 _ aE,)-l (B-ll)

12k2 2E

the dependence of P on a arises, of course, from the dependence of

E on a. Using the E-a-relationship given in the test, Equation
(3-1)-gives
780 E.J. RIPLING ET AL.

2
2 (a - a )
a (2a - ~a o ) } (B-I2)

The term in curly brackets represents the correction due to the

variable Young's modulus.

APPENDIX C

Press Design for Loading Double Cantilever

Beam Specimens

Figure A-I illustrates the press designed for use in a conven-

tional machanical test machine. The purpose of the fixture is to
provide a reliable means of applying a load to the bottom adherend
of a double cantilever beam concentric with the dowel on Which the
top adherend rests. The loading rod is machined to a hemisphere
of the same diameter of the dowel pins in the fixture. The rod is
also tapered from the hemisphere, as shown in Figure C-2, to .prevent
it from rubbing the circular opening in the top adherend, Which ro-
tates as the specimen is opened. Thus, both halves of the double
cantilever beam specimen are free to rotate with this design.
Figure C-3 shows the fixture with an LVDT coil attached to the top.
The magnetic core is an integral part of the load rod (Figure C-2)
and provides a direct measure of the load line displacement.
Figure C-3(b) illustrates that the loading rod shoulder drops to
the top adherend after the specimen fracture. The LVDT was chosen
to have a one inch (25 mm) range based gn the calculated WTB deflec-
tions for aluminum adherends (10.5 x 10 psi, i.e., 72 GPa modulus).
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 781

I
0.8150 I
_J
(FOR LVDT) I

DETAIL "A"

JETAIL "AN
---r-T---
---I 1--- DOWEL
---tO~-- CHANNEL
--, 1---

I
Cf.
Figure Cl. Point loading adhesive testing fixtur~
782 E.J. RIPLING ET AL.

118" R

~---1/4' D
STAI N LESS STEEL
THREADED END

1.000

~-- LVDT
MAGNETIC CORE
(SUPPLIED)

~--1/4' D

STAI N LESS STEEL

THREADED END

Figure C2. Loading rod for point loading adhesive double canti-
lever beam specimen.
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 783

(a) (b)

Figure C3. Point loading adhesive testing fixture with LVDT at-
tached to top (a) side view (b) front view.

APPENDIX D

Procedure for Preparing Width Tapered Beam (WTB)

Adhesively Bonded Specimens

1. Prepare pre-adhesive blanks with the geometry shown in draw-

ing D-l.

2. Ultrasonically clean the pre-adhesive blanks by immersion in

a non-aqueous solvent bath. For non-metallic adherends, fol-
low the standard recommended practice for preparing surfaces
of plastics prior to adhesive bonding: ANS1/ASTM D2093-69
(Reapproved 1976). For titanium adherends, the stabilized
phosphate-fluoride treatment (method 1-S) is used. For
aluminum adherends, the phosphoric acid anodizing treatment
is used.

3. The bonding conditions prescribed by the adhesive manufacturer

are followed on the WTB pre-adhesive blanks with the exception
of pressure which is reduced by half to minimize run-out. Two
alignment holes are provided at the end of the specimen to '
prevent the bottom adherend from slipping relative to the top
during the cure cycle. It is not necessary to use this pre-
caution when a press is used.
784 E.J. RIPLING ET AL.

4. The specimen is machined to the final configuration shown in

drawing D-2. Sawing the taper angle into the previously bond-
ed top and bottom pre-adhesive blanks provides a uniform. smooth
surface between the adherends and the adhesive. The machining
conditions can be adjusted to accommodate the adherend material;
plastic. non-metallic composite. metallic composite. or metal.

BOTTOM TOP

I 1 0 . 250
(6.35)
0.250
0.2500
(6.35)

(6.35)

0.1250

1 r
0.7500
(3.2)
((9.05)

3500
(88.9)

2.500 2.250
(63.5) (57.15)

0.125 0.125
(3.2) (3.2)
I I
- 0 - 0 - ----L L -0-0-
I

0.250 ,0.250
(6.35) (6.35)
0.500 I
0.500
(12.7) (12.7)

i i
Figure. Dl. WTB specimen pre-adhesive blanks. Dimensions are
inches (mm).
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 785

BOTTOH TOP
I
---1 L
I
0.250
(6.35)
0.2500
(6.35)

F 0.250
(6.35)
-+-

314 0
(20.0)

3.250 3.000
(82.55) (76.2)

1/16"0 NOTE CI)

(16)

0.250
•
0.125
(3.2)
(6.35)

NOTES:
CD 1116" 0 ALIGNMENT
~ DIMENSIONS ARE INCHES (mm)

Figure D2. WTB specimen design.

NOMENCLATURE

A beam cross section area.

a crack length for a double cantilever beam (DeB) fracture spe-
cimen measured from the point of loading to the end of the
cracked region.

initial starter flaw length in a DeB specimen.

a coefficient used to describe the change of apparent tensile
o
modulus as a crack grows in a DCB specimen.

b crack front length.

B maximum specimen width for a width tapered beam specimen.

786 E.J. RIPLING ET AL.

c specimen compliance (displacement per unit load).

c' rate of change of specimen compliance with crack length, dC/da.

D depth of a 3-point bend specimen (same as h in a DCB fracture

specimen) •

DCB double cantilever beam.

E tensile modulus.

E coefficient used to describe the change of apparent tensile

o
modulus as the span-to-depth increases for a three-point bend
specimen.

£ crack extension force.

h height of one beam for a double cantilever beam fracture

specimen.

I moment of inertia.

k ratio of alb for a width tapered beam (WTB) specimen.

M bending moment (Px).

P load applied to specimen.

S distance between support for a three-point bend specimen.

V opening measured at the load line for a double cantilever beam

fracture specimen.

W length of a double cantilever beam fracture specimen from the

center of loading to the end of the specimen in the direction
of the crack growth.

u span-to-depth ratio (SID) for a three-point bend specimen.

x spatial coodinate in appendix

UDCB uniform double cantilever beam fracture specimen.

WTB width tapered beam fracture specimen.

y spatial coordinate in appendix.

a, e coefficientsfor a linear least squares regression relating

compliance and crack length.
FRACTURE OF COMPOSITE-ADHESIVE-COMPOSITE SYSTEMS 787

e coefficient relating load to crack extension force for a width

tapered beam fracture specimen.

~ constant of integration.

p coefficient relating crack length to load for a uniform double

cantilever beam fracture specimen.

v Poisson's ratio.

~ correction factor to the DeB specimen behavior when the crack

length dependent tensile modulus is considered.
CHARACTERIZATION OF THE FRACTURE BEHAVIOR OF ADHESIVE JOINTS

D. L. Hunston
Polymer Division
National Bureau of Standards
Washington, D.C. 20234, USA

A. J. Kinloch and S. J. Shaw

Ministry of Defence (PE), P.E.R.M.E.

Waltham Abbey, Essex, U.K.

S. S. Wang

University of Illinois
Urbana, IL. 61801, USA

The desire to use adhesives and composites in

structural applications has led to a need for a failure
prediction capability for the polymers used in such
systems. Unfortunately, this task is greatly compli-
cated by the failure load being dependent not only upon
the specimen geometry but also on the previous history
of loading, temperature, environment,. etc. For the
tough, rubber-modified polymers that are of most inter-
est for structural applications the effects of previous
history can be dramatic. As a result, predictions
based on measurements at a single set of conditions can
lead to dangerous over or under estimates of the frac-
ture behavior. In an effort to understand this problem
the present work has studied the fracture behavior of
various polymer formulations using bulk and adhesive
joint specimens tested over a variety of different
thermal and loading histories.

789
790 D.L. HUNSTON ET AL.

INTRODUCTION

The utilization of composites and structural adhesives in

aerospace, marine and automotive systems has increased markedly
in the last decade and this dramatic growth rate shows every sign
of continuing in the future. Since it is often desirable to join
composites with structural adhesives, this growth has placed a
heavy burden on the present state of adhesive bonding in terms of
both producing new materials that meet the rigorous demands of
these applications and developing the failure models required for
efficient design with such materials. Predicting the behavior of
structural adhesives is a difficult task because they are formu-
lated as complex mixtures of polymers and they are subjected to
constraints resulting from the fact that they are restricted to
a thin layer between relatively high-modulus substrates. The
formulations are designed to give tough materials and do so by
the introduction of complex viscoelastic and plastic energy
absorption processes. This makes it difficult to develop a basic
understanding and predictive capability for failure. The situa-
tion is further complicated by the fact that the stress-strain
field in a thin adhesive bond layer is difficult to analyse even
for linear-elastic materials, and structural adhesives are clearly
not linear elastic.

In an effort to address this highly interdisciplinary prob-

lem, a cooperative program has been established among scientists
at the National Bureau of Standards, Washington, D.C.; Ministry
of Defence (PERME) U.K.; Naval Research Laboratories, Washington,
D.C.; and the University of Illinois, Urbana. The objective in
this effort is to combine expertise in the areas of materials
science, polymer mechanics, engineering and stress analysis in a
unified effort to gain a better understanding of adhesive joint
failure. To accomplish this, the program is identifying and
studying the major variables that influence adhesive joint fail-
ure. This provides useful information for design, identifies
problem areas in failure prediction and helps to build the data
base of information that must be established if true design
criteria are ever to be successfully formulated. This program
has been underway for two years and the purpose of this paper is
to briefly outline some of the areas that have been studied and to
indicate future research directions.

BACKGROUND

The polymers used in structural adhesives are generally high-

modulus glassy polymers that are highly-crosslinked to give good
elevated temperature properties and to minimize creep. Unfortu-
nately, high cross1inking can produce brittle materials and thus
FRACTURE BEHAVIOR OF ADHESIVE JOINTS 791

Figure 1. Replica transmission electron micrograph of fracture

surface of rubber-modified epoxy material (bar indicates 2~m).

it is necessary to toughen them if they are to be used in struc-

tural applications. One of the most successful methods used to
toughen crosslinked epoxy systems is the addition of a rubber
which phase-separates. The curing reactions are controlled so
that the final material is a matrix of epoxy with small rubber
particles dispersed in and bonded to the matrix, as shown in
Figure 1. The reason for utilizing this type of two-phase system
can be understood as follows~-3 It is not difficult to generate
materials that are highly resistant to crack growth but in doing
so it is often necessary to make considerable sacrifices in other
properties such as modulus, elevated temperature properties and
creep resistance. The advantage of the rubber-modified material
is that the two-phase nature of the system makes possible toughen-
ing mechanisms that do not occur in a single-phase material.
These mechanisms can greatly increase the resistance to crack
growth in the material. Equally important is the feature that
these toughening processes are restricted to localised regions at
the tips of flaws, cracks or other stress concentrations. Thus,
792 D.L. HUNSTON ET AL.

since the matrix itself contains little or no rubber, the bulk

properties such as modulus, heat distortion temperature, and
creep resistance are not greatly reduced compared to those of
the unmodified epoxy. Consequently, the resistance to crack
growth is increased with a minimum sacrifice in other properties.
For this reason most commercial structural adhesives are based on
a technology of this type.

Although the exact details of the toughening mechanisms are

not completely understood, it is clear thl},t non-linear visco-
elastic/plastic processes are invo1ved. 1 ,4,5 This means that
only a stress-strain analysis involving a sophisticated constitu-
tive equation can hope to provide an adequate characterization of
the system. Moreover, the region in which these processes occur
is not small comEared to the dimensions of the adhesive bond
layer thickness. Consequently, the behavior of the bulk material
is modified by the constraining effect of the substrates when such
polymers are employed as adhesives. To address this situation the
present program is divided into two parts. Part one considers the
behavior of bulk specimens of both rubber-modified and unmodified
epoxy to provide base-line data. Part two then seeks to under-
stand how the constraints imposed by restricting the adhesive
between the substrates alter the bulk behavior. Only mode-I load-
ing will be discussed here; however, many of the conclusions would
be applicable to other types of loading as well.

EXPERIMENTAL

Materials

Experiments were performed using an unmodified epoxy and two

different types of rubber-modified epoxies. All materials were
cured with piperidine (5 parts per hundred resin, phr) at 120°C
for 16 hours. Specimens of the first type of modified epoxy were
formulated by adding various amounts (0, 5', 15 and 17.5 phr) of a
liquid rubber (carboxyl·-terminated poly(butadiene-acrylonitrile),
CTBN) to the epoxy resin (dig1ycidy1 ether of bispheno1-A) before
adding the piperidine. For the second type of modified system 24
phr of bispheno1-A (BPA) was added to the epoxy together with 5
phr of CTBN, prior to the addition of piperidine. With this
second rubber-modified epoxy system the cure chemistry is such
that a different final project is obtained. Unlike the former
rubber-modified systems where rubber particles of about 1 to 5 ~m
diameter are obtained (see Figure 1), the CTBN/BPA modified epoxy
not only results in particles of this diameter but also many small
particles with diameters of 0.2 ~m and less. This microstructure
has been termed "bimodal". To minimize any effects of physical
FRACTURE BEHAVIOR OF ADHESIVE JOINTS 793

and/or chemical aging, all specimens were given the same thermal
history prior to testing, i.e., a very slow cool (>12 hrs.) from
120°C to room temperature.

Bulk Specimens

Bulk fracture tests were conducted using single-edge notrhed,

compact tension and tapered-double-cantilever-beam specimens. ,2
After inserting a sharp precrack the specimens were loaded to
failure using crosshead-speeds ranging from Smm/s to O.OOOSmm/s
and temperatures from -60°C to +60 o C. The results of these tests
were converted to fracture energies using standard procedures. 1 - 3
Although the bending modulus is relatively insensitive to time
and temperature over the range of conditions employed for the
fracture tests, care was taken to use modulus values in the
fracture energy calculation that were measured at approximately
the same temperatures and time scales as the corresponding frac-
ture experiments. All tests employ mode-I type loading.

In addition to fracture tests, the basic mechanical proper-

ties of the materials were also determined. Linear viscoelastic
properties, such as the shear storage modulus and shear loss
modulus, were measured using a specially designed and computer-
ized test device which measures the dynamic mechanical and stress
relaxation properties of polymers. 7 Non-linear properties were
also determined in compression and tension tests; for example,
the true compressive yield stress was determined from uniaxial
compression tests. This information provides data on basic be-
havior as well as input for the stress analysis calculations.

Joint Specimens

The specimen geometry employed for the adhesive joint

fracture tests was a tapered-double-cantilever-beam joint, as
shown schematically in Figure 2. The substrate material was
either aluminium-alloy or mild-steel. As for the bulk fracture
experiments, tssts were conducted over a wide range of rates and
temperatures l , an~ ~tandard procedures were used to calculate
fracture energies. - All tests employ mode-I type loading.

RESULTS AND DISCUSSIONS

Bulk Behavior

Bulk mechanical properties. The dynamic shear modulus

 794 D.l. HUNSTON ET Al.

SUBSTRATE (MODULUS = Esl

LOAD, Pc

~=====-..............~~........~~h

ADHESIVE LAYER SUBSTRATE

Figure 2. Tapered-double-cantilever-beam adhesive joint specimen.

in the linear viscoelastic range was measured over 4 to 5 decades

of frequency and a temperature range of 80°C. The results were
then examined for the applicability of time-temperature super-
position. Despite the rather complex nature of the modified
epoxies, it was found that data for both the modified and unmodi-
fied materials superimposed quite well in the range from room
temperature to somewhat above the glass transition temperature.
The shift factors, aT' needed for the superposition are plotted
against temperature in Figure 3. Also shown in this figure are
results from shear stress relaxation experiments performed on the
same materials. These experiments in the linear viscoelastic
range indicate that reasonably good superposition can be obtained
with only a horizontal shift; however, slightly better results
are produced when a small vertical shift is included.

Figure 3 also contains the results from non-linear visco-

elastic data obtained by applying time-temperature superposition
to curves for true compressive yield stress vs. time to yield
(time required to go from zero load to the yield point). In this
experiment the experimental scatter and the small curvature in
the isothermal curves made it impossible to test for a vertical
shift. The results for linear viscoelasticity suggest that a
vertical shift is present but that it is small. Consequently,
the data here were treated using only a horizontal shift. Since
the objective in this work was to investigate general trends
FRACTURE BEHAVIOR OF ADHESIVE JOINTS 795
12

!i!!!! Oynemic MocIwIbI

• Opht OOpht

.15 pht 85pht

I!J 5 pht(bimodli cIiotribution)
+ 17.5 pht

...
OOpht
~

~ 5 pht(bIInodM cIiotriburion)

• 17.5pht

- 12

- 18
+
0.002 0.003 0.004 0.005 0.008
lIT ( It)

Figure 3. Value of the shift factor, aT' needed for super-

positioning of bulk mechanical data as a function of one over
temperature.

rather than to make highly accurate extrapolations, this proce-

dure was thought to be a reasonable first step.

As shown in Figure 3, the shift factors deduced from all

three experiments on a variety of different materials agree quite
well, particularly when the difficulties involved in maintaining
similar thermal histories in the different samples are considered.
Consequently, while the absolute magnitudes of the various mechan-
ical properties depend on composition, the shift factors do not.

In the temperature range that is of interest for the fracture

behavior, below 60°C, the shift factors can be fitted to a simple
Arrhenius type relationship

aT =B e LlE / RT (1)

where B is a constant and LIE is an activation energy. Values for

LI E are given in Table I.
796 D.L. HUNSTON ET AL.

Bulk fracture properties. It is well known that the energy

required to fracture even brittle, unmodified epoxy materials is
significantly greater than that required to break the chemical
bonds involved. Most of the energy expended in fracture goes to
viscoelastic and plastic deformation processes at the crack tip.
These deformation process~~ are rate and temperature dependent and
thus the fracture energY'~I ' also varies with rate and tempera-
ture. However, in unmodifiea epoxies the magnitude of these
effects, like the size of the deformation zone at the crack tip,
is small. On the other hand, for the rubber-modified epoxies the
fracture energy may exceed that of the unmodified material by more
than an order of magnitude 8 ,9. This arises because the deforma-
tion zone is considerably larger and this results in the rate and
temperature dependence of the fracture energy being considerably
greater. Figure 4, for example, shows the values of~I for two
modified epoxy formulations as a function of temperaturg measured
at three crosshead speeds: 2nrrn/s (F), 0.08 nrrn/s (M) and 0.0008
nrrn/s (8). These data demonstrate not only that the fracture

.....
'"E
-...,
.><:

>- 4
0)
"-
CD
<:
W

CD
:;
(.J
/ F
'"
"-
IL
2 '"
... t'"

o
- 60 - 30 o 30 60
Temperature ( ' e)

Figure 4. Fracture energy versus temperature at three crosshead

speeds for bulk specimens of 17.5 phr CTBN-epoxy: open symbols,
solid lines; and of 5 phr CTBN-BPA-epoxy: filled symbols, dotted
lines.
FRACTURE BEHAVIOR OF ADHESIVE JOINTS 797

energy varies dramatically with temperature and loading rate but

also that different formulations can have very different rate and
temperature dependences. Consequently, the conventional method
of testing materials at one set of conditions and presenting the
resulting values in tabular form can be very misleading. For a
given formulation the data obtained in one test may provide very
little information about how the material will respond in a test
under different set of conditions. Moreover, comparisons between
formulations can be totally incorrect if the temperature and/or
loading rate is chan~d. Consequently, only when a complete
characterization of is performed can various formulations be
c
evaluated and compare •

One method we have developed to characterize the loading rate

and temperature dependence is to perform a series of constant
crosshead speed experiments at various temperatures and speeds.
The data may then be presented in the form shown in Figure 4, but
may also be modelled by the expression:

11Ic =11Ics +A(tf)m

aT
(2)

where~IC = the measured value of the mode-I fracture energy,

l11 = theminimum value Of~I at low temperatures and

cs high speeds (see FigurecS),

A,m = constants,

t = the time taken for the load to increase from zero to

f that required for the onset of crack growth,

a = the shift factor taken from time-temperature

T superpositioning of the bulk mechanical data
(see Figure 3).

Typical plots Of~ versus Log (tf/aT) are shown in Figure 5 and
within the limits o~ the experimental uncertainty, the data fall
close to a single curve. Consequently, the time-temperature shift
factors obtained from measurements of viscoelastic properties can
be used to reduce the fracture data. It should be noted that with
the fracture data to an even greater extent than with the yield
data, the experimental scatter makes it difficult to judge the
accuracy of extrapolations based on time-temperature superposition;
~.e., it is impossible to estimate the uncertainty in values of
DI inFigure 5 for times to failure outside the range directly
me~sured. What is important here however is that the general
relationship between temperature and time for the fracture data is
roughly equivalent to that for the basic viscoelastic properties.
Now the shift factor, ~, may in turn be expressed as in Equation
798 D.L. HUNSTON ET AL.
9

8 • -saoe
)( - SO
6 -40
0-30
7 'J - 20
• 0
20
6 " 30
"
& 40 Rle8ER-MOOIFIEO EPOXY

"'....."-e 5
• 60 (15 phrl

:It

~4
C)

2
,
I
I

Figure 5. Fracture energy versus time-to-failure, tf/shift

factor, a y Data shown for 15 phr CTBN-epoxy and unmodified
epoxy.

(1) to give:
~
Ic -
-11Ics+ m -L'lEm/RT
C tf e (3)

where the parameters C, m and L'lE largely characterise the

magnitude, rate dependence and temperature dependence of,ljI •
As an example, when the different formulations shown in c
Figure 4 and 5 are analysed in this way, they give similar
values of L'lE but very different values for m, c, andl1J as may
be seen from Table I. Consequently, an expression sucnc~s
Equation (3) provides a useful way to evaluate the bulk mode-I
fracture behavior of different formulations.
FRACTURE BEHAVIOR OF ADHESIVE JOINTS 799

Table I. Fracture Parameters.

Formulation -4 cs
(kJ/m 2 ) C m L1E(kJ/mol)

5 phr CTBN + 24 phr 0.2 21650 0.15 160.5

BPA (bimodal)
5.0 phr CTBN 0.2 1236 0.12 160.5
15.0 phr CTBN 0.6 2935 0.12 160.5
17.5 phr CTBN 0.6 4064 0.12 160.5

Adhesive Joint Behavior

Effect of rate, temperature and joint geometry. The mode-I

fracture behavior of unmodified epoxy in an adhesive joint is in
most ways a predictable extension of the failure behavior of bulk
specimens: the fracture energy has the same value and is not
significantly dependent upon the detailed geometry of the adhesive
joint. However, with rubber-modified epoxies the size of the de-
formation zone is often similar to the thickness of the adhesive
layer and thus interactions must be considered.

For the rubber-modified epoxy, the mode-I adhesive fracture

energy,l1 I (joint), is shown as a function of adhesive bond thick-
ness, h, at various temperatures (specimen width, H=12mm; cross-
head speed, Y=0.002mm/s) in Figure 6, at various rates (tempera-
ture=22°C; specimen width=12mm) in Figure 7 and at various specimen
width (temperature=22°C; rate=0.02mm/s) in Figure 8. In all cases
for the joints the adhesive fracture behavior is dominated by a
strong dependence upon the adhesive bond thicknass, h, employed,
as previously found by Bascom and co-workers 6 ,1 for tough adhe-
sive sys~lms. The adhesive fracture energy passes through a
maximum'~I (joint), at a certain bond thickness, h. The value
of this maxlmum'~I (joint), is however dependent u~on the
temperature, rate a~~ specimen width, as may be seen from Figures
6, 7 and 8 respectively. Finally, in discussing the experimental
results, it is noteworthy that throughout these tests the locus of
joint failure was cohesive in the adhesive layer.
800 D.L. HUNSTON ET AL.

40
'23' K I ~O'C ) 20

I~

10

10

,., 0
) 10 ' K l7
I ' C) 20

0 I~

10

298' K(25'C) 20

15

1; 10
...
0:

......
Z
10
'!iN Z,
.....
... E
U ....
o:~
, n 0
r
'"~
20 Z
~
iii "
l. I~

10

10

0
0
253' KI-W'C) 20

"
10

10 5

0 0
233' K 1·40'C ) 20
30
15

20
10

10 5

0
00 005 010 0 15 020 025 030 0"
BONO THICKHESS. ft. 001

Figure 6, Adhesive fracture energy versus bond thickness at

various temperatures (H = 12mm; y = O.02mm/s).
FRACTURE BEHAVIOR OF ADHESIVE JOINTS 801

•
2
•
DISPLACEMENT RATE, Y .e·33 .10- 0 mi.

l...
I .lI
<:t
y .1.57.10-1 10/.
li
II:
III 0
Z
III

..
III
! 3
U
<C
II:
"-
•
III
>
iii
III
2 •
%
0
<C

Y.1·670 10- 4 mi.

0

Glcm
I JOINTI
3

•
2
•

o 2
BOND THICKNESS Imml

Figure 7. Adhesive fracture energy versus bond thickness at

various crosshead speeds (T = 20°C; H = 12 rom).
802 D.L. HUNSTON ET AL.

WIDTH, H. 3....

D~-----------r----------~------------~

1
~

It H .1",.
:i
I
•
'"
0
".
It;
i
~
!'"Q
C

N.12",,,,

D~-----------r----------~~----------~

N.SO ... In

DL-------------~------------~2~------------~3
IOND THICKNESS , ..""

Figure 8. Adhesive fracture energy versus bond thickness at

various joint widths (T = 20 o e; y = O.02mm/s).
FRACTURE BEHAVIOR OF ADHESIVE JOINTS 803

Theoretical analysis. Detailed calculations have shownll

that it is possible to estimate the optimum bond thickness, ~,
(for obtaining~ (joint» for a wide range of test rates
and temperatures 6~ using a plane-stress elastic-plastic model
of the crack tip deformation zone (Figure 9). Based on the
conventional "quasi-elastic" approach to estimate small scale
crack-tip yielding, the radius of the deformation zone, r I c'
can be calculated. The optimum bond thickness is then pre8icted
by assuming that the maximum adhesive fracture energy is obtained
when the size of the zone, 2rIyc' matches the bond thickness.
The parameters in the model are taken to be time dependent and the
time to failure, tf' in the fracture test is used to select the
appropriate values:
~Ic(tf)
h 2r Iyc (4)
m 7T cry (ty)~(tf)

where cry is the yield stress, and e y is the yield strain.

Values of ~ calculated from Equation (4) are shown together

with the experimental values in Table II. .fos may be seen, the
correlation is very good. The decline in<JI (joint) values at
adhesive bond thicknesses, h, less than h m ~y now be readily
understood. Namely, the presence of the high-modulus substrates
restricts the full volume of the plastic zone from developing;
and, since the toughness is mainly controlled by the energy
dissipated in forming the deformation zone, the adhesive fracture
energy is steadily reduced as the bond thickness is decreased. 6 ,lO

Beyond the conclusions discussed above, there are three

questions concerning the adhesive fracture behavior which can only
be answered by r.onsidering the details of the stress-strain field
that is present in the adhesive bond layer. The questions are:
(i) why does the value Of~I (joint) decrea~e at bond thicknesses
greater than hm' (ii) why iS~Icm (joint) > DIc (Bulk), a~d (iii)
why does the value of ~IC~ (joint) depend upon specimen w1dth
when the value of ~ c (bulk) is ~ significantly depen~ent
upon specimen width7 Although complete answers to these questions
are not yet possible, some interesting qualitative notions can be
offered. As implied in the opening statement of this paragraph,
the stress-strain field ahead of a crack in an adhesive layer is
804 D.L. HUNSTON ET AL.

AT MAXIMUM ADHESIVE FRACTURE ENERGY, Glcm:-

" "
SUBSTRATE "-
'" " -

Figure 9. Elastic-plastic model for deformation zone at the crack

tip. As drawn, h = h •
m

greatly influenced by the constraints imposed by the presence of

the high-modulus substrates. Wang and co-wol~ers have conducted
such a stress analysis using both an elastic and an elastic-
plastic13 finite element analysis. These analyses demonstrate
that as the adhesive bond thickness decreases, the high crack-tip
stress levels extend farther down the adhesive layer than they do
in bulk specimens or thick adhesive bonds. This results in a
change in the shape of the deformation zone in the thin adhesive
bonds; i.e., the zone extends farther down the bond line.

From the discussion above it is clear that there are two

effects that must be considered when estimating the size of the
deformation zone and the expected value of fracture energy: the
dimensions of the zone in the adhesive bond thickness direction
and its length down the adhesive bond layer. Very thick bonds
give results similar to bulk specimens. However, as the bond
thickness is decreased, the first effect is the increasing degree
of constraint which causes an extension of the high stress levels
further down the bond layer and in response to this the deforma-
tion zone extends further down the bond. This results in an
FRACTURE BEHAVIOR OF ADHESIVE JOINTS 805

Table II. Comparison of Calculated and Exper!mental Values of

Adhesive Bond Th~kness, hm' at Maximum Adhesive
Fracture Energy, I cm (joint).

Temp, T Crosshead Joint width Value of h m (mm)

speed, y
(OC) (mm/s) H(mm) Theoretical Experimental
(from eqn (4) )

20 0.0008 12 0.85 1.0

20 0.02 12 0.70 0.8
20 0.2 12 0.49 0.55
20 1.2 12 0.43 0.4

20 0.02 3 0.75 0.85

20 0.02 6 0.72 0.65
20 0.02 12 0.70 0.8
20 0.02 25 0.65 0.7
20 0.02 50 0.61 0.8

50 0.02 12 1.6 1.1

37 0.02 12 1.16 0.9
25 0.02 12 0.57 0.6
0 0.02 12 0.39 0.5
-20 0.02 12 0.15 0.25
-40 0.02 12 0.05 <0.1

increase in the zone size and the adhesive fracture energy. As

the bond thickness continues to decrease eventually a point is
reached where the extension down the bond layer becomes secondary
to the restricting effect of the decreasing space available in the
thickness direction. The zone size and fracture energy then begin
to decrftse after passing through a maximum. Experimental obser-
vations of the growth of the crack-tip deformation zone made
with high-speed movies are in good qualitative agrfzment with the
predictions of the elastic finite-element analysis and a more
detailed quantitative analysis that is to be published by Wang
et.a1. 13 • Finally, the degree of constraint is obviously also a
function of joint width, H. For example as H increases the
degree of constraint increases and thus~I (joint would be pre-
dicted to increase in value compared tol1I~m(bu1k), and this is
indeed observed as may be seen from Figure 10. Thus, the same
arguments may be used to explain the effects of both adhesive bond
thickness and specimen width upon the adhesive fracture energy.
Although a quantitative explanation of these effects will probably
require a detailed stress-strain analysis with a more realistic
constitutive equation, the simple idea presented here is qualita-
tively consistent with all of the experimentally observed results.
806 D.L. HUNSTON ET AL.

--------------
Gte (BULK ADHESIVE)

o 10 20 30 40 50
WIDTH OF SPECIMEN, H (mm)

Figure 10. Fracture energy versus specimen width, H (T = 20°C;

Y= 0.02mm/s).

CONCLUSIONS

The studies on the bulk epoxy materials have demonstrated

that there is a correlation between temperature and time-to-
failure that makes it possible to apply time-temperature super-
position techniques to help characterize the fracture behavior of
these materials. This characterization has enabled a semi-
empirical equation to be developed which provides a means of
modelling the properties of the materials over a wide range of
test rates and temperatures. Moreover, since the behavior of
rubber-modified epoxies changes substantially when the formulation
is altered, these techniques should provide a basis on which to
establish structure-property relationships that will assist in the
development of optimized materials.

The studies on the adhesive joints fabricated with rubber-

modified epoxies have shown that their fracture behavior is
dependent upon (i) the fracture energy of the bulk epoxy adhesive
and (ii) the degree to which the geometry of the bonded joint
enhances and/or restricts the toughening processes normally pre-
sent in the bulk adhesive. The importance of this second para-
meter has been demonstrated experimentally and qualitatively
modelled using both elastic and elastic-plastic stress-analyses.
FRACTURE BEHAVIOR OF ADHESIVE JOINTS 807

REFERENCES

1. J.L. Bitner, J.L. Rushford, S. Rose, D.L. Hunston and C.K.

Riew, J. Adhesion, 13, 3 (1981).
2. A.J. Kinloch and R.T. Young, "Fracture Behavior of Polymers",
Applied Science Publishers, London, 1983.
3. D.L. Hunston, J.L. Bitner, J.L. Rushford and J. Oroshnik,
J. E1ast. Plastics, 12, 133 (1980).
4. A.J. Kinloch and J.G:-Wi11iams, J.Mater.Sci., 15, 987 (1980).
5. A.J. Kinloch, S.J. Shaw and D.L. Hunston, in "Proceedings
Int. Conf. on Yield, Deformation and Fracture", held in
Cambridge, Plastics Rubber lnst., London, 1982.
6. W.D. Bascom and R.L. Cottington, J. Adhesion, l, 333 (1976).
7. D. L. Hunston, W.D. Bascom, E.E. Wells, J.D. Fahey and J.L.
Bitner, in "Adhesion and Adsorption of Polymers", L.H. Lee,
Editor, Pt. A. p.321, Plenum Press, New York, 1980.
8. E.H. Rose, A.R. Siebert and R.S. Drake, Mod. Plastics, 47,
llO (1970).
9. J.N. Sulton, R.C •.Laib1e and F.J. McGarry, J. Appl. Po1ym.
Sci., ~, 127 (1971).
10. W.D. Bascom, R.L. Cottington, R.L. Jones, and P. Peyser,
J. App1. Po1ym. Sci., 19, 2545 (1975).
11. A.J. Kinloch and S.J. Shaw, J. Adhesion, 12, 59 (1981).
12. S.S. Wang, J.F. Mandell and F.J. McGarry,-rntern. J. Fracture,
14, 39 (1978).
13. S.S. Wang, D.L. Hunston and A.J. Kinloch, to be
published.
14. D.L. Hunston, A.J. Kinloch and J. L. Rushford, to be
published.
STRUCTURAL PRECURSORS TO FRACTURE IN ADHESIVE JOINTS

Wartan A. Jemian

Materials Engineering
Department of Mechanical Engineering
Auburn University, AL 36849

Sructure-property relations in adhesive joints

are reviewed for the purpose of explaining the general
nature of conditions and structural processes leading
to fracture. Reference is made to selected homogeneous
materials and adhesive joints to illustrate specific
features. The adhesive joint is described as a com-
posite structure composed of adherend and adhesive
layers and various chemical and structural transition
layers. The nature and dimensions of materials struc-
tural features are considered. Traditional and modern
methods of analysis are summarized with special empha-
sis placed on the use of materials characteristic cur-
ves to indicate the existence of structural precursors
to fracture. These features, in relatively small quan-
tities, severely alter the mechanical properties of the
adhesive joint. In only a few instances have struc-
tural precursors to fracture been directly observed and
identified due to their small dimensions, distribution
in the system and limited quantity. It is concluded
that identifying the presence of structural precursors
to fracture by the use of materials characteristic cur-
ves may aid the avoidance of adhesive joint failure.

809
810 W.A.JEMIAN

INTRODUCTION

The general study of structural precursors to fracture is

important in adhesive joints as well as any structural materials
system because it promises the basis for a method to anticipate and
hence avoid fracture. Even before this objective is fully
realized, it is advantageous to better understand the nature of
events leading to fracture. This study requires a definition of
the significant features of the original structure and a review of
structural changes that are known to occur in the separate, homoge-
neous materials of the adhesive joints.

An adhesive joint is a layered structure in which strong,

elastically stiff materials are chemically bonded (covalent, ionic,
metallic bonding) through a second substance, which has a lower
strength and stiffness. These are called the "adherend" and
"adhesive", respectively. The principles governing properties of
the joint relate also to more general cases of adhesion, such as
plating, in which the term, "joint", is not appropriate. The adhe-
sive joint usually is a combination of any basic type of material
such as metals and alloys, glassy or crystalline ceramics, and any
of the many organic plastics and rubbers or other high molecular
weight compounds. Therefore, this review draws upon common ideas
and similarities in the special fields touched upon.

In a symmetric loading system which is frequently the case for

lap-shear specimens with a center of inversion symmetry at the
center of the overlap, deformation and fracture are usually also
symmetric. Figure 1 depicts the nature and sequence of progressive
changes that take place in aluminum-epoxy-aluminum lap shear
samples under load, as reported by Wilcox and Jemian.l,2
Initially, crack nucleation begins by a void condensation process,
followed by slow crack growth toward the center of the overlap
until a critical condition develops for rapid crack extension. The
initial processes start at the ends of the overlap close to the
center of the adhesive and rapid crack growth occurs close to the
adherend at the interface with the higher shear stress. The final
separation at the center of the overlap is a tearing process under
overload conditions after the joint has effectively lost all of its
original strength. Bending, initial separation, crack extension
and final decoupling of the two alloy strips occur symmetrically,
within the precision of observation. There are similar,
progressive changes that take place in a non-symmetric joint, where
separation occurs along one surface only. Therefore, in this ana-
lysis a single adherend-adhesive composite is used to represent an
entire assembly or joint.

Materials structure and phenomenology represent two principal

considerations in understanding the characteristics of materials.
Materials are articulated systems where the nature of the response
to any action depends on the details of physical arrangement at all
STRUCTURAL PRECURSORS TO FRACTURE 811

o : ;....
---:-- .'
-/- ,; .,

LEGEND
o ADHEREND

~ ADHESIVE
,
D ADHESIVE
",
.0 NOMINAL LOADING DIRECTION

~)

Figure 1. Configuration of the lap shear joint.

(a) Macro-structure under load
(b) Profile of the fracture path

structural levels. This aspect is emphasized in this presentation.

However design requires numerical values which are obtained by the
application of phenomenological relations, which are identified
with each deformation process.

Geometry, dimensions and configuration are important. The

mechanical actions transferred through the bulk structural members
to the region of the interface produce components of normal and
shear strain. Table I lists the basic elements of the terminology
of solid mechanics that are referred to. Figure 2 illustrates the
relationship of these variables to a volume element of the
material. The volume element represents any portion of the adhe-
sive joint, or other materials system, under load. Loading con-
figuration and sample deformation control the proportions and
distribution of stress components, including the interfacial
region. These change with loading and sample shape. In addition
to their general stress distribution, local features such as cracks
and interphase boundaries between phases of markedly different
stiffness produce stress fields that are locally complex and
include concentrated stress components. 3 ,4
812 W.A.JEMIAN

(0) (b)

K
I
I
---
....

I
I
I
Pi,
I , .....
.....
..... ........................
I . , / ---

(c) (d)

Figure 2. Mechanical Actions and Materials Response.

(a) Stress components (c) Shear, showing the shear angle, y
(b) Strain components (d) Dilatation under hydrostatic pressure

Adhesion testing is no longer limited to the measurement of a

few, arbitrary sample properties. Adhesion scientists recognize
structural similarities in bulk constituents and transition zones
of adhesive joints although terminology varies considerably. At
different stages in the loading a variety of changes take place.
Some are the same as those that occur in the bulk components,
separately loaded. Others have their origin in the unique chemical
and structural conditions of the adhesive joint.

Mechanical Characteristics

Flow (stress-strain), creep and stress relaxation curves pro-

vide comprehensive information about loading response and sample
behavior. 3 Certain features are expected for each type of sample
by material and configuration. In the lap-shear configuration the
separation surface has features of the form shown in Figure 1.
Different combinations of materials and dimensions produce dif-
ferent combinations of these features and others not shown in this
STRUCTURAL PRECURSORS TO FRACTURE 813

Table 1. Forces, Stresses, Strains, Strain Rate and Deformations in

Materials. (Refer to Figure 2)

FORCE STRESS STRAIN STRAIN DILATATION SHEAR

RATE

Normal <111 £11

.
£11 *
Normal <122 £ 22 ~22 *
Normal <133 £ 33 £ 33 *
Shear '12 112 112 *
Shear '23 123 123 *
Shear '31 131 131 *

*Deformation associated with the listed variables.

NOTE: The subscripts, i, j, in <1ij etc. refer to force or displa-
cement in the j-direction on a surface normal to the i-direction.

sketch. There is greater variety possible. It is frequently

possible to interp~et these differences in terms of the relating
magnitudes of stress and strain components of Table 1 and Figure 2,
and their time derivatives. At a crack at the edge of the overlap,
at the beginning of adhesive failure, the opening stress is maximum
and the shear stress is zero, as. shown schematically in Figure 3.
This distribution was described by Masubuchi and Keith 4 , in a
review of deformation mechanisms, confirmed by finite element
structural analysis by Grimes 5 , photoelastically by Jemian and
Ventrice 6 , and supported by a closed-form mathematical solution
procedure by Renton and Vinson 7 • The photoelastic analysis showed
that the stress distribution form is constant and follows the crack
tip as it advances in the fracture process. 6

In a long joint, a transient "steady state" may be

established until terminal conditions prevail. The conditions of
the symmetric, tapered double cantilever fracture toughness test
sample and the peel test sample represent more clear examples of
the action of steady state stress distributions during the fracture
process.

The materials characteristic is the sequential response of

the sample to a programmed action under controlled conditions.
Stress-strain curves are one form of mechanical materials charac-
teristic. Other materials characteristics depict response to ther-
mal exposure, magnetic or electric fields, etc. There are three
related types of materials properties; (i) slopes and parameters of
the continuous portions of the characteristic curve, (ii) critical
values of either the action or reaction variable where the features
of the curve change, and (iii) a critical value of some condition;
e.g. temperature or pressure, that causes the characteristic curve
to have a different shape or show some new feature.
814 W.A. JEMIAN

I
a:
T

ADHESIVE

ADHERENO

Figure 3. Stress distribution in the lap shear sample.

Well known examples of these properties are provided by the

standard test procedures for homoge.neous alloy samples. 3 The typi-
cal tensile materials characteristic for a ductile alloy sample is
shown in two ways in Figure 4. The engineering stress-strain curve
is the more common representation and is equivalent (it presents
the same information in a different form) to the true stress-
natural strain curve which relates better to phenomenological
representations. The two curves coincide closely in the range of
elastic deformation, the first smooth region below the elastic
limit, 0e. The other regions, uniform plastic flow and non-uniform
plastic flow differ in form and deformation process.
Phenomenological relations are identified with each region in
Figure 4.

Three transitions are also shown with critical value proper-

ties identified. The elastic limit marks the onset of slip, the
ultimate tensile strength, which is the engineering measure of the
maximum load that can be supported, marks the onset of significant
production and accumulation of voids. Breaking, the last tran-
sition involving crack nucleation and growth, is usually overlooked
in testing alloy samples. This terminal detail is frequently
displayed in tests of adhesive joints because response times are
different. This final portion of the characteristic in the alloy
sample involves crack extension under the action of essentially
constant displacement, a stress relaxation process.

In the alloy sample, the first deformation process is the

distortion of the crystal structure in the grains without any
changes between nearest neighbor atoms. Uniform plastic defor-
mation, typified phenomenologically by a constant strain hardening
exponent, n, is principally a constant volume process with a small
increase in elastic distortion proportional to the level of
STRUCTURAL PRECURSORS TO FRACTURE 815

~ 0;
,,~ /
/
/
/
/
/
/
/
/
/
/
/
/

Figure 4. Mechanical Characteristic of ductile alloy. ae • Elastic

limit; UTS, ultimate tensile strength; SB. aBo breaking
strength; KI. K2. constants.

loading. Other processes such as vacancy production occur but are

only important ~n the later stages of deformation. Vacancy produc-
tion leading eventually to void formation results in the typical
fibrous fracture surfaces observed in ductile materials. The UTS
marks the onset of significant void production as the yield
strength marks the onset of slip.

The ductile-brittle transition temperature of the Charpy test

or dropping weight impact test is an example of the third type of
related materials property. In this procedure. a separate sample
is tested at different temperatures spanning the temperature range
where the fracture work. as well as the dynamic stress-strain
characteristic curve change. The temperature associated with the
changing characteristic is reported as a property of the material. 3

A constant linear slope or smooth constant form of non-linear

characteristic indicates that structural changes within the system
are of the same kind. Changes in curvature or line continuity
indicate the action of some new internal structural process.
"Pop in" in a fracture toughness test is such a feature as is the
more subtle deviation from linearity observed in deformation curves
816 W.A.JEMIAN

in the region of the elas.tic limit or in the rate of change of spe-

cific volume with temperature at the glass transition temperature
of a glassy phase. The observation of such a feature indicates a
significant change in the material.

Phase Structure Model

The aluminum-epoxy system is referred to as a typical adhe-

sive joint system. The epoxy is a thermosetting polymer with a
multiply connected network of primary bonds along aliphatic and
aromatic molecular segments. In addition to the strong covalent
bonds, distributed secondary bonds also stabilize the phase struc-
ture. The epoxy is an amorphous materials system. The adherend is
polycrystalline. Experience shows that surface preparation of the
adherends is as important as formulation of the adhesive to the
ultimate joint properties. Degreasing, followed by chromic acid
etching and finished by controlled washing and drying are critical
to the final result. The objective is to form a coherent layer of
hexagonal Bayerite (~-AI203 • 3H20), which strongly absorbs the
epoxy, rather than pseudo-Boehmite (A1203 • X H20, X~2) that is
formed in boiling water. B,9

The final joint involves the transition from alloy to adhesive

through an hydrated oxide and an internal surface of attachment where
epoxy molecular segments are chemically bonded to the oxide. This
attachment is highly organized in comparison to the bulk epoxy
phase. The regular termination of the epoxy branches at the oxide
surface produces a region in the epoxy with a non-typical molecular
organization. In this region the molecular segments are closely
aligned and positioned to match the oxide surface.

The attachment density at the interface is estimated to be of

the order of IOIB m-2, taking the molecular spacing to be approxi-
mately I nm. A typical hydrated alumina surface provides 3.2 x IOIBm-2
sites. By comparison there are 1.4 x I019m-2 attachments across
(III) in aluminum.

These features are represented schematically in Figure 5.

The alloy and coherent oxide layer of the adherend are represented
by arrays of dots. It is assumed that this region is within a single
grain of the alloy. The epoxy is represented by lines, intended to
indicate molecular conformation in a general way. As a natural
consequence of the structures of the phases and their rela-
tionships, a number of layered regions can be identified. These
are:

M metallic substrate
OMI oxide-metal interface
AOI Adhesive-oxide interface, referred to above as the
internal plane of attachment.
STRUCTURAL PRECURSORS TO FRACTURE 817

OA Ordered Adhesive, where the effects of the structural

regularity of the adhesive molecule and the plane of
attachment persist.
BA The remaining adhesive phase with a structural organi-
zation not influenced by the interfacial attachment •

. ..
. . . . . . . . . . . . . .. . . .
. . . .. .. . . . . . . . . . . M
. . . . . . . . . . .. . . . . .
Figure 5. Phase structure in the regions of the aluminum-epoxy
adhesive joint (schematic). BA, bulk adhesive; OA,
ordered adhesive;AOI, adhesive-oxide interface; 0, oxide
layer; OMI, Oxide-Metal Interface; and M, metallic
substrate within the adherend.

If the adhesive layer is thin the effects of surface organization

affect the structure of the entire adhesive layer. This, in turn,
produces changes in adhesive joint characteristics, which are typi-
cally an increase in measured strength and a shift in fracture
locus toward or into the adherend or oxide.

Interfacial contaminants, such as the monolayers or partial

monolayers of Langmuir 10 ,11, Blodgett 12 , Langmuir and Schaeffer 13 ,
and Jacquet 14 interrupt the interface bonding and lower measured
adhesive strengths, as well as lead to apparently clean interface
separations. Jacquet used this effect to initiate a peel test ("T"
test) of an electrodeposit. Jemian 1 5 found that interfacial con-
tamination controllably reduced dislocation interactions with
metal-rnetal interfaces. These effects, however, are in a separate
category from the structural features that develop out of a well
formed joint under load.
818 W.A.JEMIAN

Stress Effects

The application of forces to the adhesive joint invariably

results in complex stress distributions including both general
stress field details and local variations due to the presence of
structural features. There are many examples of the effects of
simplified, homogeneous stress fields on various materials. In
some situations these effects are attributed to the stress field
alone and in other cases the effects are complicated by ambient
conditions or external factors. The stress effects include enhan-
cement or development of the following processes and structural
features:

Elastic deformation,
Flow,
Void formation,
Impurity production,
Transformations, and
Transition macrostructures.

These processes are described in a manner consistent with the dif-

ferent phase structures involved in homogeneous materials and
corresponding regions of the adhesive joint. Table II is an
approximate size scale for structural features in adhesive joints.

Elastic deformation, in reference to alloys and other engi-

neering materials, is extension and bending within the yield
strength or elastic limit, and is usually in direct proportion to
the stress. Elastomers, rubber modified adhesives, and a variety
of other materials undergo initial deformations, when the load is
gradually applied, that are continuous, non-linear functions of the
stress and may result in pronounced shape changes within the appli-
cable range. These processes are recoverable distortions of inter-
component spacing in the phase structure and changes in component
shape. The energy content of the elastically deformed phase is
increased due to both potential energy and entropy factors asso-
ciated with spacing and shape changes, respectively.
Phenomenologically, the possible deformations expand into the realm
of finite elasticity16 which includes the effects of major shape
changes and other nonlinearities.

In the adhesive joint the adherend and oxide layer, which are
the stiffest constituents, undergo linear elastic deformation. The
bulk polymer is capable of some recoverable extension by rotations
about carbon-carbon bonds within molecular branches. This process
also involves some reversible intercomponent shear. The polymer at
the oxide surface is considerably restricted and has a preferred
orientation and structure. These factors increase the elastic
stiffness in homogeneous samples and have the same effect in the
adhesive joint.
STRUCTURAL PRECURSORS TO FRACTURE 819

Table II. Size scale for features of materials structure in adhe-

sive joints

SIZE FEATURE
ELEMENTAL MACROMOLECULAR
SOLID PHASE
Imm

CRACK
O.lmm GRAIN SPHERULITE
TRANSFORMATION NEEDLE
DISLOCATION CLUSTER
10JLm SU P LI NE

CRAZE
IJLm

O.IJLm SHEAR BAND

MICROCRAZE
VACANCY CLUSTER VOIDS
10 nm FIBRILS

I nm MOLECULAR SPACINGS
BURGERS VECTORS
ANIONS
0.1 nm ATOMS
CATIONS
LATTICE VACANCIES
820 W.A.JEMIAN

Flow processes differ in mechanism in solid and glassy phases

due to crystal compatibility requirements in the former and lack of
such requirements in the latter. Flow includes any structural pro-
cess involving a sequence of phase structure equilibrium states.
When the load is removed the system retains its immediate con-
figuration. Thus, flow results in permanent set. Flow is enhanced
by an increase in temperature, especially if the increase involves
the glass transition. Gent 17 finds that the effect of an applied
stress is to enhance flow mechanically, thus effectively reducing
the glass transition temperature. Crystalline materials flow by
slip, twinning, and stress induced martensite transformation. This
is true of both metallic and non-metallic crystals. These changes
occur by progressive action and involve dislocation motions which
mark the extent of the crystallographic shear.

When flow occurs in either a crystalline or glassy phase,

rotations are usually induced. Molecular branches tend to align
with the principal normal stress direction and crystallographic
deformation directions align with the direction of maximum shear.

Flow in a glassy macromolecular substance involves local

relaxations between neighbor branches. Permanent set is achieved
if the new configurations are similar to the original. This is
viscous flow in which the strain rate, € or y, is a function of the
stress. The relation is linear for Newtonian liquids. The tensile
viscosity, ~, and shear viscosity, ~, are defined as ~ii = aii/Eii
and ~ii - Yii/Tii, where ~ = 3~.3
Alfrey discusses general relations between bulk polymer
mechanical properties and adhesive joint properties, including
yield strengths. 18 The principal deformation processes in glassy
polymers are shear band formation .and crazing. 19 - 22 Microshear
bands, which generally form at about 40° to the compression direc-
tion are illustrated in Figure 6. Diffuse shear bands which also
form under compression, but closer to 45°, are illustrated in
Figure 7. Crazing, which forms under tension is illustrated in
Figure 8. The craze surface is perpendicular to the loading direc-
tion as shown. Oxborough and Bowden found that all - v (a2l + a23)
= EEll, where E'is Young's modulus and v is Poisson's ratio, provi-
des a meaningful criterion for crazing.23 Figure 9 shows the cri-
teria for polymer deformation in two dimensions. The elliptical
boundary for shear band formation is similar to the Von Mises cri-
terion for yielding in metals. Its center, however, is displaced
towards biaxial compression. Craze formation occurs at a much
lower stress magnitude and in the tensile region.
STRUCTURAL PRECURSORS TO FRACTURE 821

Figure 6. Microshear band formation in bulk plastic under

compression. Typical bands are inclined approximately
40° to the compression surface.

Figure 7. Diffuse shear band formation by compression of bulk

plastic. Shear band is inclined approximately 45° to
the compression surface •

. a .

o " ..... 0
. 0 ' 0 •

Figure 8. Crazing. The craze is an opening crossed by oriented

fibrils in the plastic.
822 W.A. JEMIAN

(/22

/ SHEA
/
/
/ YIELDING

Figure 9. Criteria for shear yielding and crazing in polymers.

The effect of temperature on crazing is indicated.

A further distinction between the two deformations is that

crazing is dilatational and associated with a decrease in density.
There is also a loss in cohesion and strength but not to the same
extent as in crack formation. The craze contains substance and is
capable of supporting loads to a reduced extent. Crazing develops
from voids. The craze volume is marked by many fibrils oriented
across the surface. The fibrils have dimensions of the order of 10
nm thickness and the spacing between is of the same order.

Crazing is a deformation mode observed in thermoplastic

polymers or weakly crosslinked thermosets. Void formation, which
is a precursor to crazing has been observed in thermosetting adhe-
sive systems.

Flow in the ordered adhesive is restricted. A number of

investigators have reported an increase in fracture strength with a
decrease in adhesive thickness. 24 - 27 Bascom and Cottington report
deviations in this relationship and provide a more detailed
STRUCTURAL PRECURSORS TO FRACTURE 823

explanation. 28 The increase in strength with decreasing joint

thickness that is generally observed is attributed to a different
specific mechanism in each system. However, the trend is general
and the fracture mode changes from ductile to brittle. Possible
explanations are that the flow processes are eliminated in these
adhesive layers due to a boundary effect which is an apparent mani-
festation of the Bikerman weak boundary layer theory.29 Another
contributing factor is the ability of the ordered adhesive to sup-
port a higher stress due to the increased elastic limit in the
ordered structure (flow is severely restricted). A third possible
reason is that since the thickness of the adhesive layer is reduced
and the principal deformation occurs in this layer, the effective
strain rate is increased for any given loading rate. This favors
brittle fracture.

Void formation is an included process in the deformation and

fracture of both alloys and polymers. 3 ,19,30 The increase of
entropy of mixing accounts for an equilibrium vacancy content in
crystals that have an Arrhenius temperature dependency.
Dislocation interactions and diffusion provide mechanisms for the
attainment of these concentrations. Continued vacancy production
by dislocation interactions in the loaded member cause vacancy
super-saturations that are relieved by diffusion and clustering.
These structural features are usually of the order of 15 nm in
small dimension. Macromolecular glassy phases include structural
features that provide for analogous effects. Void production is
accompanied by decreases in specific gravity and strength.

As a point of interest, vacancies are generally found to have

a strengthening effect in metals and dilute solid solutions. 30
They are associated with local elastic distortion in the crystal so
that the volume of the missing particle is partially occupied.
These lattice distortions cause dislocation interactions which
reduce dislocation mobility. On the other hand in a tensile test
void production opposes work hardening, leading to the pronounced
necking observed beyond the ultimate load. The center gage
thinning reported by Miklowitz,31,32 MacGreggor,33 and others is
attributed to other factors.

Impurity production is more a result of external factors,

such as atmospheric humidity and the absorption of ultra-violet
radiation than the application of stresses alone. The degradation
of plastics is comprehensively reviewed by Pinner. 34 In addition
to these factors the dilatation of a portion of the joint due to
the stress field favors the accommodation of molecules and molecu-
lar fragments that alter the local bonding, phase stability and
mechanical properties. Impurities are expected to be of the order
of molecular spacings as represented in Table II.
824 W.A. JEMIAN

Transformations, including crystallization, stress-induced

martensite transformations of crystalline phases and chemical reac-
tions, occur spontaneously to reduce the energy of a substance with
stored elastic energy. Under stress the system is in a higher
energy state. If a new configuration is possible at a lower
energy, it will form. Polymers react to form new molecular groups
by mechanisms involving free radicals or chemical agents already in
the system. These changes, as is the case of impurity production
are usually eliminated from adhesive joints by joint design and
materials selection. It is possible, however, to develop an unde-
sirable layer and lose strength and toughness. The transformation
product may be large enough to be included in the next category.

Transition macrostructures are features in the macroscopic

size range that form out of the elements of the original phase
structure under the conditions that prevail. They are
distinguished by their structure, size, and shape and are tran-
sitional since they lead naturally to other features connected with
the fracture process. Of these features listed in Table II it is
possible to single out

shear bands,
crazing,
voids, and
crystallites or other transformation products.

These are are the structural precursors to fracture and are detectable.
Critical dislocation clusters are also precursors to fracture but
not in adhesive joints. The failure envelope for the adherend in
the well designed adhesive joint is much larger than that of the
adhesive so that adhesive failure shields against adherend failure.

The actual development of a craze or other macrofeature makes

a profound change in terms of stress concentrations and system
behavior. Beardmore and Rabinowitz show that the stress for
crazing in PMMA is approximately 80% of the fracture stress. 35
This is analogous to experience in metal fatigue failure in that
90% of the life of the item elapses before the fatigue crack is
nucleated. The appearance of the precursor is a positive indica-
tion of imminent fracture.

METHODS OF CHARACTERIZATION

The most notable development for adhesion science is the

availability of a large variety of sophisticated, reliable instru-
ments, that provide information well beyond the capabilities of the
STRUCTURAL PRECURSORS TO FRACTURE 825

classical microscope, radiograph and universal test machine. The

older instruments, never-the-less, retain their importance in basic
analytical procedures. The new instruments provide sensitivity and
method out of the range of the earlier models. The functions that
are now available are in the following categories:

chemical and structural analysis

observation
process monitoring
materials characteristic determination.

Chemical and structural analyses may be applied to the in

situ determination of submicroscopic sample regions. This includes
the electron microprobe, EDAX attachments to scanning electron
microscopes, Auger electron spectroscopy, selected area electron
diffraction, ellipsometry and other surface characterization tech-
niques. To this list can be added three facilities that require
removal of portions of the sample for analysis but that can be
used to identify the existence of molecular fragments and tran-
sition structures. These include the scanning calorimeter, NMR
spectrometer, chromatograph and IR spectrometer.

Modern observational techniques include the transmission and

scanning electron microscopes, polarized light, phase contrast and
other interference techniques for the enhancement of contrast. There
are many such methods and they all contribute information.

Process monitoring may now be accomplished by such methods as

acoustic emission in which characteristic mechanical waves are
emitted in conjunction with specific processes. These sounds can
be isolated from a noisy environment to provide a reliable signal
of the occurrence of some specific event in the sample. Alternate
methods, in which the sample interacts with radiation, which is
monitored, are also available. Ultrasonics provide reflection
techniques, useful if new surfaces are formed and attenuation
methods that have proven useful in grain size determination and
similar applicatons. A variety of physical properties may now be
employed along with sensitive transducers and signal processing
equipment. In many cases signal data can be computer processed to
isolate and enhance the desired information.

This same signal enhancement can be used to record the fine

structure of a materials characteristic. Although last on this
list, the materials characteristic is the most direct analytical
test and provides the most useful information to identify separate
deformation processes and to recognize the formation of structural
precursors to fracture.
826 W.A.JEMIAN

CONCLUSION

The science of materials has progressed to such an extent that

a number of transition structures are recognized in materials under
load. Techniques are available to systematically identify their
presence and formation. The first indication of the presence of
new structural entities is provided by discontinuities in materials
characteristic curves or changes under different conditions even
though direct confirmation may not be possible. Other, more speci-
fic information may be gained by modern methods of fracture surface
observation, microchemical analysis and macroscopic structural
determination. Eventually in-process monitoring can be used to
detect the presence of structural precursors to the fracture pro-
cess and thus avoid adhesive joint failure.

ACKNOWLEDGEMENT

The author is grateful for helpful discussions with B.Z. Jang,

who very generously provided information related to this subject
from his PhD dissertation at Massachusetts Institute of Technology.

REFERENCES

1. R.C. Wilcox and W.A. Jemian, Metall. Rev., 1,13 (1972).

2. R.C. Wilcox and W.A. Jemian, Polym. Eng. Sci., li, 40 (1973).
3. A.H. Cottrell, "The Mechanical Properties of Matter," John
Wiley & Sons, New York, 1964.
4. K. Masubuchi and R.E. Keith, Final Report on "Fundamentals of
Deformation Characteristics of Adhesive-Bonded Joints and
Metal-Adhesive Interfaces," Contract No.
DA-01-021-AMC-14693(Z), Battelle Memorial Institute, Columbus,
1967.
5. G.C. Grimes, Aero. Adhes. and Elast., 249-258 (1970).
6. W.A. Jemian and M.B. Ventrice, J. Adhesion, 1,.190 (1969).
7. W.J. Renton and J.R. Vinson, "The Analysis a;-d Design of
Composite Material Bonded Joints under Static and Fatigue
Loadings,"Contract No. AFOSR-1760-72, University of Delaware,
1973.
8. F.J. Boerio, C.A. Gosselin, R.G. Dillingham, and H.W. Liu, J.
Adhesion, 13,159 (1981).
9. W.T. McCarville and J.P. Bell, J. Appl. Polymer Sci., ~,335
(1974).
10. I. Langmuir, J. Amer. Chem. Soc., 38,2221 (1916).
U. I. Langmuir, J. Amer. Chem. Soc. , 39,1848 (1917).
12. K. Blodgett, J. Amer. Chem. Soc., 56,495 (1934) •
13. I. Langmuir and V. Schaeffer, Chem. Rev. , ~, 181 (1939) •
14. P. Jacquet, Trans. Elect rochem. Soc., ~, 393 (1934) •
STRUCTURAL PRECURSORS TO FRACTURE 827

15. W.A. Jemian, "Adhesion and Bonding of Coatings on Metals," PhD

Thesis, Rensselaer Poly. Inst., Troy, 1956.
16. Y.C. Fung, "Foundations of Solid Mechanics," Prentice Hall,
1965.
17. A.N. Gent, J. Mater. Sci.,~, 925 (1970).
18. T. Alfrey, in "Treatise on Adhesion and Adhesives," R.T..
Patrick, Editor, Vol. 1, Marcel Dekker, New York, 1966.
19. C.B. Buchnall, "Toughened Plastics," Applied Science
Publishers, London, 1977.
20. R.P. Kambour, J. Polymer Sci., D7,1 (1973).
21. S. Rabinowitz and P. Beardmore,ICRC Crit. Revs. Macromol.
Sci., 1 ,1 (1972).
22. A.S. Argon, J.G. Hannoosh and M.M. Salama, Fracture 1977, l,
445 (1977).
23. R.J. Oxborough and P.B. Bowden, Phil. Mag., 28, 547 (1973).
24. J.J. Bikerman, "The Science of Adhesive Joints," Academic
Press, New York, 1961.
25. J.L. Gardon, in "Treatise on Adhesion and Adhesives," R.L.
Patrick, Editor, Vol. 1, Marcel Dekker, New York, 1966.
26. R.W. Bryant and W.A. Dukes, in "Structural Adhesives Bonding,"
M.J. Bodnar, Editor, p 81, Interscience, New York, 1966.
27. W.C. Wake, in "Adhesion," D.D. Eley, Editor, Oxford University
Press, p .191, 1961.
28. W.D. Bascom and R.L. Cottington, J. Adhesion, 7, 333 (1976).
29. J.J. Bike rman , in "Adhesion and Adhesives," R.-Houwink and G.
Salomon, Editors, Vol. 1, 2nd Ed., Elsevier, 1965.
30. J. Friedel, "Dislocations," Pergamon Press, 1964.
31. J. Miklowitz, J. Appl. Mech., Trans, ASME, lQ, 274 (1948).
32. J. Miklowitz, J. Appl. Mech., Trans. ASME, 72, 159 (1950).
33. C.W. MacGreggor, Proc. ASTM, 40, 508 (1940)-.-
34. S.H. Pinner, "Weathering and Degradation of Plastics," Gordon
and Breach, 1966.
35. P. Beardmore and S. Rabinowitz, J. Mater. Sci., lQ, 1763
(1975).
FRACTURE TOUGHNESS OF ELASTOMER MODIFIED EPOXY ADHESIVES

A.A. Donatelli~ C.T. Mooney* and J.C. Bolger**

*University of Lowell, Lowell, MA 01854

**Amicon Corporation, Lexington, MA 02173

Elastomer modification is an effective way of tough-

ening glassy epoxy resins. The toughening mechanisms of
phase separation and gellation have been studied exten-
sively. This paper reports the effects of carboxyl ter-
minated acrylonitrile rubber on several single component
epoxy adhesives, based on Epon 828 and cured with dicy-
andiamide. Variables include accelerator type and level,
cure temperature and elastomer level. There is an opti-
mum amount of rubber which may be added to each system
to yield maximum shear strength. Peel strength is max-
imized in two regions. Rapid gellation of these adhe-
sives may be detrimental to the phase separation and
chain extension mechanism and thereby cause lowering of
fracture toughness and strength properties.

829
830 A.A. DONATELLI ET AL.

INTRODUCTION

Elastomeric modifiers have been used extensively in the adhe-

sive industry to improve the mechanical properties of glassy epoxy
adhesives. Many structural adhesives have benefitted from the
addition of small amounts, ca. 5-20 parts per hundred resin (phr)
of elastomer.

Toughening is achieved by introducing an elastomer such as a

carboxyl terminated copolymer of butadiene and acrylonitrile
("RYCAR CTBN") into a glassy epoxy resin. The elastomer which is
initially dissolved in the resin precipitates to form a second
phase during cure. This second phase appears as small domains
(5000-30000 ~) within the glassy matrix. The formation of this
second phase is dependent on suchlf2ctors as concentration, cure
temperature and rate of reaction. '

The importance of carboxyl termination in the toughening

mechanism stems from the fact that the precipitated phase is not
pure rubber but rather an epoxy terminated liquid rubber. The
following reactions have been proposed to account for this phenom-
enon. 3

o
II /0\ IIC -
"'- C - OR + CR - CR - -+ 0 - CR - CR ,..",. (1)
2 2
I
OR

-- CR......, (2)
I
o OR
I I

Reaction 1 is the chain extension reaction which occurs be-

tween the carboxyl terminated elastomer and the epoxy resin. Re-
action 2 is a cross linking reaction which competes with reaction
1 for epoxide groups.

Although carboxyl termination is an important step in the

toughening mechanism, the molecular weight of the starting resin
also plays an important role. Toughening has been increased by
adding Bisphenol A into liquid epoxy resins which are then modi-
fied with an elastomer. 4 ,5 In theory, the addition of Bisphenol A
causes chain extension which further increases toughness.

In practice, mixtures containing Bisphenol A have unusually

high viscosities and are difficult to handle. The increased
 FRACTURE TOUGHNESS OF EPOXY ADHESIVES 831

strength generally does not justify the handling disadvantages.

EXPERIMENTAL

Several single component epoxy systems were prepared. All

systems studied contained 100 parts of a diglycidyl ether of
Bisphenol A (Epon 828) plus 10 phr dicyandiamide and 2 phr
CAB-O-SIL.

The first variable studied was a to 100 phr elastomer modifier

(HYCAR CTBN x 13). An "optimum" system was then selected and two
catalysts were added. The catalysts studied were 3-(p-chloro-
phenyl)-l,l-dimethylurea and nonyl phenol. Both were varied from
0-8 phr. Systems were cured at temperatures ranging from l20 0 C
to l80 o C.

Performance properties measured were Shore D Hardness, l60 0 C

hot plate gel time ("stroke cure") and two strength tests to mea-
sure adhesive joint fracture; i.e; tensile shear strength per ASTM
D1002 and "T" - peel strength per ASTM D1876.

100 ~--------~------~'--------'--------~--------~--------1

NO ACCELERATOR

160°C CURE TEMP.

75

50
SHORE D
HARDNESS
AT 25°C

25

o
o 20 40 60 80 100 120
RUBBER LEVEL
PARTS PER HUNDRED RESIN

Figure 1. Shore D Hardness versus rubber level.

 832 A.A. DONATELLI ET AL.

RESULTS AND DISCUSSION

The addition of small amounts of elastomer modifier increases

the toughness of glassy epoxy resins. In addition, elastomer mod-
ifiers also have a softening or plasticizing effect. Figure I
shows Shore D hardness measured at 2S o C versus rubber level in an
unaccelerated epoxy resin matrix. A nearly linear relationship
exists between rubber level and hardness. Maximum softening oc-
curs at addition levels above 20 phr. Between 0-20 phr of elas-
tomer a 11.8% decrease in hardness is observed, but between 20-40
phr of elastomer a 24.4% decrease in hardness occurs.

This toughening versus softening effect is best illustrated

in Figures 2 and 3. Figure 2 plots tensile shear strength versus
rubber level for an unaccelerated epoxy resin matrix. These
systems were cured for one hour at 180 0 C and one hour at 160 o C.
Both exhibit the same behavior. A bell shaped curve is observed
with shear strength increasing sharply between 0 and 20 phr rub-
ber. This supports the concept of softening rather than tough-
ening at addition levels above 20 phr. Above this level the
materials lose their rigidity, which is almost as important to
shear strength as the crack stopping imparted by the small rubbery
domains present in cured system with low addition levels of rubber.

421.0 r--------r-------.r-------,-------~------~r_------~
NO ACCELERATOR

210 . 9
TENSILE
SHEAR
STRENGTH
KQ/cm 2

O ~ _______ L_ _ _ _ _ _ _ _ ~ ______ ~ ________ ~ ______ ~~ _ _ _ _ _ __

o 20 40 60 80 100 120
RUBBER LEVEL
PARTS PER HUNDRED RESIN

Figure 2. Tensile shear strength versus rubber level.

 FRACTURE TOUGHNESS OF EPOXY ADHESIVES 833

7 . 16

NO ACCELERATOR

5 . 37
RUBBERY
"Tn PEEL REGION
Kg/em CRACK
STOPPING
3 . 58
REGION

1. 7 9

o ~ ______ ~~ ______-A________-L________ ~ ______ ~~ ______-4

o 120
RUBBER LEVEL
PARTS PER HUNDRED RESIN

Figure 3. "T" peel strength versus rubber level.

Figure 3 presents "T" peel strength plotted aginst rubber

level for the same systems presented in Figure 2. Peel strength
increases between 0 and 20 phr rubber, levels off, decreases to
60 phr then starts to increase again. In the first region at low
addition levels, peel strength is being increased by the crack
stopping effect of the precipitated rubbery domains. Unlike shear
strength, peel strength is not as adversely affected by a soft
material. This is supported by the fact that peel strength in-
creases after the 60 phr addition level. The fracture mode indi-
cates that rigidity is somewhat of a disadvantage and the same
effect can be achieved through the internal strength of the ma-
terial. These data indicate that there is a transition period
from a discontinuous precipitated rubber phase which depends on
crack stopping to achieve its strength, to a continuous rubbery
phase which depends on its internal strength.

Several adhesives possess a good balance of strength proper-

ties. One disadvantage of dicyandiamide is the high temperatures
at which it must be cured. Dicyandiamide has a minimum activation
temperature of approximately ISOoC. To obtain a realistic cure
time, 160 0 C must be used.
834 A.A. DONATELLI ET AL.

60 r-------,--------r------~------_.r_----~

......----aal-------4Q:t-----4t1~--- o - -NONYL
PHENOL
45

;EL TIME
11 NUTf;S

30 GEL TP1E
AT 160'C

15

()-- U.qEA
o ~ ____ ~ ____ ~
----().---
____ ________
~
()-
L __ _ _ _ __4

o 2 4 6 9 10
CATALYST LEVEL
GRAMS PER 10 G~AMS orcy

Figure 4. 160°C hot plate gel time versus catalyst level.

421 . 8 ~----r------~----,------,------~

UREA ACCELE~ATOR
0 180'C
Q 160'C
A 140'C
316 . 35 ., 120'C

4
210 . 9 a
TENSILE A A
SHEAR
STRE GTH
Kg/cm 2
1
V .,
105.45

0 L-_ _ _ _L-_ _ _ ~ _ _ _ _~ _ _ _ ~ _ _ _ ~

o 2 4 6 8 10
CATALYST LEVEL
GRA~S DE~ 10 GP~S OICY.

Figure 5. Tensile shear strength versus urea type accelerator

level, at constant 20 phr rubber level.
FRACTURE TOUGHNESS OF EPOXY ADHESIVES 835

Two accelerators were used in this study, 3-(p-chlorophenyl)

-l,l-dimethylurea and nonyl phenol. Figure 4 shows the effect of
the addition of these accelerators to a system which contained
20 phr of elastomer modifier. The addition of nonyl phenol had
no effect on the l60 0 C hot plate gel time. The urea accelerator
significantly reduced the gel time between 0 and 2 parts of urea
per 10 parts of dicyandiamide, then heading towards a plateau at
the 8 parts addition level indicating that additions of urea above
this level would yield no advantage. It is evident that the acid
catalyst (nonyl phenol) has no accelerating effect whereas the
basic catalyst (urea type) has a significant effect on the dicy-
andiamide cure rate.

The accelerators have a detrimental effect on tensile shear

strength. Figure 5 shows the effect of the urea accelerator level
on shear strength at four different cure temperatures. There is a
large decrease in strength between the control which contains no
accelerator and the adhesive which contains two parts of urea
accelerator. Although there is a large decrease in strength ini-
tially there is no appreciable strength difference between the
2 and 8 parts addition level of the urea catalyst at l80 0 C and
l40 0 C cure temperatures. There is a significant decrease in
strength in the systems when the cure temperature is decreased to
120~C. As with the higher cure temperatures, there is no appre-

7 . 16

UREA ACCELERATO~

5 . 37

3 . 58
" T " PEEL D 180·C
Kq/crn
~----------------~O~160·C

1. 7 9

o
o 2 4 6 8 10
CATALYST LEVEL
GRAI>1S PER 10 r,~A."'S orcy.

Figure 6. "T" peel strength versus urea type accelerator level,

at constant 20 phr rubber level.
836 A.A. DONATELLI ET AL.

421.8
NONYL PHENOL

316 . 35

180'C

210.9
160'
TENSILE
SHEAR
STRENGTH
Kq/em 2

105 . 45

o
o 6 8 10
CATALYST LEVEL
GRA~S PER 10 GRAMS DICY.

Figure 7. Tensile shear strength versus nonyl phenol level, at

constant 20 phr rubber level.

7.16 I I I I

NONYL PHENOL

5 . 37 ~ -

3.58 r-
"T" !'EEL I"l 0
Kg/em ~~~------~D=--------i~--------------~ 180'C

o
1. 79 r o
~------~O~------------------------- 160'C
-

o I I I I
o 2 6 8 10
CATALYST LEVEL
GRN'S PER 10 GRAMS DICY.

Figure 8. "T" peel strength versus nonyl phenol level, at con-

stant 20 phr rubber level.
FRACTURE TOUGHNESS OF EPOXY ADHESIVES 837

ciable strength difference observed between the 2 and 8 parts

addition level of the urea catalyst. The accelerator may in-
terfere with the phase separation and hence the toughening mech-
anism.

The strength loss effect of the urea accelerator is not ob-

served with "T" - peel strength. Figure 6 shows a slight but not
significant increase in peel strength with urea level. As was
shown previously in Figure 3 there are two regions where peel
strength can be maximized, the crack stopping region and the rub-
bery region. If the phase separation mechanism is inhibited by
the presence of an accelerator than a rubbery phase would be pres-
ent in the cured system and the peel strength can be maintained.

Nonyl phenol does not function as an accelerator but it does

have an effect on the strength properties of the system. Figure 7
plots tensile shear strength versus nonyl phenol level. There is
a decrease in strength as the level of nonyl phenol is increased.
It is interesting to note that the data points at 4 phr are well
above the respective curves. It would appear from this, plus the
n on accelerating effect observed in Figure 4, that the nonyl
phenol functions as a plasticizer.

Figure 8 shows the effect of the nonyl phenol on "T" peel

strength. Nonyl phenol has no effect on the peel strength.

CONCLUSIONS

1. Rubber addition reduces hardness, increases flexibility.

2. Tensile shear strength is maximized at ca 20% rubber addition

level.

3. Peel strength is maximized in two regions, the crack stopping

region and the rubbery region.

4. Increased cure temperature (up to l80 0 C) increases all strength

properties.

5. Accelerators may be used to speed cure rate and to permit cure

at as low as l20 0 C to l40 0 C.

a. Urea is a more effective accelerator than nonyl phenol.

b. Nonyl phenol appears to function as a plasticizer.

c. Both accelerators cause a reduction in tensile shear

strength.

d. Accelerators have a negligible effect on peel strength.

838 A.A. DONATELLI ET AL.

REFERENCES

1. F. J. McGarry and J. N. Sultan, Polym. Eng. Sci., !l (1), 29

(1973) •
2. R. S. Drake, E. H. Rowe and A. R. Siebert, Mod. Plast., ~,
110 (1970).
3. A. R. Siebert and C. K. Riew, "The Chemistry of Rubber
Toughened Epoxy Resin r", B. F. Goodrich Research Center,
Brecksville, Ohio, Paper presented at the 161st American
Chemical Society Meeting, (1971).
4. L. T. Manzione and J. K. Gillham, J. Appl. Polym. Sci., ~,
907 (1981).
5. C. B. Bucknall and T. Yoshii, Br. Polym. J., lQ (1), 53
(1978).
6. W. D. Bascom, R. L. Cottington, R. L. Jones and P. Peyser, J.
Appl. Polym. Sci., l!, 2545 (1975).
A THREE-DIMENSIONAL ANALYSIS OF A BUTT JOINT WITH A FLAW

R. S. Alwar* and K. N. Ramachandran Nambisan**

*Applied Mechanics Department, lIT, Madras

600036, India
**Civil Engineering Department, R.E.C., Calicut
673601, India

A three-dimensional finite element method is

employed to analyze an adhesive butt joint with a
through-the-thickness crack in the center of the adhe-
sive layer. The finite element model used accommodates
both the extremely small dimensions of the adhesive
layer and the larger dimensions of the rest of the
region. The crack tip region is modelled by isopara-
metric quarter-point degenerate brick elements which
give the required square root singularity for stresses
at the crack tip. The results from a fracture mechanics
approach corroborate the earlier experimental results
that the strength of an adhesive joint increases with
decrease in adhesive layer thickness. It is also
observed that the stress intensity factor varies in a
non-linear fashion along the crack front.

839
840 R. S. ALWAR AND K. N. RAMACHANDRAN NAMBISAN

INTRODUCTION

Significant savings in weight and fabrication cost can be

realized through the use of adhesive bonding in aircraft struc-
tures. The definition of an adhesive bond encompasses a number of
situations from bonded structural components to fibers bonded
together by the matrix in a fibrous composite. When a composite
is made with an excellent bond between the fibers and the matrix,
it is found to be brittle and have very low toughness value.
Hence the application of the principle of linear elastic fracture
mechanics in the design of bonded structures becomes very im-
portant r Extensive studies have been conducted by Mostovoy and
Ripling on flaw tolerance and crack growth in adhesives under
opening mode. Their studies showed that the cracks tend to grow
in a center of bond (CoB) fashion.

Trantina 2 employed a two-dimensional analysis, using con-

ventional six noded triangular elements, and analyzed a single
edge notched (SEN) specimen. He calculated the compliance rate
for various adhesive layer thicknesses and compared the results
with experimental values.He showed that compliance rate decreased
as the adhesive layer thickness became thinner. This corroborates
the experimental finding from a fracture mechanics point of view
that the bond strength 3 increases with decrease in adhesive layer
thickness. Wang et ale analyzed a Double Cantilever Beam (DCB)
Specimen with an adhesive inter layer by a two-dimensional hybrid
finite element method. They found the stress field close to the
crack tip in the adhesive to be similar to that of a monolithic
system. However, they found no influence of the adhesive layer
thickness on the stress intensity factor in the case of the DCB
specimen. But, for a stretching field, an increase in adhesive
streng~h with decrea~e of adhesive layer thickness was noted by
Gardon and Williams • The same result was reported by many
others who worked wi~h adhesive structural joining. In fact,
Meissner and Baldauf had demonstrated that at very low adhesive
layer thickness, the ultimate tensile strength of a butt joint
can exceed the ultimate tensile strength of the adhesive. A
discussion of the factors which may contribute to the strength of
the adhesive joint 7with decre~se in adhesive thickness was given
by Anderson et ale • Trantina ,using a two-dimensional finite
element analysis, studied the effect of the location of the crack
tip for an inclined crack in a butt joint.

Two-dimensional analysis for an adhesive joint may not be

adequate, especially for large thickness to crack length ratios
and with sudden change of material properties in a region close
to the crack front as in the case of bonded joints. The object of
the present investigation is to analyze a two material (aluminum-
epoxy-aluminum) system in the form of a butt joint using three-
dimensional finite element method and to apply the fracture
mechanics concept to the strength of butt joints.
ANALYSIS OF A BUTT JOINT WITH A FLAW 841

METHOD OF ANALYSIS

The geometry and dimensions of the problem analyzed are

shown in Figure (1). Edge-cracked and center-cracked bonded
joints are analyzed with the crack situated in the middle of the
adhesive layer thickness. In the multi-material body, the stress
intensification near the crack tip is produced from the geometric
discontinuity (crack) and the material discontinuity. The
numerical analysis becomes complicated by the vast disparity in
dimensions between the adherend boundaries and the region across
the adhesive layer thickness where the stress distribution must
be accurately described. The finite element mesh must accommodate
both the small dimensions of the adhesive thickness and the
larger dimensions of the rest of the specimen. It is essential to
model the adhesive layer and the crack tip region accurately with
a number of elements across the adhesive thickness to discern the
effect of the material discontinuity.

Using the symmetry about x-z and x-y planes, only 1/4 of the
plate is considered for solution in the case of edge crack; and
using the additional symmetry about y-z plane, only 1/8 of the
plate is taken for solution in the case of center-cracked plate.

co =1

t---r-- - W: 100
ADHEREND
(ALUMINIUM) Ez ."JZ
L:90 Crack

AOHESIVE LAYER
(EPOXY) E l • Yl
L=90
E2/ E l: ZO
Vz='1'1 = 0 .33
a = 1 0 (Small crack)

ao =1
Figure 1. Geometry of the bonded joint and the coordinate system
at the crack tip.
842 R. S. ALWAR AND K. N. RAMACHANDRAN NAMBISAN

The adhesive layer thickness (=2T) has been represented by eight

layers of elements in the crack tip region and by four layers in
the rest of the region. The crack tip element size is taken as
0.1 with the crack tip element size to crack length ratio of
0.00333 for large crack 9and 0.1 for small crack.16rack tip region
is modelled by singular and transition elements with a proper
grading of the element sizes to avoid large distortions of the
elements. Because of the interaction of the singularity with the
material discontinuity in the immediate neighborhood of the crack
tip, a more refined mesh grading has been adopted in the region
between the crack tip and adherend-adhesive interface. The mesh
pattern adopted for the analysis is shown in Figure (2).

The ratio of the modulus of elasticity of the adherend

(aluminum) to that of the adhesive (epoxy) is taken to be 20 and
the Poisson's ratio v for both the materials is taken as 0.33.
The analysis assumes both adherend and adhesive to be linear
elastic materials.

EXTRACTION OF STRESS INTENSITY FACTORS

The three-dimensional stress and displacement fields in the

close neighborhood of the crack front have the following form.
(See Figure (3) for the crack tip coordinates).

t I
w= 100

~ / ~ /
I

0
~

"
...l

_I!\_/

~T
,
.')); :;((
• . . \ .
~
'm =O.l---J ~
a :30 CRACK TIP L iNTERFACE

Figure 2. Finite element model of the bonded joint (not to

scale).
ANALYSIS OF A BUTT JOINT WITH A FLAW 843

K1 (z) 9 (1 i 8 . ~)
(2r)1/2 cos 2" - s n 2" S1n 2

K2 (z) 8 e 38
(2r)1/2 sin 2" (2+cos 2" cos :2) + 0(1) • • (1)

K1 (z) e e 38
cry cos (1 +sin -2 sin -2
(2r)1/2 2
. 8 e 38
S1n 2 cos 2" cos:2 + 0(1) • • (2)

crz v (cr + cr ) •• (3)

x Y
K1 (z) 8 8 38
(2r)1/2 sin 2 cos 2" cos :2

K2 (z) e e 38
+ /
cos - (I-sin - sin - ) + 0(1) •• (4)
(2r) 1 2 2 2 2
K3 (z) e
T
(2r)1/2 sin 2 + 0(1) •• (5)
xz

K3 (z) 8
T
(2r)1/2 cos 2 + 0(1) •• (6)
yz

4Gu = K1 (z) (2r) 1/2 cos 2"e (I<: -

2 e
1+2sin 2)
1/2 8 2e
+ K2 (z)(2r) sin 2" (I<: +1+2cos 2") +0 (r) . • • (7)

4Gv K1 (z) (2r)

1/2 e 2 8
sin 2 (I<: +l-2cos 2)

- KZ (z){2r)
1/2
cos 2"8 (I<: -1-2sin
2
2"e )+0 (r) •• (8)

Gw K3 (z) (2r)
1/2
sin 2"e + O(r) •• (9)

where I<: = 3-4v for plane strain

3-v
= l+v for generalized plane stress

and G = Modulus of rigidity.

844 R. S. ALWAR AND K. N. RAMACHANDRAN NAMBISAN

v ,v

Crack front

x,u

z,w

Figure 3. Coordinate system near the crack front.

11
An exact 3-D solution incorporates K = 3-4v and hence
this value has been used in the present analysis.

The finite element values of displacements or stresses are

matched with the elastic singular solutions in the vicinity of
the crack tip (Equations 1-9), yielding the stress intensity
factors (SIF) as a function of distance from the crack tip.
Extrapolation to the crack tip, using the exact stress or dis-
placement expressions, yields accurate values for the stress
intensity factors. Stresses are calculated at two consecutive
gauss points very close to the crack tip and lying on a radial
line from the crack front, using a 10xl0xl0 gauss point rule.
Substituting the computed values of cr in Equation (2) and
extrapolating using square root singu!ar behavior of the
stresses, the value of SIF at r = 0 is obtained. Alternatively
the crack opening displacements (COD) at the quarter-node and
end-node of the crack tip element may be used to calculate the
stress intensity factor~ Kl 1
and Kl 2 respectively using
Equation (8) and the extrap& ated value of the stress intensity
factors Kl at r=O is given by Kl = 2Kl 1_- Kl 2. The evaluation
of Kl in this manner incorporateS the 'rr dis~Iacement variation
near the crack tip and eliminates the rigid body motion and
linear terms from the displacements. The Kl variation along the
crack front is obtained by calculating the continuous displace-
ment field on the traction-free surface of the crack tip element
from the element nodal displacement values and the shape func-
tions of the element.

RESULTS AND DISCUSSION

A study of the accuracy of the present analysis has been

made by comparing the results with those of the two-dimensional
ANALYSIS OF A BUTT JOINT WITH A FLAW 845

2
analysis by Trantina • The comparison has been made for the case
of very large crack length for which 2-D analysis may be accurate
enough. For small crack length with large h/a value, the problem
should be treated as a three-dimensional onel2 • Though the two-
dimensional analysis by Trantina was done using conventional
elements to model the crack tip region without singular and
transition elements, the application of energy principle to get
the compliance derivative is expected to give accurate results.
Moreover, experimental verification of the numerical results had
also been obtained in the above analysis. Figure (4) shows the
comparison of the present 3-D results with those obtained by
Trantina. The present analysis gives SIF values which are not
uniform along the crack front. The maximum value of the SIF is
obtained in the mid-plane of the plate and the value decreases as
we proceed towards the free surface of the plate. It may be
observed from the figure that there is reasonable agreement
between the 3-D results and the 2-D results by Trantina. The 3-D
mid-plane values are found to be higher than the 2-D values by
8-12 percent. It may be observed that 3-D values are higher than
the corresponding 2-D values in the case of monolithic plates.

18,-------------------------------------,
1 6 2 0 Anal ys Is - Trantlna

o 3 .0 Analysis Max valucz (middlcz)

14 • 3 0 AnalYS I S Mm . vol ucz (surfa~)

a l w =0 , 3

1 2
o

1.0

~
o
-: 0 ,8
o
"0
u
"0 0.6

04

0 .2

T / W ----

Figure 4. Compliance derivative dc/da as a function of adhesive

thickness.
846 R. S. ALWAR AND K. N. RAMACHANDRAN NAMBISAN

0 .5
I '"
"" 1\
0 .4
'\,
I
\' \

t '\ \
\\
0. 3
V Thickn<lsS of sp~c'men =10
"0 =1.0
I- It t
or.
r ..1--- 100 _ \I
N
I
0 . 2 I- 90..
7
ADHEREND
I
1e
Qdhl! S1 ~
- ~8 II -x
- 0=1.0 T =1.0
I
1

lI
0.1 - - - SINGLE LAVER
o O.F 1035 I
I
- - - TWO L AVERS
- 00 F 1632
I

o I I I I I
0.4 0.6

Figure 5. Convergence study for SIF variation along the crack

front.

Figure (5) shows the convergence of SIF variation in the

interior of a center-cracked plate for single and two layer
solutions for a/W = T/W - 0.01. Since 'a' is of the same order of
magnitude as the adhesive layer thickness and less compared to
the specimen thickness, the crack-tip stress distribution will be
three-dimensional in nature. Convergence study could not be done
with more number of layers of elements across the specimen thick-
ness because of the prohibitive computer time involved in such an
analysis. The experience in analyzing the monolithic plates with
a center crack has shown that two and three layer solutions agree
well for the SIF variation throughout the thickness of the plate.
It is seen from the figure that in the interior of the plate up
to z/h = 0.2 from the mid-plane, the agreement between one and
two layer solutions is excellent. The reduction towards the
surface for the two layer solution compared to one layer solution
is about 6 percent. The results presented in this paper are
corresponding to single layer solutions and hence the values in
ANALYSIS OF A BUTT JOINT WITH A FLAW 847

the mid-plane region are expected to be accurate, whereas the

surface values are expected to be higher than the actual by about
6-10 percent. From the two-layer solution, the decrease of SIF at
the surface compared to the mid-plane value is about 16 percent.
A decrease of SIF value at the surface, of the order ?3 la-30
percent, is reported in the case of monolithic plates ' •

Figure (6) gives the 3-D results for the stress intensity
factor as a function of adhesive layer thickness for a large
crack length (a = 30.0). It can be seen from the figure that the
value of SIF decreases with the decrease in adhesive layer
thickness which corroborates the earlier experimental results
that the strength of an adhesive joint increases with the
decrease in adhesive layer thickness.

Figure (7) shows the mid-plane SIF variation as a function

of adhesive layer thickness for the case of a very small crack
length (a = 1.0). It is observed that the stress intensity factor
reduces with decrease in adhesive layer thickness for this case
also. Comparing both Figures (6) and (7), it can be seen that as
h/a increases K1/crQ i:a at~o increases which is true for the
case of a monolith1c plate •

Figures (8-11) show the stress distribution across the plate

thickness in the neighborhood of the crack front for points lying
around the crack tip. It can be seen from these figures that the
transverse shear stresses are extremely small compared to the
other stresses. It may be observed that for bending and stretch-
ing loads, these shear stresses are bounded and since the points
selected are close to the axis of symmetry, the transverse shears
have to be very small. Figure (8) shows that cr ~ 0y at points
very close to the line of crack extension (ie.xat e = 0.74°) and
that the condition of pseudo plane strain, i.e., 0 = v( 0 + 0 )
is also satisfied. The pseudo plane strain conditi~n is fo~nd t6
exist at all the pOints in the neighborhood of the crarf front.
This is in accordance with the 3-D analytical solution that
a = a along the line of crack extension and that a pseudo plane
s~rainYcondition exists in the close vicinity of the crack front.
The figures also show that the values of 0y increases from e = 0°
to approximately e = 60° and then gradually decreases to zero .
value at e = 180°. The value of 0 decreases from e = 0° to
approximately e = 75 ° and then slows an upward trend up to e =
135° and then drops to zero at e = 180°. The value of Txy
increases up to e = 30° and then decreases to assume a maximum
negative value at e = 100° and then drops to zero at e = 180°.
The above e~ariation of stresses follow the theoretical e-
variation as represented in reference 16. This indicates that a
correct ~variation of stresses in the neighborhood of the crack
front is obtained with the finite element mesh pattern adopted in
the analysis.
848 R. S. ALWAR AND K. N. RAMACHANDRAN NAMBISAN
0.6,-- - - - - -- - -- - - - - - - -----,

0 .5

~-

---
.---- --
~- -
-~ ,.. -

......... --

o Edg2 crock mid p lana: valuQ:S

• Edgcz crock su rfoc<z ",oluQS

0 3 Il Cczn lrl2 Cfock midplon<r: valUCZ5

.. Cczn trcz croc k sur facCl 'I,IoluCl:S

o / w.O 3

hlo .0 33

T/W

Figure 6. KIlo 0 Ii. as a function of adhesive thickness for a/W

= 0.3.
o 1r------------~

06
Edgq cracko:d
h 10.10 .0
W .100 .0
I!: L .900
T -0.6,10,1.4
W
Q .1 .0
,z-
0 .5

O· t LO-05...l..-....L.--lL-.l.-'::'
0 .",
0"'10,-l--L---'-.....".O,:015
.
T/ W

Figure 7. K1/0;.a (mid-plane) as a function of adhesive

thickness for a7W =0.01.
ANALYSIS OF A BUTT JOINT WITH A FLAW 849

0.5.---~----.---------------------'-----------~'----------'

0.4

Tyz

.t::.
N

0.2
r= 0.000017

Q = 0.74748

0.1 Edge crack

T/W= 0.014
a/w =0. 3

Figure 8. Stress distribution a/a near the crack front at r =

0.000017 and e = 0.74748° - edge gracked plate.

0.5,--.--.-----.-----~._--._----------------------~----------,

0.4

.t::.
N

r = 0.0000239
0.2
Q = 45. 37627°

Tyz TlW= 0.014 Edge crack

a/W = 0.3
0.1

Figure 9. Stress distribution a / a 0 near the crack front at r

0.0000239 and e = 45.376° - edge cracked plate.
850 R. S. ALWAR AND K. N. RAMACHANDRAN NAMBISAN

\
0.5

\ \ 1\
0.4
,

'Txy 'Tyz -r xz OX ozl

I
I
\ \OY
--
.c
N

0.2 ,
I
r = 0.000017
I 9 = 90.74746·

0.1 T!W= 0.014

I
a/W = 0.3

o
-0.6 -0.4 -0.2 o 0.2 0.4 0.6 0.8 1.0 1.2

Figure 10. Stress distribution a/a near the crack front at r

o
0.000017 and e = 90.74746° - edge cracked plate.

0.5
I \

0.4 I--
"1 rr/ \
0.3 -

.c
Txy vC'z "ax
N l-
0.2 I--
'ryz? r = 0.0000239

9= 135.3736"

T/W=0.014
o. lr.-
a/w =0. 3

r EDGE CRACK

o I I I I I I I
-0.8 -0.6 -0.4 -0.2 o 0.2 0.4 0.6 O.B 1.0 1.2

Figure 11. Stress distribution a/a near the crack front at r

o
0.0000239 and e = 135.3736° - edge cracked plate.
ANALYSIS OF A BUTT JOINT WITH A FLAW 851

Figure (12) compares the stress-based and COD-based values

of stress intensity factors across the plate thickness and it is
seen that there is good agreement between the two results, the
maximum difference being about 3 percent.

CONCLUSIONS

The 3-D analysis gives results which are 8-12 percent higher
compared to 2-D values in the interior portion of the adhesive.

The SIF is uniform in the central region of the adhesive and

decreases towards the free surface. This reduction is about 16
percent for the case studied.

For an adhesively bonded joint the value of SIF decreases

with decrease in adhesive layer thickness. This corroborates the
experimental observations that the strength of a bonded joint
increases with decrease in adhesive layer thickness.
o5

\
\-\
-

0.4
Vt 00
t f f t
r~ r--- 1OO
o3 0 "

1:
a>

t:.
.... I
N

0.2
~ o .~
T: 0 . 6
- x

- COD BASED VALUES

• STRESS BASED VALUES
GAUSS PO I T CO - OROI ATES FOR
STRESSES
0 .1 r: 0 ,000017
9 : 90 . 75'

-
•
o
0 . 36 0 ,4
•
0.5
I I I 0 .6
KJ! "0 ra -
Figure 12. Comparison of stress-based and COD-based SIF values
across plate thickness.
852 R. S. ALWAR AND K. N. RAMACHANDRAN NAMBISAN

APPENDIX

NOTATION

2a Crack length

E Modulus of elasticity

G Shear modulus
h Plate thickness

Kl ' Kll , KIll Stress intensity factors

2W Width of Plate

2T Thickness of adhesive layer

u,v,w Displacement components in X,Y,Z

direction

r, e Polar coordinates

v Poisson's ratio

REFERENCES

1. S. Mostovoy and E. J. Ripling, J. Appl. Polymer Sci., lQ,

1351 (1966).
2. G. G. Trantina, J. Composite Mater., 6, 192 (1972).
3. S. S. Wang, J. F. Mandell and F. J. McGarry, Int. J.'
Fracture, 14, 39 (1978).
4. J. L. Gardon, in "Treatise on Adhesion and Adhesives", R.
L. Patrick, Editor, Vol. 1, pp. 269-324, ¥.arcel Dekker, New
York, 1967.
5. M. L. Williams, J. Appl. Polym. Sci., 14, 1121 (1970).
6. H. P. Meissner and G. H. Baldauf, Tran~ Soc. Amer. Mech.
Engr., 697 (1951).
7. G. P. Anderson, K. L. DeVries and M. L. Williams, Int. J.
Fracture, 10, 565 (1974).
8. G. G. Trantina, J. Composite Mater., 6, 371 (1972).
9. R. S. Barsoum, Int. J. Numerical Meth. Eng., 11, 85 (1977).
10. P. P. Lynn and A. R. Ingraffea, Int. J. Numerical Meth.
Eng., .,!l, 1031 (1978).
ANALYSIS OF A BUTT JOINT WITH A FLAW 853

11. G. c. Sih, M. L. Williams and J. L. Swedlow, "Three-

dimensional stress distribution near a sharp crack in a
plate of finite thickness", Air Force Materials Lab.,
Wright-Patterson Air Force Base, AFML-TR-66-242 (1966).
12. G. C. Sih and R. J. Hartranft, Int. J. Fract., 7, 39 (1971).
13. P. D. Hibitt, in "Mechanics of Fracture, Plates-and Shells
with Cracks", G. C. Sih, Editor, Noordhoff Int. Pub.,
Leyden (1977).
14. I. S. Raju and J. C. Newman, Jr., "Three-dimensional finite
element analysis of finite thickness fracture specimens",
NASA-TN-D 8414, May (1977).
15. G. C. Sih and R. J. Hartranft, Int. J. Fract., 9, 75 (1973).
16. R. S. Barsoum, Int. J. Numerical Meth. Eng., 10~ 551 (1976).
THE INFLUENCE OF LAYER THICKNESS AND INTERNAL STRESSES ON THE BOND

STRENGTH OF METAL-TO-CERAMIC JOINTS

w. Diem, G. Elssner, T. Suga and G. Petzow

Max-Planck-Institut ~Ur Metallforschung

D-7000 Stuttgart, W. Germany

Sandwich-like joints between Zr,Zr-Nb-Zr or Hf foils

of different thickness and rectangular bodies of hot-
pressed silicon nitride (HPSN) were fabricated by solid-
state bonding at 1150 °c or 1200 DC. Bend bars of
30x5x2 mm were cut from the welded pieces and notches
of variable depth a were introduced at the metal-to-
ceramic interface. The bond strength was described by
critical energy release rates using experimentally deter-
mined correction functions and correction functions cal-
culated by means of the FEM method. The ef~ects of lay-
er thickness and crack position on the correction func-
tions of materials joints are discussed from a more ge-
neral point of view by a comparison with the correction
functions of isotropic materials and bimaterials. Mea-
surements of the critical energy release rates o~ the
metal-to-ceramic joints and microstructural investiga-
tions indicate that the reduction of internal stresses
by microcrack ~ormation and by allotropic transformation
of the metal component control together with the metal
layer thickness the bond strength o~ the sandwich-like
joints. The bend test method is also applicable to adhe-
sively bonded joints. The in~luence of the adhesive thick-
ness and of the crack position on the correction function
can be evaluated according to the principles described in
thjs paper.

855
856 W. DIEM ET AL.

INTRODUCTION

In most of the materials joints thermal stresses are produced

during the manufacturing process. They may lead to a reduction in
bond quality. Thermal stresses promote slow crack growth and the
formation of microcracks in the interface region. Their magnitude
is not only affected by the differences in the elastic and thermal
properties of the bonded materials but depend also on the thickness
of intermediate layers.

Solid-state bonded metal-to-ceramic joints consisting of two

bodies of hot-pressed silicon nitride (HPSN) and an intermediate
foil of a refractory metal were chosen to study the influence of
layer thickneRs and of an allotropic phase transformation on the
bond strength 1 •

The bond strength of the metal-to-ceramic joints was character-

ized by the critical energy release rate GC obtained by a fracture
mechanics testing method. The critical release rate GC and the frac-
ture resistance KC are closely related to each other. Both can be
taken as a measure of the bond strength of metal-to-ceramic joints
and of semi-brittle joints. Generally,the energy release rate G and
the stress intensity factor K are described by the relations 2

G = (1-B 2 )K2 /E* (1)

G = Ko 2yG/E* (2)

K2= K 2y /(1-B 2 ) (3)

o G

where S is a Dundur's composite parameter given by the elastic pro-

perties of the two materials adjacent to the crack front, E* the
effective modulus of elasticity, and K a constant proportional to
the appli~d stresses. Y is a so-calleg correction function which
depends on the crack leggth at the interface, the elastic proper-
ties of the bulk materials of the joint, and the specimen geometry.
For bend test specimens the correction function YG can be ob- 2
tained experimentally by compliance measurements using the equation

YG = dC*/d(a/W) (4)

where C* is a normalized compliance, a the crack or notch depth, and

W the height of the specimen.

To get true data of the critical energy release rate G the de-
pendence of the correction function YG on both the ratio a/~ and the
layer thickness d must be determined. If different materials combi-
INFLUENCE OF LAYER THICKNESS AND INTERNAL STRESSES 857

nations are investigated this procedure will be rather tedious. How-

ever, some combinations as refractory metal/ceramic joints show re-
latively small differences in the elastic properties of their con-
stituents when compared to others as for example adhesive joints.
Therefore, it may be useful to estimate beforehand the maximal de-
viation of the correction function YG from the wellknown correction
function Y. of isotropic materials. This feature is discussed from
1SO. .. .
a more general p01nt of V1ew 1n the next sect1on.

THE CORRECTION FUNCTION AND LAYER THICKNESS EFFECTS

The correction function YG for a crack at the interface of a

materials joint depends on the elastic properties of all materials
of the joint. Consequently for a joint consisting only of the two
materials 1 and 2 the stresses under in-plane deformations and the 2
correction function YG can be described by the Dundurs' parameters
0. 12 and B12 •

Consider the case that a crack is situated at an interface be-

tween the two materials A and B of an intermediate multilayer of
thickness d between the materials 1 and 2 (Figure 1). The thickness
d is negli~ibly small compared to the specimen size. If the elastic
p~operties of the materials A and B are known, a simple relation can
be derived from considerations of the continuity of the energy re-
lease rate during continuous changes of the materials geometry and
the crack tip position.

d /W
m
« 1

where a. .. is a Dundurs' parameter related to the materials i and j

and Y.. 1J the correction function ~G for an interfacial crack speci-
fied 1J by the materials i and j. Tfiis relation stands also for a
diffusion layer, if its thickness d is small compared to the speci-
men size and its elastic propertiesmat the crack tip are known.

In cases where the crack runs irregularly ~hrough a thin inter-

mediate multilayer of thickness d (Figure 2) and a specification
of the elastic properties at the ~rack tip is not possible the cor-
rection function Y12 can only be calculated. In Figure 2 two other
possible fracture paths are shown where the cracks are running
parallel to the interface into the materials 1 and 2. They may be
designated as cohesive failures. When the layer thickness d is small
~ompared.with the specimen height W the. correction function YG of
1nterfaC1al cracks and cracks near the 1nterface can be related to
the correction Y. for isotropic specimens using one composite
param~ter 0. 12 fofs~he description of the elastic properties of
mater1als 1 and 2.
858 W. DIEM ET AL.

Figure 1. Cracks in intermediate layers of thickness d between

the two materials 1 and 2. m

I d.JW<I ~
dm d dm

f---....L'-1..'-='-,--------, h IW
0- 0 O·

Figure 2. Fracture paths within an intermediate layer and parallel

to the interfaces in the materials 1 and 2.
INFLUENCE OF LAYER THICKNESS AND INTERNAL STRESSES 859

A more general view of the relationship between the isotropic

correction function Y. and the correction functions of materials
joints is given in Fi~~~e 3. Thin intermediate layers which may exist
between the materials 1 and 2 are omitted in the sketches of the
specimens. Yb ·, Y1 ' and Y2 denote the correction functions YG for an
interfacial cfack in a simple bimaterial or for cracks in the mate-
rials 1 or 2, respectively. As can be seen, YG for interfacial cracks
at variable ratios d/W is bounded by the correction functions
Yb.(d/W + 00) and (1+a)Y. (d/W + 0). If the influence of stress con-
cefltrations at the bond~aOedges is not large the value of Yb . will
.
be sl~ghtly smaller than the value 0 f Y. • For a center crac~k ~n .
an infinite bonded plane the relation Yb~~o> 0.842 y. holds, e. g.
Yb · = 0.97 4 Y. for glass / epoxy and Yb . ~ = 0.99 4 Y. ~so for HPSN / Hf
joints. At cofl~~ant layer thickness and~variable r~t~os h/W (h =
distance of the crack from the interface) the correction function
YG changes its value discontinuously from Y12 /(1+a) over Y12 to
Y12 /(1-a) when the crack position shifts from material 1 to material
2. With increasing distance of the crack from the interface the cor-
rection function Y12 /(1+a) will be substituted by Y1 and finally by
Y. if material 1 ~s considered or Y12 /(1-a) will be substituted
byS~2 if the crack is situated in an ~ntermediate layer material 2
between two bodies of material 1.

EXPERIMENTAL
Sandwich-like joints consisting of two hot-pressed silicon
nitride (HPSN) bodies and an intermediate laye 3 of a refractory
metal were manufactured by solid-state bonding • Zr and Hf foils
and Zr-Nb-Zr triple layers so different thickness ranging from
0.125 to 2.0 mm were chosen as intermediate metal layers. The HPSN/
Zr/HPSN and the HPSN/Zr,Nb,Zr/HPSN joints were solid-state bonded
at 1150 °e and the HPSN/Hf/HPSN joints at 1200 °e. The metal and
ceramic parts were welded toghether under a pressure of 10 N/mm 2
in a high vacuum. Some properties and the composition of the ma-
terials used for the solid-state bonded joints are given in Table I.

Three-point bend test specimens of the dimensions 30x5x2 mm

were cut from the solid-state bonded materials. Notches of width 50
to 100 ~m and of depth a were introduced into the metal-to-ceramic
interface (Figure 4) by a precision cutting device. The bend tests
were carried out with a model 1474 Zwick machine at a cross-head
speed of 1 ~m/min. In the set-up load and deflection signals were
treated in real time by a directly connected microprocessor. To ob-
tain true load-deflection curves and compliance data, the measured
deflections were corrected for the deformation of the test device
during loading.
 CD
0)
o

- - - - cohesive failure in moterial 2 -+ interfacial cohesive failure in material I

I failure

a. denotes a.12

..,Materiall
~Material 2
(f-:!JYiSO (I· (llYn .Yiiso• • •
o o
•
I Y2 I
d/W d/W

j Y~o YiSO 1
oo~ 00

1 v", v, Inlv" ~, : YI 1
d/W ~ LJI] I. c:::=-:::::::J )1 d/W
I~IV"• LI-J
i I - C1
I Yiso ( I - 0 IYiso I (~IYiso I
o c=c:=J c=c:=J: Lr=:J o
) -CID. h/W 0- 0 I 0+ h/W • +CID I
Figure 3. Correction ~unctions o~ materials joints as ~unctions o~ the ~
normalized crack position h/W,and the ratio o~ layer thick- o
m
ness to specimen height d/W. s::
m
-I
>
['""
INFLUENCE OF LAYER THICKNESS AND INTERNAL STRESSES 861

Table 1. Materials used for solid-state bonded part s •

Material Composition (w /0) E(GPa) V aTe 1Q-6/K)

Zr 99.8 Zr,0.1 0,0.06 Fe, 99 0.33 7.3

0.01 Cr,N,H
Hf 97.0 Hf,2.8 Zr,0.015 0, 141 0.30 6.0
0.01 Fe,Ta,C,N
Nb 99.8 Nb,0.05 Ta,0.03 W, 108 0.38 8.4
0.01 Zr,Ti,Fe
HPSN 97.0 Si3N4,0.98Mg, 314 0.28 2.7
0.14 Ca,Y 20 3 ,Si 2ON 2

E is the Young's modulus, V the Poisson ratio, and aT the thermal

expansion coefficient.

NUMERICAL ANALYSIS

The stress analysis of a materials joint with thin intermediate

metal 'layers requires a fine mesh of finite elements for the speci-
men. The present calculations were performed with about 800 quadra-
tic isoparametric elements under the condition of plane strain. For
the crack tip eight quarter-point singnl~r elements of length
0.005 W were employed. The correction functions were calculated by
means of the virtual extension and the J-integral method. The cal-
culated results for four-point bend test specimens of an isotropic4
material differ within + 1 %from the wellknown analytical values
(Figure 5). Figure 6 shows a mesh of finite elements used for the
stress analysis of a HPSN/Hf/HPSN three-point bend test specimen.

RESULTS AND DISCUSSION

In Figure 7 the normalized compliance C* and the corresponding

correction function YG of HPSN/Hf/HPSN jointb are plotted against
the ratio of crack depth to specimen height a/W for different ratios
of metal layer thickness to specimen height d/W. The curves are cal-
culated by means of the finite elements method. The dots represent
experimental data. Minor deviations between the calculated and the
experimental values seem to be due to the small specimen size and
the assumption of a sharp crack in the FEM calcul~tions. For
HPSN/Hf/HPSN joints the composite parameters a and Bare -0.366 and
0.098 respectively. As previously mentioned, the correction func-
tions are bounded by two extreme curves: the curve of the correction
function Ybi near the curve for the isotropic case and the curve of
YG = (1+a)Y. for d/W = O.
J.so
862 w. DIEM ET AL.

HPSN

Hf
notch

Figure 4. SEM picture of a notch cut into the interface of a

HPSN/Hf/HPSN joint.

3r-------------------------~

_
-j? 01--- __0 /.:
j2 ,--_o-
s' -1
~
o vIrtual t'Xt_on
-2
o J. ,ntegal

o 0.2 OA 0.6
a/W
Figure 5. Comparison of the calculated correction function YG (calc)
for an isotropic four-point bend test specimen with the
correction function YG obtained by Gross and Srawley •
 INFLUENCE OF LAYER THICKNESS AND INTERNAL STRESSES 863

l, 28mm 1

1
E

.
'"_i

i d,0.125,.".c
I
d ,n'~

d ,nI:J.<.
HPSN I HI I HPSN d, 1.0mm- r--
d: 2.0mm

Figure 6. FEM mesh used for the stress analysis of a HPSN/Hf/HPSN

three-point bend test specimen. Heavy lines indicate the
integral paths.

~r---------------------------,
N
HPSN/Hf/HPSN HPSN/Hf/HPSN

..otTO c
IW.oo

:;::~ ;~
dJW .0 135
•u
d1W,0.025
dJW.O

CSl'--_ _-

CSl

0 .0 0 .2 0.4 0.6 0 .0 0 .2 0.4 0 .6

a/W a/W

Figure 7. Normalized compliance C* and correction function YG

versus ratio a/W for HPSN/Hf/HPSN joints. Three-polnt
bend test. Dots are measured and lines calculated values.
864 W. DIEM ET AL.

Figure 8 shows experimental data of the critical energy release

rate GC of a HPSN/Hf/HPSN joint with a Hf layer of 0.125 mm thick-
ness as a function of the ratio a/We The observed independence of
the GC data on the ratio a/W demonstrates the applicability of the
correction function YG•

Metal-to-ceramic joints fabricated by solid-state bonding at

1150 to 1200 °c from bulk HPSN pieces with an intermediate Zr or Hf
layer or a triple layer Zr-Nb-Zr are characterized by a thin reac-
tion layer of about 5 ~ thickness which is composed of silicides,
nitrides 5and ternary metal-silicon-nitrogen compounds of the metals
Zr of Hf • Furthermore, a very thin Ca- and Si-oxide rich glassy
layer develops during the welding process on the HPSN interface
which may lead to a reduction of bond strength. The formation of
these glassy layers is caused by the hot-pressing aids added to the
HPSN material. To separate the influence of layer thickness and allo-
tropic transformation of the metal component different model combi-
nations were chosen. Triple intermediate layers Zr-Nb-Zr consist of
a rigid niobium foil of variable thickness coated on both sides by
Zr foils of 0.125 mm thickness. They cause a thickness-independent
intimate contact of the mating ceramic and metal interfaces during
welding which is only achieved by the plastic deformation of the Zr
layers without any visible deformation of the Nb foil or the HPSN
parts. HPSN/Zr,Hf,Zr/HPSN joints are characterized by the same type
of reaction layers independent of the total metal layer thickness.
Nb and Zr form solid solutions at 1150 DC. Therefore, no reaction
layers develop at the Zr/Nb interfaces. HPSN/Zr/HPSN joints of
different metal layer thickness and, for comparison, HPSN/Hf/HPSN
joints were used to study the influence of an allotropic transforma-
tion of the metal on the bond strength of the joint. Zr transforms
during cooling down from the welding temperature from a cubic body-
centered B-structure to a hexagonal close-packed a-structure. The
transformation of the pure metal at 862°C is accompanied by an
elongation of about 0.1 %. Hafnium as another metal of the group IV A
shows a B/a-transformation at 1750 DC. Therefore welding at 1200 °c
and cooling down is not influenced by a phase transformation. How-
ever, both metals form the same types of reaction layers because of
their chemical similarity.

GC data of HPSN/Zr,Nb,Hf/HPSN, HPSN/Hf/HPSN, and HPSN/Zr/HPSN

joints were evaluated from the fracture load FC' the specimen geo-
metry, and the correction functions YG according to the relation 2
2
2 9 e 'IT YG
G = F • (6)
C C B2W3 E*

where 2e is the lower span length in the case of three-point bend

testing, B the width, and W the height of the specimen.
INFLUENCE OF LAYER THICKNESS AND INTERNAL STRESSES 865

HPSNJHfI HPSN
HI : d.o.12Smm

150

100
N
__-=-_e_ •-T--' ---.--
E •
...,

50

OL-__~____~__~~__~~
o 0.2

Figure 8. Critical energy release rate GC of a HPSN/Hf/HPSN joint

with a Hf layer of 0.125 mm thlckness as a function of
the ratio a / W.

I
• HPS I Zr I If'SN
& HPSN I HII HPSN

300 o HPSN I Zr.NII.Zr l HPSN

tdtr • O'2Smm 1

t
I
200

c.J
(!)

6/ .!. - - - .- - - - - - _ 0
,
,
,
. I

0' ,
... I

O~ ____ ~ ____ ~ ____ ~~

Figure 9. GC data of refractory metal/HPSN joints as a

function of the intermediate metal layer thickness d.
866 W. DIEM ET AL.

Figure 9 shows the experimentally determined GC data of the

joints as a function of the layer thickness d. The plotted data re-
present mean values obtained from 10 to 15 measurements per layer
thickness. The scatter is ~ 10 %. For HPSN/Hf/HPSN and HPSN/Zr,Nb,Zr/
HPSN joints the critical energy release rate GC decreases from
94 J/m 2 at d = 0.125 mm to 54 J/m 2 at d = 0.6 mm and increases again
to a value of 107 J/m 2 at d = 1 mm. For layer thickness of d > 1 mm
up to d = 3.25 mm nearly the same G values were observed. HPSN/Zr/
HPSN joints show for Zr layers of tte thickness 0.055, 0.385, and
0.64 mm a GC value of 100 J/m 2 • From d = 0.64 mm to d z = 1.0 mm
a steep increase to a Gc value of 328 ~7m2 occurs and a~ra thick-
ness d Zr = 1.56 mm the GC value is reduced to 244 J/m 2 •

During cooling down from the welding temperature considerable

thermal stresses are built up in the interface region due to the
differences in the thermal expansion coefficients of the ceramic
material and the refractory metals (Table I). They lead to the for-
mation of microcracks perpendicular to the interface. Figures 10
and 11 show two examples of microcracks which terminate at the
a-Zr/reaction layer interface (HPSN/Zr/HPSN, d = 0.125 mm) or at
the reaction layer/HPSN interface (HPSN/zr,Nb,~f/HPSN, d= 0.6 mm).
For HPSN/Zr,Nb,Zr/HPSN and HPSN/Hf/HPSN joints the lengths of the
cracks increase with increasing metal layer thickness from 1 ~m
at d = 0.125 mm to about 30 ~m at d = 3.25 mm and the tips of the
microcracks shift with increasing metal layer thickness from the
metal/reaction layer interface over the reaction layer/HPSN inter-
face into the adjacent regions of the ceramic materials. Similar
observations were made with HPSN/Zr/HPSN joints. The length of the
microcracks increase from about 1 ~m at d z = 0.125 mm to about
4 ~m at d = 1.0 mm. However, the tips ofrthe microcracks were
always si~frated at the metal/reaction layer interface independently
of the thickness of the Zr layer. Fractographic investigations
showed that the failure of all three types of materials joints
occured predominantly in the interfacial regions where the micro-
cracks terminate. This observation and the dependence of the micro-
crack position on the metal layer thickness can be used to explain
the dependence of the bond strength data on the layer thickness.
For the HPSN/Hf/HPSN and the HPSN/Zr,Nb,Zr/HPSN joints the mini-
mum in bond strength at a metal layer thickness of about 0.6 mm
(Figure 9) is due to a fracture at the interface between the re-
action layer and the HPSN. This interface is rather weak caused by
the already mentioned formation of a thin glassy layer. At a metal
layer thickness of 1.0 mm microcracks surmount the reaction layer/
HPSN interface and enter the ceramic part of the joint. Thereby a
stress relief of the interface region occurs which leads to an
increase in bond strength. HPSN/Zr/HPSN joints show up to a metal
layer thickness of 0.645 mm a constant GC value of 100 J/m 2 and a
constant fracture behaviour. The fracture of the specimens occurs
in the metal/reaction layer interface. The steep increase in bond
strength for specimens with a Zr layer thickness of 1.0 mm can be
INFLUENCE OF LAYER THICKNESS AND INTERNAL STRESSES 867

a-Zr

reaction
layer

HPSN

Figure 10. Microcrack terminating at the interface between a-Zr

and reaction layer. HPSN/Zr/HPSN joint with d Z =0.125 mm.
SEM picture. r

a-Zr

reaction
layer

HPSN

Figure 11. Microcrack terminating at the reaction layer/HPSN

interface. HPSN/Zr,Nb,Zr/HPSN joint with a metal
layer thickness of 0.6 mm. SEM picture.
868 W. DIEM ET AL.

explained by the existence of rather short microcracks terminating

again at the metal/reaction layer interface. The rather weak re-
action layer/HPSN interface is not influenced by microcracks. At
the critical load the crack propagates into the reaction layer al-
ternating its fracture path from the reaction layer/HPSN interface
to the reaction layer/Zr interface. The high GC value of 328 J/m 2
seems to be due to these irregularities in the fracture behaviour.

Table II. Microcrack densities N in solid-state bonded HPSN/metal

joints.
Intermediate d N A NoA /:,T/:'arr /:'T
metal layer (nm) ( 1/mm) (jJ.m) ( 10-3 ) ( 10-3 ) (OC)

0.35 17 0.15 2.6 4.0

Zr,Nb,Zr 0.6 31 0.15 4.7 5.0 1120
1.25 35 0.15 5.3 5.5
0.125 36 0.12 4.3 4.0
0.385 34 0.12 4.1 4.0
Hf 0.64 33 0.12 4.0 4.0 1170
1.0 30 0.12 3.6 4.0
2.0 32 0.12 3.8 4.0

Zr 0.385 13 0.12 1.5 1.7* 830*

1.0 14 0.12 1.7 1. 7*
** Because of the plasticity of B-Zr in the temperature range
1150 to 862°C /:'T = 830 K is assumed. The elongation of 0.1%
during the Bfa transformation is subtracted from the llTlla T
value.

The magnitude of thermal stresses in materials joints can be

estimated by the strain mismatch /:,~o/:'T where /:,~ is the difference
between the expansion coefficients of the metal and the ceramic and
/:'T is the difference between welding temperature and room temper-
ature. On the other hand, the reduction of internal stresses by
microcrack formation is given by the product NoA where N is the
number of microcrack per unit length and A the mean opening of the
microcracks. A was determined by SEM measurements at a magnifi-
cation of 30 000:1. Table II shows a comparison between NoA and
/:,T/:,aT • As can be seen the agreement between the two values is good
for all materials combinations investigated. This leads to the
conclusion that the thermal stresses are nearly completely elimi-
nated by microcrack formation. The reduction of internal stesses
INFLUENCE OF LAYER THICKNESS AND INTERNAL STRESSES 869

by the B/a transformation of the Zr layer in a HPSN/Zr/HPSN joint

is also demonstrated. The measureded NeA values of HPSN/Zr!HPSN
jOints are smaller by a factor of more than 2 than the NeA values
of HPSN/Hf/HPSN and HPSN/Zr,Nb,Zr/HPSN joints.

CONCLUDING REMARKS

GC measurements by means of single edge notched bend test

specimens require a precise determination of correction functions
YG• If the elastic properties of the materials adjacent to an
interfacial crack in a metal-to-ceramic or in an adhesively bonded
joint are known the somewhat tedious experimentai determination
can be replaced by a calculation of the correction function.
Following this procedure accurate bond strength determinations are
feasible both for metal-to-ceramic and adhesively bonded joints.
Some relations between the correction function Y. for isotropic
materials and the correction function Yb . for a B!gaterial can be
derived, if the dependence of the correction function Ya on the
thickness of intermediate layers and on the crack positlon is
considered.

The results obtained here on the influence of layer thickness

and phase transformation on the bond strength of joints between
hot-pressed silicon nitride and refractory metals indicate that
the critical energy release rate GC is controlled by the position
of the cracks in the interfacial region and the reduction of in-
ternal stresses. Rather high bond strength values can be obtained
by the applications of metal layers which undergo a stress reducing
phase transformation.

ACKNOWLEDGEMENT

Support of this work by the Deutsche Forschungsgemeinschaft

is gratefully acknowledged.

REFERENCES

1. G. Elssner, W. Diem, and J.S. Wallace in "Surface and Inter-

faces in Ceramic and Ceramic-Metal Systems", J. Pask and
A. Evans, Editors, p. 629, Plenum Publishing Corporation,
New York, 1981
2. W. Diem, G. Elssner, T. Suga, and G. Petzow, these proceedings
these proceedings, pp. 87l-882
3. G. Elsqner and U. Krohn, Z. Metallkde. 70, 71 (1979)
4. B. Gross and J.E. Srawley, NASA TN D-2603 (1965)
5. H. Ncwotny, B. Lux, and H. Kudielka, Mh. Chemie 87, 447 (1956)
BOND STRENGTH CHARACTERIZATION OF METAL-TO-CERAMIC AND ADHESIVE

JOINTS BY CRITICAL ENERGY RELEASE RATES

w. Diem, G. Elssner, T. Suga, and G. Petzow

Max-Planck-Institut fUr Metallforschung

D-7000 Stuttgart, W. Germany

Small three- or four-point bend test specimens of

30 to 60 mm length were used to characterize the bond
strength of metal-to-ceramic and adhesive joints by cri-
tical energy release rates. The testing procedure in-
volves the determination of correction functions which
describe the influence of the geometry and the elastic
properties of the joint. The correction functions are
obtained by compliance measurements or can be calculated
by means of the finite element method. A fracture re-
sistance parameter KC derived from the critical energy
release rate is introduced. KC allows a comparison of
the bond strength of different joints without consider-
ing differences in their elastic properties. Examples
of a bond strength determination of glass/epoxy joints,
and of the influence of crystal orientation on the
strength of sapphire/Cu joints serve to illustrate the
utility of this bend test method to determine critical
energy release rates of semi-brittle materials joints.

871
872 w. DIEM ET AL.
INTRODUCTION

Modern technology has placed increasing demands on the strength

properties of metal-to-ceramic and adhesive joints. The soundness of
these types of joints must be assured to guarantee the structure in
which they are incorporated. Tests to determine the bond strength
properties are, therefore, of major importance for the quality as-
sessment and development of composite structures.

Conventional bend and tensile tests or peel tests provide bond

strength data which are functions of the specimen size and the test
method employed. However, in practice and for a better understanding
of the bonding behaviour of joints, strength data are needed as a
function of fabrication parameters and parameters such as chemical
and mechanical compatibility, intrinsic materials properties and
local characteristics of the interface region such as flaws and un-
bonded areas. This may be provided by fracture mechanics test me-
thods yielding values which describe the fracture resistance or
the critical energy release rate of the joints. The present paper
reports on the determination of the bond quality of specimens
notched exactly at the metal-to-ceramic or the adherend-adhesive
interface in terms of critical energy release rates using correc-
tion functions experimentally or numerically determined in three-
or four-point bend tests.

CRITICAL ENERGY RELEASE RATE

One general method of stUdying semi-brittle fracture employs

energy rate considerations which are based on a global energy ba-
lance for creating new surfaces under the assumption of an autonomy
of the fracture process (YF = 0)

where UE is the internal energy, UK the kinetic energy, YF the frac-

ture surface (or interfacial) energy incorporating dissipated energy
associated with crack extension, A the crack surface area, Q the
heat flux, and Ws the work due to applied surface and body forces.
The dots denote differentiation with respect to time. The energy re-
lease rate G is defined by
d
G = --
dA
(W S - UE ) (2)

The critical energy release rate GC corresponds to the onset of un-

stable crack extension. It represents the fracture surface or inter-
facial energy
BOND STRENGTH AND CRITICAL ENERGY RELEASE RATES 873

when the load is applied slowly and kinetic and thermal effects are
negligible.

Some continuum mechanics models for an interfacial crack in

ideally bonded elastic materials have been developed 1- 2 which allows
us to establish a correlation between G and the stress intensity
faktor K. Choosing the definition 3
. . -(X( 1)-1) .
K = KI-l KII = 12n llmr+or (Oe- l 're)e=O (4)

where A(l) is the first eigenvalue of the equation for the local
stress fleld corresponding to the chosen crack tip model. Kr and
KII are the stress intensity factors for model I and II,. the local
stress field can be described independeJatly of the elastic proper-
ties by the stress intensity factor K:
X -1
(Oe-i 're)e=O = (Kjlan)r (1) (5)

In the models mentioned above the energy release G is related to the

stress intensity factor K2 = Kf + KII by
= 1-B2 K2 (6)
G E*
where B is a composite parameter (see equation 8) and E* an effecti-
ve modulus of elasticity given by
1 K1+1 K2+1
E*
= -16 ( -
JJ.
- + --)
JJ. 2
1
where JJ., (j=1,2) are the shear moduli of the materials j adjacent
to the interface, K. = 3-4 v. (plane strain), K.=(3- V.)/(1+v.)
(plane stress) and J v . the P~isson ratios. J J J
J
The elastic behaviour of the two bonded materials can be ex-
pressed by the Dundurs' parameters a and B4,
k(K 1+1)-(K 2+1)
a = k(K
1+1)+(K 2 +1)
(8)
k(K -1)-(K -1)
B _ 1 2
- k(K 1+1)+(K 2 +1)

where k = JJ. 2 /JJ. 1 •

Although up to now the mechanisms of interfacial separation

have not been completely understood and the description of this type
of failure by a continuum mechanics model is not completely war-
ranted, it seems appropriate to utilize both parameters G and K for
the characterization of the bond quality of materials joints.
874 w. DIEM ET AL.
For engineering applications the specification of bond quality
by critical stress intensity factors or fracture resistance values
Kc may be preferred because these parameters allow a comparison of
strength between different materials joints and even between joints
and bulk materials. However, for joints which do not fracture
exactly at the interface, or for materials combinations with no
distinct interface, it is necessary to specify the chosen Dundurs'
parameter a and 8 and the effective modulus E* to avoid misinter-
pretations of the given KC data.

We used for the investigation of the bond strength of metal-

to-ceramic joints three- and four-point bend test specimens with
lengths of 30 to 60 mm. The sandwich-like specimen consist of two
rectangular ceramic bodies connected by a thin layer of metal and
were notched exactly at the metal/ceramic interface. These relative-
ly small specimens of simple geometry5 were chosen because of the
scarcity of material, their better fabricability and their adapta-
bility for high temperature tests. For comparison, glass/epoxy
specimens of similar size were tested.

THE CORRECTION FUNCTIONS

The energy release rate and the stress intensity factor depend
not only on the applied load and the elastic properties but also on
the specimen geometry. For material joints the influence of both
the specimen geometry and the elastic properties of the constituents
is described by the so-called correction functions YG. Consequently,
the energy release rate G and the stress intensity factor K can be
expressed by

G = (K 2/E*}Y
o G
and

(10)

where. K0 is a constant for a given specimen proportional to the

applJ.ed stress.

The energy rate G for three- and four-point bend test specimens
J.s described (see Figure 1) by3
YG(a/W}
G = F2 9 e 2 7f
B2 W3 ( 11)
E*

for a given materials combination with constant thickness d of the

intermediate layer. A precise determination of GC requires an accu-
BOND STRENGTH AND CRITICAL ENERGY RELEASE RATES 875

rate correction function. YG can be experimentally obtained by

measurements of the compliance C according to the well known relation
F2 dC
G=-
2B
-
da
( 12)
By comparison of Equation (11) and (12), it is seen that YG can be
described by

( 13)

where C* is a normalized compliance

(14)

and Co is the compliance of the unnotched specimens. The correction

functions YG can also be determined by means of the finite element
method.

F FI2 F/2
- 4lII+-- - e

Ge = Fe' ( 3t~ )-t.

Figure 1 Bend test configurations for GC measurements.
876 w. DIEM ET AL.

EXPERIMENTAL

Soda-lime glass/epoxy/soda-lime glass joints of adhesive thick-

ness 0.1 to 3 mm were prepared. The adhesive employed was an unmodi-
fied epoxy resin of DGEBA type (DER 332) mixed with 10 phr TETA
curing agent cured 18 h at room temperature and postcured 4 h at
60°C. Some glass and epoxy specimens were also prepared for investi-
gating their bulk properties. Sapphire/Cu and A1203/Cu joints were
fabricated by an eutectic bonding process 6 • Polycrystalline A1203_
platelets and sapphire platelets of the orientation [0001) and (0110)
and of 1.4 mm thickness were joined via Cu layers of 0.26 mm thick-
ness to two pieces of 14 mm length. Some properties and the compo-
sition of the materials are given in Table I.

Table I. Materials properties and composition.

Material Composition E v
(w/o) (GPa)

soda-lime 72 Si0 2 ,14.5 Na20, 70.3 0.24 9

glass
8.5 CaO, 3.5 MgO,
1.5 A1 20 3

unmodified epoxy equiv.172-178, 3.9 0.34 40

epoxy reSln 10 phr TETA
DGEBA curing: 18h/RT+
4h/60 °c

sapphire[0001] 99.98 A1 20 3 + 461 8.8

[0110J Cr,Ni,Sn,Mn 425 7.9
polycryst. 97.0 A1 20 3 , 350 0.25 8.2
Al203 2.9 Si0 2

E is the Young's modulus, V the Poisson ratio, and aT the thermal

expansion coefficient.

Three- and four-point bend test specimens of the dimensions

60x10x10 mm (glass/epoxy) and 30x5x2 mm (alumina/Cu) were cut from
the as-received composite materials. Notches of width 50 to 100 ~m
and of depth a were introduced into the interface or into the glass
and epoxy parts of the adhesive joints by a precision cutting device.
The notch in the epoxy was sharpened by tapping with a razor blade.
The bend tests were carried out with a Model 1474 Zwick machine at
a cross-head speed of 1 ~m/min. In the set-up load and deflection
signals were treated in real time by a directly connected micro-
BOND STRENGTH AND CRITICAL ENERGY RELEASE RATES 877

processor. To obtain time load-deflection curves and compliance

data, the measured deflections are corrected for the deformation of
the test device during loading. Calculations of the correction func-
tions by 1he finite element method were performed using the virtual
extension and the J-integral methods. The results obtained by these
two methods differ only by ~ 1 %.

RESULTS AND DISCUSSION

Figure 2 shows the dependence of the normalized compliance C*

and the correction function YG on the ratio a/W for different layer
thickness to specimen height ratios d/W of glass/epoxy/glass four-
point bend test specimens. As can be seen, experimental data agree
well with the calculated results.

In Figure 3 experimental data for the fracture resistance KC or

~IC for a glass/epoxy joint with d = 1 rom are plotted against the
distance h of the notch from the interface. The filled circles de-
note apparent values obtained by means of the isotropic correction
function Yis0 8 and the open circles are the corrected data obtained
by means of the correction function YG for the adhesive joint tested.
The corrected KIC data for the epoxy and glass regions agree very
well with the data obtained by fracture tests on the individual bulk
materials. The interfacial failure of glass/epoxy joint was charac-
terized by stable crack extension. KC data of 0.24 MN{m 3 / 2 at. a
crack velocity of 1.7.10- 3 m/s were obtained by using the correction
function shown in Figure 2. As demonstrated in Figure 3 it is gene-
rally not advisable to use isotropic correction functions for the
determination of KC data for joints. In the given example the differ-
ence between the apparent and the corrected value of the fracture
resistance of the joint is about 54 %.

GC measurements were used to study the influence of the crystal

orientation on the bond strength of A1203/Cu joints. Specimens with
a polycrystalline alumina, a sapphire (0001), and a sapphire [0110]
interface in contact with polycrystalline copper were tested. After
the determination of the correction functions YG' bond strength
measurements were carried out at 77 K to minimize plastic deformation
of the metal layers. However, the sapphire/Cu joints showed only an
interfacial fracture if the ratio a/W was larger than 0.4. For these
rather deeply notched specimens, a plastic deformation of copper
could not be avoided as can be seen in Figure 4 by the steep increase
of the measured GC values with a/W ratio. GC data free from contri-
butions of plastic deformation were obtained by an extrapolation of
the measured GC values to a a/W ratio of 0.39 following the trace of
curve for polycrystalline alumina-copper joints as shown in Figure 4
by the two dotted lines. Typical fracture surfaces of the sapphire/Cu
specimens are given in Figures 5 and 6. The copper side (Figure 5)
is characterized by numerous semispherical pores whose borderlines
878 W. DIEM ET AL.

•
IS)
Vl
Nr---------------------------,
9a ss/epoxylglass glass/epoxy/glass

M
IS)

Vl

N
IS)

o IS)
.
U
>' .

IS)
Vl
IS)

IS)

IS) IS)

IS)

-FEN
• • xpw.tn«tt

0 .0 0.2 0 .4 0 .6 0 0 0 .2 0 .4 o 6
a/W a/W

Figure 2. Normalized compliance C* and correction function Y

versus ratio a/W for glass/epoxy joints. F0ur-poin~
bend test. Dots are measured and lines calculated values.

l::: f -t-+
~ _t/t
_---1
--
u

~+--
gloss !epoxy/glass
£0.5 • apparent value

II (I- I .
o corrected YOlue

--l!l;L
I)
o ~-1------~0------~1-------2
I I I
~----~3--J

h [mm! • hom nolch~ S>pKlfrltnio

Figure 3. Comparison of apparent and corrected values ofKC and Kr

as a function of notch location h for a glass/epoxy join~
with d = 1 mm.
BOND STRENGTH AND CRITICAL ENERGY RELEASE RATES 879

f·
160 r - - - - - - - - - - - - - - z - - ,

/A -~/_I
_ poIycr AI,o,/Cu
e sapphire IOllOl/Cu
sapphire IOOO1l!Cu

J
A

_120t-

~
.;'80 .... - - - - - - - -

40 t--r-_ _ _ -=-- _-_-----"-,,1;-- Ie

0----"1-- I
o 0,2 0,4 0,6
a/W

Figure 4. GC data for A1203/Cu joints measured at 77 K as a function

of aN ratio.

are covered with copper oxide particles. More detailed investiga-

tions of the fracture surfaces lead to the assumption that a CuAl0 2
spinel of 0.1 to 1~m thickness is also formed between co~per and
sapphire. Table II shows the extrapolated GC data for the joints
corrected for the interfacial porosity in comparison to the fracture
surface ~nergies 2YF of single crystal and polycrystalline alumina.
The experimentally estimated data indicate a strong influence of the
crystal orientation on the bond strength of the joints. Furthermore,
they compare well with the fracture surface energy data for bulk
alumina and sapphire. Therefore, it can be concluded that the inter-
facial bonding forces of copper/alumina joints are similar to those
in bulk alumina. This may be explained by an intermediate thin
spinel-like layer between alumina and copper.

Table II. Critical energy release rates GC of copper-alumina joints

and fracture surface energies 2YF of alumina.

Alumina GC Interfacial 2YF References

(J/m 2 ) Porosity(%) (J/m2)

polycrystalline 35 9 43 10
sapphire [0110] 12 4 13.8 11 , 12
sapphire (0001) 93 13 102 11 , 12, 13
880 W. DIEM ET Al.

Figure 5. Copper side of an interfacial fracture of a sapphire/Cu

joint tested at 77 K. SEM picture.

Figure 6. Sapphire side of an interfacial fracture of a

sapphire/Cu joint tested at 77 K. SEM picture.
BOND STRENGTH AND CRITICAL ENERGY RELEASE RATES 881

Table III. Critical energy release rates GIC and GC and fracture
resistance data KIC and KC for some bulk materials
and materials joints.

Materials G1C [J/m 2 ) KIC[MN/m3/2 ) e" [GPa)

Al20:! (2.9%Si0 2) 33 3.5 373.3

sapphire [0110 ] 13.8 2.6 21 501.2 21

HPSN 89 5.5 340.7

soda-lime glass 5.2 0.62 74.6

amine cured epoxy
IDER 332 .10 phr TETA ) 61 0.52 4.4
plasma - sprayed
Zr02 coating 26 1.2 52.8

Materials Joints Gc [J/m 2] Kc[MN/m 3J2 ] e"[GPa] a. ~

AI 20 3 (2.9% Si 02)/Nb 32.8 3.5 183.4 -0.509 -0.064

sapphire [0110] I Cu 1211 1.6 21 227.121 -0.546 -0.144

HPSN / Hf 91 4.5 216.1 -0.366 -0.098

soda -lime glass I
amine cured epoxy 6.6 31 0.24 8.33 -0.888 -0.210

plasma - spray.d
NiCrAI Y I Cr/Zr02 coating 16 1.1 76.8 -0.458 -0.081
on steel

1) extrapolated value
2) treated as an isotropic materials joint
3) stable crack extension

Small-sized specimens of a variety of materials combinations

and bulk materials were tested. Table III summarizes some measured
data. The GC data for adhesive joints and metal-to-ceramic joints
are of the order of 10 to 100 J/m 2 • The KC data rank from rather
low values below 1 MN/m 3 / 2 for glass/epoxy joints to relatively
high values above 4 MN/m 3 / 2 for special metal-to-ceramic joints.

CONCLUDING REMARKS

Although the micromechanical processes in the interfacial

region of a materials joint are highly complex, its critical energy
release rate GC can be measured in a similar manner as for isotropic
materials. The necessary correction functions can be computed if
the elastic properties are known or if only thin intermediate layers
of unknown properties are present; they could also be derived from
experimental compliance measurements.
882 W. DIEM ET AL.

The determination of critical energy release rates gives. a more

scientific footing for bond strength measurements than conventional
testing methods. Small-sized specimens cut from brittle or semi-
brittle materials joints can be used to characterize local varia-
tions of bond quality and to describe the influence of microstruc-
ture in the interfacial region, the influence of fabrication para-
meters, intermediate layer effects, and crystal orientation effects.

ACKNOWLEDGEMENT

We thank the Deutsche Forschungsgemeinschaft for the financial

support of some portions of the work described here.

REFERENCES

1. J.R. Rice and G.C. Sih, J. Appl. Mech. 32, 418 (1965)
2. M. Comninou, J. Appl. Mech. 44, 631 (1977)
3. T. Suga and G. Elssner, unpublished results
4. J. Dundurs, J. Appl. Mech. 36, 650 (1969)
5. R.F. Pabst and G. Elssner, J. Materials Sci. 15,188 (1980)
6. J.C. Driscoll and Y.E. Sun, Proc. Int. Microelectronics
Symp., Oct. 1976, p. 44, Vancouver, Canada
7. T.K. Hellen, Int. J. Num. Mech. Engng. 9, 187 (1975)
8. B. Gross and J.E. Srawley, NASA TN D-2603 (1965)
9. W. Diem, Doctoral Thesis, University of Stuttgart (1982)
10. W. Kromp and R.F. Pabst, Z. Materialprufung 22, 241 (1980)
11. P.F. Becher, J. Am. Ceram. Soc. 22, 59 (1976-)-
12. A. Krell, M. Kikuchi and T. Nishino, phys. stat. sol. (a) 52,
K45 (1979) --
13. S. Morozumi, M. Kikuchi and T. Nishino, J. Mat. Sci. 1£, 2137
(1981)
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS

A.C. Moloney*, H.H. Kausch* and H.R. Stieger**

*Laboratoire de Polymeres
Swiss Federal Institute of Technology
CH-I007 Lausanne, Switzerland
**Brown Boveri & Cie.
CH-8050 Zurich-Oerlikon, Switzerland

The fracture properties of two commercial epoxide

resins both unfilled and filled with varying volume frac-
tion of silica, alumina and dolomite particles have been
investigated. The variation in the stress intensity
factor with the crack velocity was measured using the
double torsion test technique. In order to examine the
influence of the resin-filler adhesion on the fracture
toughness, alumina particles were treated with three si-
lane compounds. In addition the yield stresses and the
flexural strengths were measured. In an attempt to si-
mulate resin-filler interactions and to measure the frac-
ture energy for interfacial separation a scarf joint was
used with a bond angle of 45 o. This geometry produces
a mixture of mode I and mode II loading. The combined
mode strain energy release rate (G(I,II)c) was deter-
mined.

883
884 A.C. MOLONEY ET AL.

INTRODUCTION

Particulate fillers are incorporated into plastic components

principally because of their lower cost. However, in certain cases
mechanical, thermal and electrical properties may be improved, no-
tably the elastic modulus, heat deflection temperature and the arc
resistance.

Considerable work has been reported in the literature l on

the effect of particulate fillers on the mechanical properties of
thermoplastic and thermosetting polymers but the underlying mecha-
nisms are still unclear. Since the cost of plastics continues to
rise at a rate faster than that of mineral fillers, it is of great
importance that fillers and surface treatments be chosen on a more
rational basis for specific applications.

In this work we have investigated the fracture properties of

two commercial epoxide resins filled with silica, alumina and dolo-
mite particles. The principal parameters which have been varied
are volume fraction, filler surface treatment, mechanical resistan-
ce of the filler and particle size.

MATERIALS AND EXPERIMENTAL

Fracture of Unfilled Resins and Composites

Two commercial epoxide resins wer~ used in this investigation.

Restin A (solid at room temperature) was based on the diglycidyl
ether of bisphenol-A, cured with phthalic anhydride. Resin B was
an epoxide resin based on dimethyl hydantoin and was also anhydride
cured. After thorough mixing and degassing, the resins were poured
into pre-heated steel moulds which were rotated to prevent sedi-
mentation of the filler. The curing conditions were followed in
accordance with the manufacturers instructions (Ciba-Geigy). Alu-
mina, silica and dolomite particles were used as filler materials,
the salient physical properties are compared in Table I.

Silica particles pre-treated with y-glycidoxypropyltrimetho-

xysilane (A187 from Union Carbide) are available from Quarzwerke.
Alumina particles, however, were treated in our laboratories.
Three silane compounds were used in order to change the adhesion
between the phases. The first hexamethyldisilazane (HMDS) would
be expected to reduce the adhesion 2 • The other two compounds were
y-aminopropyltriethoxysilane (AIIOO) and A187 from Union Carbide.
 Z
-I
m
::IJ
"'Tl
»
n
:;
r
"'tJ
::IJ
Table I P.roperties of Filler Materials.
o"'tJ
m
::IJ
Filler Specific Form of particles Mean particle size Young's Fracture Work of :::!
m
gravity modulus toughness fracture en
o"'Tl
gcm- 3 GPa 2 G (Jm- 2)
KIc MNm-3/
Ie !!
r
r
m
Alumina 3.97 irregular-rounded 'V 6 'IJlD. 320(Ref.4) 5.3 (Ref. 7) 40 (Ref. 8) o
,
m
edges "'tJ
o
X
Silica 2.65 irregular-sharp W 10 'V 60 \.1m 94 (Ref. 5) 0.8(Ref.6) 4.4(Ref.6) om
edges ::IJ
W 6'V 100 \.1m m
en
Z
W 4'V 160 \.1m en
W 1 'V 300 \.1m

Dolomite 2.85 irregular-sharp 20 11m 78(Ref.5) - -

edges
----

00
00
C11
886 A.C. MOLONEY ET AL.

These would be expected to improve the resin-filler adhesion. The

filler particles were treated using a method based on that suggest-
ed by Trachte and DiBenedett0 3 •

The double torsion test technique was used to measure the

stress intensity factors of these composites with varying volume
fraction of filler. Certain authors 9- ll have questioned the vali-
dity of this method due to the curved crack front which is produced.
It has been shown recentlylO that depending on the dimensions of
the double torsion specimen a correction factor may be necessary to
give an accurate value for KIc' The dimensions chosen in this work
minimised the correction to within experimental scatter. The dif-
ficulty of machining these highly filled resins prevented the use
of another geometry. (In a few selected cases, comparative values
were obtained by using single edge notch specimens). The stress
intensity factors were measured using both the constant displace-
ment rate and the load relaxation methods 7 •

The relationship between KIc and the applied load (P) is

shown below:

(1)

where Wn is the moment arm, v Poisson's ratio, W the specimen width,

t the specimen thickness, tn the thickness in the plane of the crack
and kl a geometric factor.

The crack velocities were measured by a method developed in

our laboratories by Stalder et al l2 employing the change in the in-
verse of resistance with tim;-oTa thin layer of graphite applied
to the specimen by means of a spray.

The two sections of the double torsion specimen after failure

were used to measure the flexural strength by a three point bending
test. Three faces of the specimen were polished to eliminate sur-
face flaws. For resin B filled with 30 % untreated alumina and
30 % alumina treated with A187, the tensile strengths were also
measured.

The yield stresses of the composites (Oy) were determined

using the plane strain compression test developed by Williams and
Ford 13 • Employment of this test permitted the use of the failed
halves of the double torsion specimens thereby enabling KIc and 0y
to be compared for the same sample.
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS 887

Combined Mode Adhesion Test

In order to simulate the resin-filler interactions and measure

the fracture energy for interfacial separation, a modified version
of the combined mode adhesive test specimen, analysed by Trantina 14 ,
was employed. This geometry has been used to study epoxide-alumi-
nium adhesive bonds 15 ,16. However, for our purposes it was desi-
rable to investigate the interactions between alumina ceramic and
epoxide resin. It was impractical to fabricate such a specimen
from alumina so instead a block of the ceramic was inserted into a
steel support as shown in Figure 1.

Combin ed mode adhesive test spec imen.

Modification
i
5tHI 0
support

(~ramic
insf'rt

o
t- 32mm.-I

!p
Figure 1. Scarf joint for mixed mode adhesive fracture.

Blocks of alumina ceramic type AL 23 (obtained from Degussa S.A.)

were machined with diamond cutting tools to the required dimensions.
The alumina was treated by degreasing with trichloroethylene in a
Soxhlet apparatus. The steel supports were grit blasted and the
888 A.C. MOLONEY ET AL.

2D

1.5

1.0.

0.5

o 0.2 0..4 0..6

Figure 2. The variation in the stress intensity factor with the

volume fraction of filler.
Resin A oA1 20 3 (double torsion ~T); eSi02(DT); ~Si02+A187; + Si02
(single edge notch SEN).

treated alumina bonded into the support using a secondary adhesive

Ciba Geigy AV138/HV998. The specimens were assembled in aluminium
jigs to ensure correct alignment of the adherends. Resin A was
used as the adhesive in these tests. After curing, a razor blade
was used as a wedge to initiate a sharp pre-crack into the speci-
men. Since resin A is transparent, the initial crack length could
easily be measured prior to testing. The specimens were loaded in
tension in an Instron testing machine at a constant crosshead
displacement rate of 0.1 mm/min, the temperature being controlled
to 23 0C ± 2 °C. The fracture surfaces were analysed by the scan-
ning electron microscope.
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS 889

2.0
•

1.0

0.5

o 01 0.4 0,6

Figure 3. The variation in the stress intensity factor with the

volume fraction of filler.
Resin B oAl 0 (DT)· .siO (DT); ~SiO +Al87; +Si0 2 (sin g le edge notch
2 3 ' 2 2
SEN).

RESULTS AND DISCUSSION

The relationship between KIc and the volume fraction of fil-

ler was found to be linear for both resin systems as shown in Fi-
gures 2 and 3. For a given resin the data for silica and alumina
could be superimposed. This result is surprising firstly, because
of the dissimilarity in particle size of the two fillers and se-
condly, because of the discrepancy in elastic moduli of these two
filler materials (see Table I). Thirdly, the appearance of the
fracture surfaces under the SEM was completely different. The si-
lica particles were well-bonded to the resin, whereas each alumina
particle was clearly de-bonded from the resin (see Figures 4 and 5).
These results strongly suggest that for these irregularly shaped
and relatively "strong" particles neither the particle size at a
constant volume fraction of filler nor the filler-particle adhesion
greatly affects the toughness.
890 A.C. MOLONEY ET AL.

Figure 4. Scanning electron micrograph of the fracture surface of

silica filled resin B.

Figure 5. Scanning electron micrograph of the fracture surface of

alumina filled resin B.
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS 891

In order to investigate the former parameter, silica was

chosen since this is available in a wide range of particle sizes.
The mean particle size was varied between 300 ~m and 60 ~m. As can
be seen from Table II the stress intensity factor is little affect-
ed at a constant volume fraction of filler.

Table II The Influence of Filler Particle Size on the Stress In-

tensity Factor.

Resin B + 40 % Si0 2

Grade of Quartz Mean particle Mean value

size of
(Quarzwerke) ~m Klc (MN m- 3/ 2)

W 1 'V 300 1. 76 ± 0.01

W 4 'V 160 1. 74 ± 0.02
W 6 'V 100 1.87 ± 0.01
WlO 'V 60 1.83 ± 0.02

To determine the effect of resin-particle adhesion on the

fracture toughness, alumina filled resins were chosen. For both
resin systems, treatment with the silanes AllOO and A187 increased
the toughness as compared with the untreated material and treat-
ment with HMDS slightly decreased the values. However, the dif-
ferences were not large and compared with the unfilled resin are
negligible. For example, for unfilled resin B the value of Klc
was 0.6 NM m- 3 / 2 and with 30 % by volume of filler this increased
to between 1.46 and 1.67 irrespective of the filler surface treat-
ment (Table III). Scanning electron micrographs of the fracture
surfaces are shown in Figures 6 to 8. Treatment of the filler
with HMDS resulted in a fracture surface very similar to that of
the untreated filler (see Figure 5). Treatment of the alumina
with A187 improved markedly the adhesion between the resin and the
filler, as demonstrated in Figure 7. For the alumina treated with
the silane AllOO some particles were well-bonded and others poorly
bonded, as shown in Figure 8.

Broutman and Sahu 17 and Hammond and Quayle 18 have also in-
vestigated the effects of surface treatment of glass bead filled
polyester resins. In spite of the fact that they used similar re-
sins and fillers, their results were contradictory as to which
892 A.C. MOLONEY ET AL.

Table III Stress Intensity Factor of Composites Filled with Treated

and Untreated Fillers.
Resin A

Filler Treatment Mean value Klc

-3/2
MN m

30 % Alumina none 1.86 ± 0.01

HMOS 1.80 ± 0.01
A187 1.97 ± 0.04
AllOO 1.99 ± 0.04

Resin B

Filler Treatment Mean value Klc

-3/2
MN m

30 % Alumina none 1.53 ± 0.02

HMOS 1.46 ± 0.04
A187 1.64 ± 0.01
AllOO 1.67 ± 0.02

system was the toughest. We have also conducted some tests on

glass bead filled composites1 9 and found the properties to be some-
what different to those of the irregularly shaped silica and alumina
particles.

For the silica and alumina filled resins, the presence of

filler particles provokes a considerable increase in toughness which
is independent of the resin-filler adhesion. There are a number of
possible mechanisms for this phenomenon. Firstly, the toughness
may be increased by the filler diverting the crack and causing a
larger surface area of fracture. However, the increase in area is
insufficient to account for the increase in toughness. Secondly,
in the case of some filled systems, using for example metallic or
rubber particles 20 , energy may be absorbed by deformation of the
filler, thus causing increases in toughness. This mechanism is not
feasible for brittle glass and ceramic fillers. Thirdly, the in-
crease in toughness may arise from an increased plastic deformation
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS 893

Figure 6. Scanning electron micrograph of the fracture surface of

resin A filled with alumina treated with HMDS.

Figure 7. Scanning electron micrograph of the fracture surface of

resin B filled with alumina treated with A187.
894 A.C. MOLONEY ET AL.

Figure 8. Scanning electron micrograph of the fracture surface of

resin A filled with alumina treated with AllOO.

of the matrix. Fourthly, the augmentation may be due to the ob-

stacles pinning the crack and causing the crack front to bow out
between the particles.

This latter explanation was put forward by Lange 2l who pro-

posed that the increase in fracture energy was proportional to the
inverse of the interparticle spacing (2c). He proposed an equation
of the form:

T
( 2)
Ycomp = Yres + 2c

where Ycomp and Yres are the fracture energies of the composite and
pure resin respectively and T the line tension effect. However,
this analysis is not directly applicable as T is a function of the
particle size (2r). For a penny-shaped crack, Lange deduced that
the line tension was:

T = 2r ( ) ( 3)
3 Yres
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS 895

The ratio of the particle size to the particle spacing is proportio-

nal to the volume fraction Ve by Vp(l-Vp)-l. Thus, Lange's equa-
tion predicts a linear relat~onship between the fracture energy and
the ratio of r to c. However, this linear relationship has not been
observed in practice except for the case of a ceramic composite,
glass filled alumina 22 • For particulate filled plastics such a li-
near dependence is not found. The strain energy release rate
(~Ic) ~s related to the stress intensity factor as follows:

K 2(1_\1 2 )
G =~ (4)
Ic E
where E is Young's modulus.

The increase in E with volume fraction is well-known to be a

complex function of the moduli of the two phases and a parabolic
relationship is found 23 - 25. The effect that E rises at a faster
rate than Klc with increasing volume fraction of filler means that
even for "strong" fillers the value of Glc will go through a maxi-
mum at a certain volume fraction. An example of the relationship
between Glc and Vp is shown in Figure 9, a very similar result has
been obtained by Young and Beaumont 26 •

The analysis of Lange was carried further by Evans 27 who cal-

culated the increase in strain energy necessary to bow out the
crack between the particles. Evans also calculated the ratio of
Glc of the composite to that of the resin or the ratio of the ten-
sile strength,of the composite to that of the resin. Attempts to
predict changes in tensile strength in this manner are problematic
because the presen~e of the second phase inclusions increases the
size of the "inherent flaws" even if the adhesion between the re-
sin and filler is good. The flexural and tensile strengths of the
alumina filled resins are shown in Table IV. The strengths were
greatly improved by treatment with A187 and AllOO and impaired by
the HMDS treatment. The tensile strength and the stress intensity
factor are related as follows:
k
K = Y a a 2 (5)
Ic
where a is the applied stress, Y a geometric factor and a the ef-
fective "inherent flaw" size.

The tensile strength must be used here since the flexural

strengths are too close to the yield point. Inserting the relevant
values into this equation gives a flaw size of 400 ~m for resin B
filled with 30 % untreated alumina and 100 ~m for this resin with
896 A.C. MOLONEY ET AL.

0.6

0.2 •

o~----~-------.------~
0.2 Q4 0.6
Vp

Figure 9. The variation in the strain energy release rate with vo-
lume fraction of filler. 0 resin A + alumina; • resin B + alumina.

30 % A187 treated alumina. Thus, even for a composite with well-

bonded particles the effective flaw size is greater than that of
the pure resin which is about 50 ~m. The very large flaw sizes for
composites prepared from untreated alumina probably arises from the
linking of poorly bonded particles and the existence of agglome-
rates 23 •

Evans 27 has calculated the relative tensile strength in the

presence of inclusions, ocomposite/Oresin' assuming that the flaw
size is constant. This is clearly invalid for these composites
since we have seen that changing the adhesion between the phases
alters the flaw size. However, this ratio may be converted into
Klccomposite/Klcresin' The increase in strain energy (U) due to the
crack bending is:

(6)

where 0A is the stress required to propagate a crack through a se-

ries of obstacles. Thus, differentiating U with respect to c and
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS 897

Table IV Strengths of Composites Made from Treated and Untreated

Fillers.

Resin Filler Treatment Flexural Tensile

strength strength
MPa MPa

A 30% Alumina HDMS 106.8 -

Al 100 144.9 -
A187 154.7 -
B 30% Alumina none 95.7 45
HMDS 82.2 -
A187 131.9 87.8
Al 100 136.2 -

carrying out a numerical integration of the double integral, the

ratio of KIccomp/KIcresin may be obtained. These calculations have
been carried out by Green et a1 28 • The values derived from this
analysis are shown in Figu;,e 10, and compared with the experimental
values. At low filler contents the theoretical values are in very
good agreement; at higher volume fractions there is some scatter.
This may arise because the crack front bows out in a semi-ellipti-
cal manner rather than a semi-circular 28 • However, this analysis
shows that the increase in the stress intensity factor of these re-
sins may be explained by a mechanism of crack pinning.

The mechanism proposed by Lange 21 and Evans 27 is clearly in-

valid in the case of weak particles. For the dolomite filled re-
sins it was found that above a certain critical volume fraction
(~20%) a plateau is reached and the stress intensity factor no lon-
ger increases (see Figure 11). This appears to be due to the pro-
pagation of cracks through the particles. Some evidence in favour
of this explanation was provided from scanning electron micrograph
of the fracture surface (see Figure 12). Lange and Radford 2 9 have
observed a similar plateau with an aluminium hydroxide filled
epoxide resin.
898 A.C. MOLONEY ET AL.
5,0

4.0

o
o

3.0

Res~ n A 0 SlOt AIP3

RE's~nB • 5~O{'AIP3

2,0 3,0

Figure 10. The increase in the stress intensity factor required to

move a semi-circular crack through a series of obstacles ___ theo-
retical curve.

It is interesting to note that for the dolomite composites

the relationship between the yield stress and the volume fraction
of filler is completely different from that for the silica and alu-
mina composites. As demonstrated in Figure 13 the yield ,istress is
constant with increasing volume fraction of dolomite whereas for
silica and alumina the yield stress is considerably increased.
There appears to be a correlation between the dependences of KIc
on volume fraction on one hand and 0y on volume fraction on the
other hand but the precise nature of this correlation is unclear
at present.
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS 899

1.5 o

05

o~-----.------~-----,~
0.2 O.t. 0.6
Vp
Figure 11. The variation in the stress intensity factor with the
volume fraction of filler. Resin B + domolmite.

Figure 12. Scanning electron micrograph of the fracture surface of

dolomite filled resin B.
900 A.C. MOLONEY ET AL.

150
•
M Pa

100+==---..........-

50
0.2 0.4 0.6

Figure 13. The variation in yield stress with volume fraction of

filler. Resin B 0 alumina; 0 silica; + dolomite.

Combined Mode Adhesion Tests

The fracture of carefully prepared adhesive joints occurs nor-

mally in the centre of the bondline 30 • In the case of aluminium
adherends bonded with an epoxide adhesive in the tapered double
cantilever beam geometry, for example, failure was within the adhe-
sive and the value of Klc measured was in good agreement with that
of the pure bulk resin 3l • However, for the purposes of this work
centre of bond failure was undesirable and a geometry was sought
that produced fracture at the interface. With the scarf joint
Trantina 14 and Bascom et al 15 found that failure was at, or very
close to, the adherend-Surface and with a bond angle of 45 0 there
was little tendency for the crack to jump from one interface to
the other. This result was also confirmed in our work. The scan-
ning electron micrograph in Figure 14 clearly shows the grains of
alumina and, in some cases, resin is left adhering between the
grains.

Trantina 14 and Bascom et al lS calculated the combined mode

stress intensity factor (K(I-rI);) from a finite element analysis
of a homogeneous sheet containing a slant edge crack when subjected
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS 901

Figure 14. Scanning electron micrograph of the fracture surface of

a scarf joint between alumina and resin B.

to a uniform tensile stress. They justifi~d this approach since

the compliance of the scarf joint was found to be the same as that
of a solid aluminium specimen with a slant edge crack. However,
this result would be expected since the glue line is very thin with
respect to the rest of the aluminium specimen and even though the
modulus of aluminium is ~ 25 x that of the resin, it is principally
the aluminium that contributes to the compliance of the whole speci-
men. There are certain problems in determining the stress intensi-
ty factor at the interface between two materials since it is a crack
tip parameter and its value depends on the properties of the mate-
rial near the crack tip. It is extremely difficult to define the
mechanical parameters of an interfacial zone. For example, it is
unclear how the interfacial modulus may be defined. Bascom et al lS
used the modulus of aluminium in order to calculate the value of
G(I,II)c from:
K 2
(I,II)c
G (7)
(I,II)c E

However, the validity of this may be questioned, especially when

the moduli of the two phases are so different. Other authors 32
902 A.C. MOLONEY ET AL.

have used a combination of the moduli of the two materials (El and
E2) as shown below:

(8)

To avoid these problems of interfacial moduli, we decided to

calculate G(I,II)c using a compliance calibration. An extensometer
was connected to the ceramic inserts and the compliance (e) was
measured for specimens with varying crack lengths (a). G(I,II)c
was then calculated from the following relationship

p2 de
G =-- (9)
(I,II)c 2B da

where P is the load at break and B the thickness of the specimen.

The combined mode strain energy release rate for interfacial frac-
ture between degreased alumina and epoxide resin A was found to be
320 Jm- 2 • Preliminary results have indicated that this value va-
ries with the surface treatment of the alumina.

There are little data in the literature with which to com-

pare this value. The limited work reported has concerned the mix-
ed mode loading of aluminium-epoxide bonds. Mulville et a1 33 ,34
used a bending specimen where one half was aluminium and the other
epoxide. For unmodified epoxide resins they found values of
G(I,II)c of up to 300 Jm- 2 depending on the surface treatment of
the aluminium. For modified epoxide resins the values were con-
siderably higher, between 1300 and 3000 Jm- 2 depending on the angle
of 10ading 33 • Saxena 32 used compact tension and centre cracked
tension specimens which had a ratio of KII/KI of ~ 10 % and found
values of G(I,II)c for aluminium-epoxide bonds of the order of
270 Jm- 2 • As described above, Bascom and co-workers 35 used the mo-
dulus of aluminium in their calculations of G(I,II)c' however, they
obtained values between 80 and 300 Jm- 2 depending on the surface
treatment of the aluminium. This was for an unmodified epoxide
resin. For commercial epoxide resins adhesives modified with nylon
or an elastomer, they found values up to 900 Jm- 2 •

For the unmodified epoxide bonds the strain energy release

rate for interfacial fracture is generally higher than that of the
pure resin (GIc for bulk resin A is 130 Jm- 2 ). This is probably
because there is a restraint on the deformation at the crack tip
as propagation is forced to the adherend-resin interface 35 •
INTERFACIAL PROPERTIES OF FILLED EPOXIDE RESINS 903

It appears from these limited results that the scarf joint is

a promising method for modelling the fracture of particulate-filled
epoxide resins. The interfacial interactions between epoxides and
a wide range of substrates may be investigated. It is proposed to
extend this work by testing other surface treatments of alumina and
then to use quartz glass.

ACKNOWLEDGEMENTS

The authors would like to thank Professor J.G. Williams and

Dr. A.J. Kinloch for helpful discussions.

REFERENCES

1. D.C. Phillips and B. Harris, in "Polymer Engineering Compo-

sites", M.O.W. Richardson Editor, Ch. 2, AppliE}d Science,
London 1977.
2. K.L. Mittal, Solid State Technol. (May 1979) 89.
3. K.L. Trachte and A.T. DiBenedetto, Int. J. Po1ym. Mater., l,
75 (1971).
4. R.W. Davidge and G. Tappin, J. Mater. Sci., 3, 165 (1968).
5. C. Tourenq and F. Gragger, Naturstein Ind., 11, 19 (1965).
6. S.M. Wiederhorn, J. Amer. Ceram. Soc., 52, 9~(1969).
7. A.G. Evans, J. Mater. Sci., 7, 1137 (1971).
8. H.G. Tattersall and G. Tappin, J. Mater. Sci., 1, 296 (1966).
9. B.J. P1etka, E.R. Fuller and B.G. Loepke, in "Fracture Mecha-
nics Applied to Brittle Materials", STP No. 678, p. 3, 19,
ASTM, Philadelphia, PA, 1979.
10. B. Stalder and H.H. Kausch, J. Mater. Sci., 17, 2481 (1982).
11. P. Leevers. J. Mater. Sci., 17, 2g69 (1982).--
12. B. Stalder, Ph. Begue1in and~.H. Kausch, Int. J. Fract., in
press (1982).
13. J.G. Williams and H. Ford, J. Mech. Eng. Sci., ~, 7 (1964).
14. G.C. Trantina, J. Compo Mater., 6, 371 (1972).
15. W.D. Bascom, C.O. Timmons and R.L. Jones, J. Mater. Sci., 10,
1037 (1975).
16. W.D. Bascom and J. Oroshnik, J. Mater. Sci., 13, 1411 (1978).
17. L.J. Broutman and S. Sahu, Mater. Eng. Sci., S; 98 (1971).
18. J.C. Hammond and D.C. Quayle, Second Int. Coni. on Deformation,
Yield and Fracture of Polymers, P.R.I., March 1973, Cambridge,
UK.
904 A.C. MOLONEY ET AL.

19. A.C. Moloney, H.H. Kausch and H.R. Stieger, manuscript in pre-
paration (1983).
20. S. Kunz-Doug1ass, P.W.R. Beaumont and M.F. Ashby, J. Mater.
Sci., 15, 1109 (1980).
21. F.F. Lange, Phil. Mag., 22, 983 (1970).
22. F.F. Lange, J. Amer. Cera;. Soc., 54, 614 (1971).
23. A.C. Moloney, H.H. Kausch and H.R.:Stieger, J. Mater. Sci.,
18, 208 (1983).
24. J":A. Manson and L.H. Sperling, "Polymer Blends and Composites",
Plenum Press, New York, 1976.
25. O. Ishai and L.J. Cohen, Int. J. Mech. Sci., 9, 539 (1967).
26. R.J. Young and P.W.R. Beaumont, J. Mater. Sci~, la, 1343 (1975)
and 12, 684 (1977).
27. A.G.~vans, Phil. Mag., 26/6, 1327 (1972).
28. n.J. Green, P.S. Nicho1s~ and J.n. Embery, J. Mater. Sci., 14,
1657 (1979).
29. F.F. Lange and K.C. Radford, J. Mater. Sci., 6, 1197 (1971).
30. A.J. Kinloch, J. Mater. Sci., 15, 2141 (1980)~
31. R.A. Gledhill, A.J. Kinloch, S:-Yamini and R.J. Young, Polymer,
19, 574 (1978).
32. A.K. Saxena, Fibre Sci,. Technol., 12, III (1979).
33. n.R. Mu1vi11e, P.W. Mast and R.N. Vaishnav, J. Eng. Mater.
Techno1., 100, 27 (1978).
34. n.R. Mulvi11e and R.N. Vaishnav, J. Adhesion, 7, 215 (1975).
35. w.n. Bascom, R.L. Cottington and C.O. Timmons,-J. App1. Po1ym.
Sci. App1. Po1ym. Symp., ~, 165 (1977).
ABOUT THE CONTRIBUTORS

R. D. Adams is Reader in Mechanical Engineering, University

of Bristol, Bristol, England. He received his Ph.D. degree in 1967
from Cambridge University. He and his team have been working on the
stress and failure analysis of adhesive joints since 1970 and much
of their work has involved the use of finite element techniques.
His other interests are in composite materials, nondestructive
testing, and vibration damping. He is a Fellow of the Institution
of Mechanical Engineers and the Institute of Physics.

WaLtep ALtho! is Chief Scientist of the working group

"Adhesive Bonded Joints" at the DFVLR - Institute for Structural
Mechanics, Braunschweig, W. Germany which he jointed in 1955. His
research interest is in the elastomechanic properties of structural
aircraft adhesives, and has numerous publications dealing with the
engineering aspects of adhesive bonding. Because of his experience
in testing adhesive joints, he has participated in the
standardization of test methods of adhesively bonded joints in
Germany.

R. S. ALwap is presently Professor in the Department of

Applied Mechanics at the Indian Institute of Technology, Madras,
India which he joined in 1963. He received his Ph.D. degree in 1963
from the Indian Institute of Science, Bangalore. He has guided the
research of 9 Ph.D. scholars and 2 Master's candidates and has 40
publications to his credit. He is a consultant to various organi-
zations like Bharat Heavy Electricals, Indian Space Research
Organization, Ministry of Railways, Defense Research and Develop-
ment Organization, and several other private industries.

Gappon P. Andepson is presently on the staff at Thiokol Corp.

in Brigham City, UT. He received his ph.D. degree in Mechanical
Engineering from the University of Utah in 1973. His current
activities include development of analytical and experimental
techniques for evaluation of adhesives and solid propellants,
development of service life evaluation techniques, and development
of numerical stress analysis techniques. He has been a consultant

905
906 ABOUT THE CONTRIBUTORS

to a num~er of industrial and government establishments, and his

publications include 30 technical papers and one book written
mainly in the areas of adhesion testing, analysis of bond stresses,
and the mechanics o£ solid propellants.

Edga~ And~ews has been Professor of Materials at Queen Mary

College, University of London since 1968 when he established the
Department of Materials at that college. He was Dean of Engineering
at Queen Mary College, 1970-1973. His publications include Fracture
in Polymers, a monograph published in 1968 and over 90 research
papers and review articles. He has also published several books of
a theological nature. He was awarded the A. A. Griffith Silver
Medal by the Materials Science Club of Great Britain in 1977 for
his work on the fracture of materials.

W. L. Baun is with the Air Force Materials Laboratory,

Wright-Patterson Air Force Base, Ohio. He carried out graduate
studies at the Air Force Institute of Technology and Ohio State
University. After five years of commissioned service in the Air
Force, he accepted a position as a research chemist in the Air
Force Materials Laboratory where he has carried out research on
bulk and surface characterization of solid state materials.

A. Beeve~s is at the Oxford polytechnic, Oxford, England

where in 1968 he established a specialist facility in joining and
fastening technology. In addition to providing a teaching labora-
tory for engineering students, this facility undertakes a variety
of R&D programs. He carried out postgraduate work in resistance
welding at British Welding Research Association and was awarded an
M.Sc. for his research on spot welding of hardenable steels.

J. P. Bett is Professor at the University of Connecticut,

Storrs, which he joined in 1969 and has been very active in both
the Chemical Engineering Department and the Institute of Materials
Science. Prior to the University of Connecticut, he worked at
Monsanto's Chemstrand Research Center for 3-1/2 years and the
duPont plastics Department for 6-1/2 years. He received his D.Sc.
degree in Chemical Engineering from MIT. His publications in
textile fibers and epoxy resin structure-property relationships are
rather well known, and he has held extended professional
appointments in West Germany, Italy and Israel.

Etisabeth J. Be~ge~ is currently an advanced scientist at the

Owens-Corning Fiberglas Technical Center in Granville, OH. She
received her Ph.D. degree in Physical Chemistry from the
Pennsylvania State University in 1980.

F. James Boe~io is presently Professor of Materials Science

at the University of Cincinnati which he joined in 1970. He
received his Ph.D. degree in Macromolecular Science from Case
ABOUT THE CONTRIBUTORS 907

Western Reserve University in 1971. His current research interests

include adhesive bonding of metals and polymers and surface
analysis, especially using infrared and Raman spectroscopy. He has
published about 50 papers describing his research results.

Justin c. Botge~ is currently managing the division which

produces epoxy and other formulated products at the Amicon Corpora-
tion, Lexington, MA which he cofounded in 1962. He received his
Sc.D. degree in Chemical Engineering from MIT.

Joseph Boutitie~ is a Research Associate with Singer-Kearfott

in Little Falls, NJ where he has been a materials and processing
technician for fourteen years. He has worked with the development
of experimental instrumentation for the testing of materials. He
has a certificate in Electronic Technology from Newark College of
Engineering, and an Associates Degree in Applied Science, Elec-
tronic Technology, from the County College of Morris, Randolph, NJ.

Watte~ B~oakmann has been the Head of the Structures and

Composites Department in the Fraunhofer-Institut fur angewandte
Materialforschung in Bremen, W. Germany since 1970. He graduated as
a Diplom-Ingenieur in 1966 followed by work as a scientist with
Prof. Matting in the area of adhesion and adhesives, and obtained
his ph.D. degree in 1969 in this field. He has published a book and
more than 100 papers dealing with adhesion, bonding techniques and
composites.

P. B. C~ostey is a consulting engineer at the Materials

Research Laboratory, Inc., Glenwood, IL. He received his ph.D.
degree in Metallurgical Engineering from Illinois Institute of
Technology, Chicago, in 1966. He has over a dozen publications
dealing with the measurement of fracture toughness in a variety of
material systems and the prediction of failure in structures
exposed to different service conditions.

R. G. Davidson is currently leader of the optical spectro-

scopy section, Polymer Research Group, Materials Research Labora-
tories, Ascot Vale, Victoria, Australia. He graduated in Applied
Chemistry in 1962 from the Royal Melbourne Institute of Technology
and his interests have included dyes and pigments, explosives, and
synthetic rubber. He is presently interested in the application of
UV-VIS-IR spectroscopy to the characterization of polymeric mater-
ials, and to the study of polymer degradation.

K. L. DeV~es is presently Professor at the University of

Utah, Salt Lake City and was Department Chairman during 1969-1981.
He received his Ph.D. degree in Mechanical Engineering from the
University of Utah in 1962. During 1975-1976, he headed the Polymer
Program at the National Science Foundation, and has served on
several NSF, NBS, and NRC advisory committees. He served as the
908 ABOUT THE CONTRIBUTORS

Chairman of Gordon Research Conference on Science of Adhesion in

1976, as well as Gordon Conference on Basic Failure Mechanisms in
Polymers and Composites in 1980. He is the coauthor of the book,
Analysis and Testing of Adhesive Bonds (1979), and his publications
include approximately 150 papers, chapters in books, proceedings,
etc. on fracture, polymers, adhesives, biomaterials, and dental
materials.

WoLfgang Diem is with Degussa, Hanau, W. Germany ad was a

research assistant at the Max-Planck-Institute for Metals Research
in Stuttgart, W. Germany. He obtained his doctorate at the Univer-
sity of Stuttgart in 1982. His interests include physical metal-
lurgy, mechanical engineering, and fabrication and testing of
metal-to-ceramic joints and adhesively bonded joints.

R. G. DiLLingham is currently a candidate for a Ph.D. degree

in Materials Science at the University of Cincinnati where he
received his M.S. degree in Materials Science in 1982. His research
interests include adhesive bonding of metals and the influence of
metal oxides on the curing reactions of thermosetting adhesives.

A. A. DonateLLi is Associate Professor of Chemical Engineer-

ing at the University of Lowell, Lowell, MA. He received his Ph.D.
degree in Chemical Engineering from Lehigh University. His research
interests are in the areas of polymer science and mass transfer.

Yona Eck8tein is currently an advanced scientist at the

Owens-Corning Fiberglas Technical Center in Granville, OH. Before
her current position, she spent two years as a Research Associate
at the Institute of Polymer Science, University of Akron. She
received her Ph.D. degree from Hebrew University, Jerusalem.

Gerhard ELssner is a research staff member at Max-Planck-

Institute for Metals Research, Stuttgart, W. Germany. He obtained
his doctorate in physical metallurgy at the University of Stuttgart
in 1967. His group is involved in research on the mechanical pro-
perties and the microstructure of metal-to-ceramic transitions. His
interests include materials science, composite materials, ceramo-
graphy and metallography. He has published about 50 papers.

R. A. Everett, Jr. is currently involved in fatigue and frac-

ture research of adhesively bonded structures at the U. S. Army
Structures Laboratory located at NASA Langley Research Center,
Hampton, VA. He received an M.S. in Engineering Mechanics from Old
Dominion University in 1980. Prior to his current position, he
worked at the Army Missile Command, the Army Aviation Laboratory,
Lockheed-Georgia Company, and Lockheed-California Company. He has
authored or coauthored about 15 technical papers.
ABOUT THE CONTRIBUTORS 909

R. Exatto is with Fokker B.V. Technological Centre, Schiphol-

Oost, The Netherlands.

Eugene C. Fpancis is Chief Structural Research Engineer,

Chemical Systems Division, United Technologies, Sunnyvale, CA. He
received B.A. in Physics from the University of California. He has
published about 50 papers involving finite and nonlinear analysis
and experimentsl testing of filled and unfilled polymer systems,
viscoelastic fracture mechanics, nonlinear constitutive theories,
experimental stress analysis, service life analysis, adhesive
bonding, etc.

Fpancois Gaittapd is a Research Associate in the Applied

Chemistry and Chemical Engineering Department, University Claude
Bernard, Lyon, France where he received his M.S. in Materials
Science in 1980. He is involved in the characterization of stain-
less steels submitted to various passivation treatments.

Yvon Gitibept is currently Assistant Professor in Civil

Engineering at Rheims University. He graduated in Engineering from
C.N.A.M., France in 1968 and obtained a Science Doctorate from
Rheims University in 1978. From 1965-1972, he worked in far infra-
red and Raman effect spectroscopy at Rheims University, and has
worked in the field of experimental mechanics of adhesive joints
since 1973. He is a consultant in Civil Engineering.

D. Gutieppea-Lemini is Senior Structural Research Engineer,

Chemical Systems Division, United Technologies, Sunnyvale, CA.
Received Ph.D. degree from the University of Utah.

J. A. Happis is Research Assistant in Mechanical Engineering,

University of Bristol, England, where he received B.Sc. degree
(First Class Honours) in 1977. He has several years experience in
adhesives research with particular interest in impact performance
and failure of bonded joints.

Otto-Diedpich Hennemann has been since 1978 in the Department

of Nonmetallic Materials and Adhesive Techniques in the Fraunhofer-
Institut fur angewandte Materialforschung in Bremen, W. Germany. He
was graduated as a Diplom-Physicist in 1972 followed by work in the
area of spinwaves with Prof. Urban and received his Ph.D. in 1975
in this field from the University of Bonn.

Edwapd J. Hughes is Principal Scientist with Singer-Kearfott

in Little Falls, N.J. where he has been for 17 years and has con-
ducted extensive research into the microstrain behavior of metals,
adhesives and composite materials. Prior to that he was a Senior
Scientist at the Materials Research Corp., Orangeburg, N.Y. He
received his Ph.D. degree in Physical Metallurgy from Birmingham
University, England.
910 ABOUT THE CONTRIBUTORS

D. L. Hunston has been with the National Bureau of Standards

since 1980 where he is currently helping to establish a program in
adhesives and composites. Prior to coming to NBS, he was (1971-
1980) at the Naval Research Laboratory, Washington, D.C. where he
studied polymeric materials and between 1978 and 1980 directed the
Adhesives and Polymer Composites Section. He received his Ph.D.
degree in Physical Chemistry from Kent State University, followed
by 2 years at Northwestern University on an NIH Postdoctoral
Fellowship.

R. W. Hytands has since March 1980 been Principal of

Portadown Technical College, one of the largest further educational
establishments in N. Ireland. Prior to his current position, he has
had a number of academic appointments including Principal Lecturer
and Director of Studies and Head of School of Building at the new
Ulster Polytechnic. At the Polytechnic he carried out a program of
research associated with prestressed timber beams which led to the
award of Ph.D. He has published papers and booklets dealing with
strength characteristics of adhesives, home grown timber, deflec-
tion of non-uniform beams and buckling of non-uniform struts.

T. Igapashi is Associate Professor at Gunma University,

Kiryu, Japan where he has been since 1974. He graduated from Tokyo
Institute of Technology in 1953 and holds a ph.D. degree. Before
coming to Gunma University, he was with the Research Laboratory of
Hitachi Cable Ltd. (1961-1974). He is an Associate Editor of Jour-
nal of Adhesion, and his publications include: Theory of Pressure
Sensitive Adhesion and its Applications, 1978, Keieikaihatsu
Center, Tokyo; Testing Methods and Estimation of Polymeric Mater-
ials, Mechanical Test Methods, 1980, Baihukan, Tokyo; and Ultra-
sonic Wave and Nondestructive Testing, Keieikaihatsu Center,
Nagoya, to be published.

Ravi B. Jathap is currently Quality Control and product

Development Manager at Elmendorf Board Corp., Claremont, N.H. Prior
to his present position, he was Technical Manager, Forest Indus-
tries Pvt. Ltd. (India), and Research Assistant/Trainee at the
Research Center of Bison-Werke, Springe, W. Germany. He received
his M.S. degree in Wood Technology from the University of
Minnesota.

J. P. Jeandpau is an ingenieur at the Centre Mecanique-

Chimie-Materiaux, Etablissement Technique Central de l'Armement,
Arcueil Cedex, France where he is responsible for Laboratoire
"Collages et Adhesifs". He received Doctorat de 3eme Cycle
(Chimie-Physique) and the subject of the thesis was kinetic and
synthetic aspects of the photoreduction of aromatic ketones.

Waptan Jemian is Professor of Mechanical Engineering and

Materials Engineering at Auburn University. He received his Ph.D.
ABOUT THE CONTRIBUTORS 911

degree in 1956 from Rensselaer polytechnic Institute. He was Senior

Fellow and Head of the Power Rectifier Fellowship of Mellon Insti-
tute with major interests in materials and junction structure-
property relationship until 1962. He is active in research in
adhesion, dynamic characteristics of materials and solidification
morphology.

w. S. Johnson is currently the group leader of the Adhesive-

ly Bonded Structures Group of the Fatigue and Fracture Branch at
NASA Langley Research Center, Hampton, VA. He received a Ph.D. in
Solid and Structural Mechanics from Duke University. He has pre-
viously worked as an engineer at the U.S. Naval Ship R&D Center and
at General Dynamics/Fort Worth. He has authored or coauthored
approximately 30 technical papers in the area of fatigue and frac-
ture of materials. He is currently chairman of an ASTM task group
concerned with standardizing durability testing of adhesively bond-
ed joints.

Hans-Henning Kausch has been Director of the Polymer Labora-

tory at the Swiss Federal Institute of Technology, Lausanne,
Switzerland, since 1976. Prior to his current position, he was
associated with the Battelle Institute eV, Frankfurt am Main and
the Deutsches Kunststoff-Institute, Darmstadt. He received his Dr.
rer. nat. in Nuclear Physics from the University of Gottingen, and
his interests turned to polymeric materials with his affiliation
with the Mannesmann-Research Institute in 1960. He is predominantly
concerned with the various interrelations between molecular
properties, physical modifications of polymers through orientation,
reinforcement and processing and macroscopic mechanical properties.

A. C. P. Kho is currently carrying out research on the

dynamic performance of bonded box-section structures. He graduated
in Engineering from Oxford Polytechnic in 1981.

A. J. KinLoch is with the Ministry of Defense at Waltham

Abbey, U.K., where he is currently leader of a group engaged in
research and development in the field of adhesives, adhesion and
polymer fracture. He received his Ph.D. degree in Materials Science
in 1972 from the Queen Mary College, London University.

RonaLd A. KLine is Assistant Professor at the University of

Oklahoma, Norman, OK. Before his current position, he was Senior
Research Engineer with General Motors Research Laboratories,
Warren, MI (1979-1982) and during 1978-1979 he was Senior Research
Scientist with General Dynamics in Fort Worth, TX. He received his
Ph.D. degree in Mechanics and Materials Science from Johns Hopkins
University in 1978. His research interests are nondestructive test-
ing, mechanics of composite materials, and adhesive bonding.
912 ABOUT THE CONTRIBUTORS

wotfgang G. Knau88 is presently Professor at California

Institute of Technology which he joined in 1965. He is very well
known in the field of fracture of polymers and viscoelasticity. He
has been a consultant to many industrial organizations, and has
also served as a consultant on the AFML/MBC sponsored program on
"Fatigue Behavior of Adhesively Bonded Joints" and on the AFML/MBM
sponsored program on "Time-Dependent Environmental Behavior of
Graphite/Epoxy Composites". He is currently associate investigator
in the Integrated Methodology for Adhesive Bonded Joint Life
Predictions program. He received his Ph.D. degree in 1963 from
Caltech.

Han8geo~g Kottek has been since 1976 with the Fraunhofer-

Institut fur angewandte Materialforschung in Bremen, W. Germany. He
received his degreees in Analytical Chemistry fom the University of
Goettingen, W. Germany. He is currently working in the area of
chemistry of adhesion and adhesives, especially the aging behavior
of metal bonds, and has published several papers dealing with this
subject.

Dimite~ L. Kotzev is Research Scientist at the Scientific-

Industrial Center for Special Polymers, Sofia, Bulgaria. Hold
doctorate degree.

A. Kwake~aak is with Fokker B.V. Technological Centre,

Schiphol-Oost, The Netherlands.

Lieng-Huang Lee is a Senior Scientist of Xerox Corp.,

Webster, N.Y. He is the author of over 50 technical papers and 25
patents and editor of six books on adhesion, friction and wear of
polymers. Recently, he was selected as the distinguished scholar to
participate in the U.S.-China Exchange Program by the National
Academy of Sciences and the Chinese Academy of Sciences.

K. M. Liechti is currently Associate Professor, Department of

Aerospace Engineering and Engineering Mechanics, University of
Texas at Austin. Before this he was with the Materials Research
Laboratory of General Dynamics, Fort Worth, TX which he joined in
March 1980 to work on the Integrated Methodology for Adhesive Bond-
ed Life Prediction Program. He has been interested, inter alia, in
the development of failure criteria to explain the failure modes
observed in the model lap shear joints. He received his ph.D.
degree in Aeronautics from California Institute of Technology in
1980.

Yu. S. Lipatov is with the Institute of Macromolecular

Chemistry, The Ukrainian SSR Academy of Sciences, Kiev, USSR.

S. Matt is presently engaged in fatigue and fracture research

of adhesively bonded structures at NASA Langley Research Center,
ABOUT THE CONTRIBUTORS 913

Hampton, VA. He received his Ph.D. in 1977 from the University of

Washington, Seattle, and thereafter joined the faculty of the
University of Maine, Orono and became Associate Professor in 1981.
He obtained M.Sc. in Mechanical Engineering from Banaras Hindu
University, India in 1966. He has authored and coauthored more than
25 technical papers in fatigue and fracture mechanics.

N. T. McDevitt is with the Materials Laboratory, Wright-

Patterson Air Force Base, Ohio. He received his graduate degree in
Chemistry from Lehigh University. After graduation, he accepted a
commission in the United States Air Force. Following his military
tour in the Air Force Materials Laboratory, he remained in the
laboratory as a civilian research chemist. During his career in the
Materials Laboratory, he has specialized in molecular spectroscopy
(IR and Raman) and more recently research on adhesive bonding.

J. Dean Minfopd is a Scientific Associate in the product and

Process Engineering Division at Alcoa Laboratories, Alcoa Center,
PA. He received his ph.D. degree in Bio-Organic Chemistry from the
University of Pittsburgh in 1951. He has focussed his attention for
over 25 years on the study of joining aluminum to itself and other
materials through the use of adhesives; and has published more than
50 technical papers dealing with the durability of aluminum bonded
joints. He is the author of the Alcoa Handbook on Bonding and has
written a number of chapters for various books. He is the recipient
of the 1982 Adhesives Age Award for outstanding contributions to
the adhesives industry.

KashmiPi La! Mitta!* is presently employed at the IBM

Corporation at Hopewell Junction, N.Y. He received his M.Sc. (First
Class First) in 1966 from Indian Institute of Technology, New
Delhi, and Ph.D. in Colloid Chemistry in 1970 from the University
of Southern California. In the last ten years, he has organized and
chaired a number of very successful international symposia and in
addition to this volume, he has edited nineteen more volumes as
follows: Adsorption at Interfaces, and Colloidal Dispersions and
Micellar Behavior (1975); Micellization, Solubilization, and
Microemulsions, Volumes 1 & 2 (1977); Adhesion Measurement of Thin
Films, Thick Films and Bulk Coatings (1978); Surface Contamination:
GeneSis, Detection, and Control, Volumes 1 & 2 (1979); Solution
Chemistry of Surfactants, Volumes 1 & 2 (1979); Solution Behavior
of Surfactants - Theoretical and Applied Aspects, Volumes 1 & 2
(1982); and Physicochemical Aspects of Polymer Surfaces, Volumes 1
& 2 (1983); Adhesion Aspects of Polymeric Coatings, (1983);
Surfactants in Solution, Volumes 1, 2 & 3 (1984); and Polyimides:
Synthesis, Characterization and Applications, Volumes 1 & 2 (1984).
Also he is Editor of the series, Treatise on Clean Surface
Technology, a multi-volume work in progress. In addition to these

* As editor of this volume.

914 ABOUT THE CONTRIBUTORS

books he has published more than 50 papers in the areas of surface

and colloid chemistry, adhesion, polymers, etc. He has given many
invited talks on the multifarious facets of surface science,
particularly adhesion, on the invitation of various societies and
organizations in many countries allover the world, and is always a
sought-after speaker. He is a Fellow of the American Institute of
Chemists and Indian Chemical Society, is listed in American Men and
Women of Science, Who's Who in the East, Men of Achievement and
many other reference works. He is or has been a member of the
Editorial Boards of a number of scientific and technical journals.
Currently, he is Vice-President of the India Chemists and Chemical
Engineers Club (USA).

Anne C. Moloney has been at the Swiss Federal Institute of

Technology, Lausanne, Switzerland since 1979 where she is studying
the fracture behavior of particulate filled epoxide resins. She
obtained her Ph.D. degree in 1979 for work conducted on the effect
of carriers on the environmental stability of adhesive joints. From
1976 to 1979 she worked at the Leicester Polytechnic, England.

Charles T. Mooney is R&D Manager for the Electronic Materials

Group at Amicon Corp., Polymer Products Division, Lexington, MA. He
recp.ived his B.S. degree in Chemistry/Biology from Boston College
and is currently a candidate for the M.S. degree in Chemical
Engineering at the University of Lowell, Lowell, MA.

C. E. M. Morris currently heads the structural adhesives work

at the Materials Research Laboratories, Ascot Vale, Victoria,
Australia. Before joining MRL, she spent three years at the Corpor-
ate Laboratory of I.C.I. Ltd. in Cheshire, U.K. She received her
Ph.D. degree in Physical Chemistry in 1967 from the University of
Sydney. she has conducted research programmes on the kinetics and
mechanisms of free radical polymerization, surface modification of
polyolefins for improved adhesion and characterization procedures
for polymers, and is currently interested in epoxy-based systems
for ae.rospace applications.

William J. O'Brien is Professor of Dental Materials at the

University of Michigan and Director of the Surface Science labora-
tory in the Dental Research Institute. He received his Ph.D. from
the University of Michigan in 1967 and has been on its faculty
since 1970. He is coeditor of Outline of Dental Materials and Their
Selection with Gunnar Ryge and coauthor of Dental Materials: Pro-
perties and Manipulation with Drs. Craig and Powers. He has served
as a consultant to the American Dental Association and World Health
Organization, and has been very active in A.D.A. and I.A.D.R. He is
currently a representative of the American Association of Dental
Schools to the American National Standards Institute Committee on
Dental Materials. His current research is in the optical properties
and adhesion of dental materials and has published in the areas of
ABOUT THE CONTRIBUTORS 915

precious metals, sealants, surface phenomena, ceramics, and

polymers.

Jae M. Pa~k has recently completed his Ph.D. degree in

Chemical Engineering at the University of Connecticut, Storrs with
thesis research on epoxy resin bonding to copper. Prior to coming
to the United States, he worked for three years as a Scientist at
the Korean Institute for Science and Technology. His undergraduate
work was in chemical engineering at Seoul National University.

P. J. Pe~ce is affiliated with the Materials Research Lab-

oratories , Ascot Vale, Victoria, Australia which he joined in
1978. Before joining MRL, he was at the Explosives Research and
Development Establishment, Waltham Abbey, U.K. and later at Joint
Tropical Research Unit, Innisfail, Australia carrying out research
which included the effects of a tropical environment on a wide
variety of polymeric materials. He received his M.S. deg~ee in
Polymer Chemistry from Salford University in 1974, and his current
studies involve the characterization and cure kinetics on a variety
of structural adhesives and composites. He recently completed a 15
month attachment to the Army Materials and Mechanics Research
Center, USA.

s. S. Pesetskii is a Senior Researcher in the Institute of

Mechanics of Metal-Polymer Systems, Byelorussian SSR Academy of
Sciences, USSR, where he has been since 1972. He defended his
candidate thesis in 1980. He is interested in technological pro-
blems of polymers and metal-adhesive bonds, and has more than 50
papers and 25 USSR patents to his credits.

Gunte~ Petzow is a member of the board of directors of the

Max-Planck-Institute for Metals Research in Stuttgart and al~o
Professor at the Universities of Stuttgart and Berlin. He received
his doctorate in 1959 from the University of Stuttgart. His re-
search activities have been concerned with problems in the field of
physical metallurgy, powder metallurgy, special ceramics and with
phase diagrams of metallic and ceramic materials. He has received
several awards among them the Kuczynski Diploma, the honorary mem-
bership of the Korean Institute of Metals, doctor honoris causa of
the Tokyo Institute of Technology, the Hume-Rothery Prize of the
Metals Society, London, and the ASM Fellowship. He has published
more than 250 research papers and four books and holds eight
patents.

Roscoe A. Pike is a Senior Materials Scientist at the United

Technologies Research Center, E. Hartford, CT which he joined in
1967. He received his Ph.D. degree in Organic Chemistry from MIT in
1953. His fields of activity include high temperature resin based
composites, adhesive technology, thermoplastic resins and micro-
electronic packaging. His areas of publications include silicone
916 ABOUT THE CONTRIBUTORS

chemistry, grinding fluids, advanced composites and adhesives; and

he has been awarded 35 u.s. patents.

A. Pizzi is with the National Timber Research Institute,

Pretoria, South Africa. He had his education at the University of
Rome where he obtained a Doctorate in Physical Chemistry of Poly-
mers. He emigrated to South Africa in 1969 and continued his
studies at the University of the Orange Free State in Bloemfontain
and obtained a Ph.D. degree in Applied Organic Chemistry. He was
awarded the 1980 Forest Products Research Society (USA) Bork Award
for his work on wood adhesives based on bark tannin extracts, and
was nominated for the 1982 Sir Stuart Mollingson Forest Research
Medal (U.K.). He is on the editorial boards of IntI. J. Adhesion
Adhesives and J.Wood Chemistry Technology, and is the editor and
principal author of a book entitled Wood Adhesives Chemistry and
Technology to appear in 1983. His research interests are wood
adhesives, chemistry of wood preservation, wood chemistry, polymers
and conformational analysis, and has published 150 papers.

K. N. Ramachandpan Nambissan is presently Assistant Professor

in the Civil Engineering Department of Regional Engineering
College, Calicut, India. He completed his Ph.D. degree at the
Indian Institute of Technology, Madras, India. His dissertation was
in the field of fracture mechanics and the present paper forms a
part of the research undertaken for his doctoral degree. He is the
author of several publications.

R. Ramaswamy is a Senior Scientist in the Polymers and

Special Chemicals Division of Vikram Sarabhai Space Centre (Indian
Space Research Organization) at Trivandrum. He received his Ph.D.
in 1971 from Poona University, India. He and his group are engaged
in the development of specialty adhesives and sealants for space
and aircraft applications. His interests include adhesion mech-
anisms and structure-property relationships of adhesives. He has
published papers in the fields of ion exchange resins and adhesion.

Stephen T. Rasmussen is Assistant Professor of Basic Dental

Technology at the School of Dentistry, Case Western Reserve Univer-
sity, Cleveland, OH which he joined in 1979. He received his Ph.D.
degree in Materials Science and Engineering in 1970 from the Univ-
ersity of Utah. His current research deals with the fracture pro-
perties of human tooth structure and dental materials, and the
adhesion of dental materials and has published in these areas.

E. J. Ripting is President of the Materials Research Labora-

tory, Inc., Glenwood, IL. He received his Ph.D. degree in Metal-
lurgy in 1952 from the Case Institute of Technology. Work on the
application of fracture mechanics to adhesive joints has been
conducted at the Materials Research Laboratory over the past two
decades and has resulted in over a dozen publications in the open
ABOUT THE CONTRIBUTORS 917

literature. In recognition of his work on fracture mechanics, he

was elected a Fellow of the American Society of Metals in 1970. He
is a member of several NMAB Committees, and chairman of an ASTM
Task Group on Crack Arrest.

Alain A. Roche is with the Applied Chemistry and Chemical

Engineering Department, University Claude Bernard, Lyon, France and
received his thesis degree in 1978 from the same university. During
1979-1980, he was Research Scientist at the Air Force Materials
Laboratory, Wright-Patterson Air Force Base, Ohio; and he was
appointed "Attache de Recherche" at the CNRS in 1980. He is the
author of over 30 papers dealing with the characterization of solid
surfaces and the mechanical properties of metal/adhesive systems.

Stanislav I. Rokhlin is with the Materials Engineering

Department of the Ben-Gurion University of the Negev, Beer-Sheva,
Israel. He received his Ph.D. degree in Physics in 1972. He has
published over 30 papers and holds some patents. His research
interests include wave propagation and diffraction and application
of ultrasonic waves for materials evaluation especially thin films
and composite materials.

Maupice J. Romand is Head of the Materials Science Division

of the Applied Chemistry and Chemical Engineering Department,
University Claude Bernard, Lyon, France where he received his Ph.D.
in Physical Chemistry in 1970. His present research activities are
mainly devoted to the characterization of passive films formed on
various substrates, and to the development of low-energy electron-
induced x-ray spectroscopy (LEEIXS) for characterizing solid sur-
faces and thin films, and has published some 70 papers, most of
them dealing with applications of x-ray emission spectroscopy and
surface characterization.

John Romanko is currently Engineering Specialist Sr., Mater-

ials Research Laboratory, General Dynamics/Fort Worth Division, TX.
He received his Ph.D. in Molecular Physics in 1954 from the Univer-
sity of Toronto. He has conducted studies in full field displace-
ment (strain) analysis in materials and composites using laser
holographic and interferometry techniques. He is currently involved
in adhesive bonding and is program manager on three USAF programs
on adhesive bonding. He has published on holography and adhesive
bonding.

Klaus Ruhsland is a certified engineer for plastic material

engineering and head of a research team for adhesive bonding of
metals and welding of plastics at the Central Institute for Welding
Engineering of the GDR, Halle (Saale), E. Germany. In 1977 he was
awarded the Goethe Prize for Science and Technology of the City
COuncil of Berlin for developing the adhesive Epasol FV/ZIS 939. He
has published about 60 technical papers and is coauthor of the
918 ABOUT THE CONTRIBUTORS

technical manual Plastwerkstoffe in der Feingeratetechnik published

in 1973.

John L. RutherfoPd is Director, Materials and Process Labora-

tory, Singer-Kearfott in Little Falls, N.J. where he has been for
20 years. Prior to that he was a Research Physicist at the Franklin
Institute Laboratories in Philadelphia where he helped develop
floating zone-refining techniques for iron and titanium. He has a
Ph.D. degree in Metallurgy from the University of Pennsylvania. He
is listed in American Men of Science.

J. S. Santner is a Research Engineer at the Materials Re-

search Laboratory, Inc. Glenwood, IL. He received his Ph.D. degree
in Materials Science and Engineering in 1975 from Northwestern
University. His correlating the fatigue and fracture properties of
structural materials to their microstructures characterized by
light, scanning and transmission electron microscopy has resulted
in over a dozen publications in the open literature.

J. P. Sargent is currently doing postdoctoral research at

Bristol University, England. He received his Ph.D. degree studying
the linear thermal expansion coefficients of carbon fiber rein-
forced plastic. He is presently interested in the application of
interferometry to the study of the behavior of adhesive joints when
exposed to humid environments and also has a keen interest in
microprocessors.

P. Sasidharan Achary is a Scientist in the Polymers and

Special Chemicals Division of Vikram Sarabhai Space Centre (Indian
Space Research Organisation), Trivandrum. He received B.Sc.
(Special) in 1970 from Kerala University, India. He is engaged in
the research and development of various adhesives and sealants
suitable for aerospace applications. He has about 10 years exper-
ience in the development of various adhesives and sealants based on
epoxies, phenolics, acrylates and polysulfides.

L. M. Sergeeva is with the Institute of Macromolecular

Chemistry, The Ukrainian SSR Academy of Sciences, Kiev, USSR.

GedaLia Sharon is currently on leave at the University of

Utah from Rafael Armament Division, Haifa, Israel. He received B.S.
in 1967 in Chemical Engineering from Technion, Israel and M.S. in
Industrial Engineering in 1981. He has conducted studies in the
areas of adhesives and joining.

S. J. Shaw has been with the propellants, Explosives and

Rocket Motor Establishment, Ministry of Defense, Waltham Abbey,
U.K. since 1976 where he is currently engaged in research and
development in the field of structural adhesives. He obtained his
M.Sc. in Polymer Technology in 1975 from the Institute of Polymer
ABOUT THE CONTRIBUTORS 919

Technology, Loughborough University, Loughborough, U.K.

S. V. Shchepbakov is with the Institute of Mechanics of

Metal-Polymer Systems, Byelorussian SSR Academy of Science, USSR
where he heads the Design Office. He has a Candidate of Science
degree. He is interested in designing and producing machine
elements from polymer-based composite materials, and in improving
the strength of reinforced systems. He has over 50 USSR patents to
his credit and has coauthored three monographs: Plastic Gearings
(1965); Metallopolymeric Materials and Items (1978); and
Metallopolymeric Gearings (1981).

Mapgapet Shepidan is currently a graduate student in Ph.D.

program in Chemistry at VPI&SU, Blacksburg, VA. She received her
B.S. degree from Gannon College.

F~pncois P. Sidopoff is Head of the Solid Mechanics Labora-

tory, Ecole Centrale de Lyon, Ecully, France which he joined in
1978. He received his Ph.D. degree in Mechanics in 1976 from the
Ecole Nationale Superieure des Arts et Metiers (ENSAM) in Paris,
and was appointed "Charge de Recherche" at the CNRS. He is con-
cerned with wide-ranging research programs including solid material
viscoelasticity, plasticity, viscoplasticity and behavior modeling,
and has published over 40 papers in these areas.

V. E. Stapzhynskii has been with the Institute of Mechanics

of Metal-Polymer systems, Byelorussian SSR Academy of Sciences,
USSR since 1959. He defended his Candidate Thesis in 1968, and
since 1971 has been the Head of the Laboratory of Metallopolymeric
Structures. He is interested in the development of engineering
composite materials, designing of metallopolymeric machine parts,
and increasing the load bearing capacity of drives. He has publish-
ed numerous papers and has coauthored two monographs Metallopoly-
meric Materials and Items (1978), and Metallopolymeric Gearings
(1981) •

H. R. Stiegep is presently Head of the polymeric Materials

Department, Brown Boveri & Cie, Zurich, Switzerland which he joined
in 1975. He received his Dr. phil. II in Chemistry from the Univer-
sity of Zurich in 1973.

Tadatomo Suga is research assistant at the Max-Planck-

Institute for Metals Research in Stuttgart, W. Germany. He obtained
his master of engineering at the University of Tokyo in 1979.
Recently he has been working on fracture mechanics problems of
metal-to-ceramic and adhesively bonded joints.

Robept Y. Ting is presently Head, Materials Research, Under-

water Sound Reference Division of the Naval Research Laboratory in
Orlando, FL. In his current position he is responsible for R&D work
920 ABOUT THE CONTRIBUTORS

on materials used in the Navy's underwater acoustic systems. He

received his Ph.D. degree from the University of California, La
Jolla. He has published extensively on the various subjects such as
drag reduction, cavitation, polymer rheology, composite materials,
fracture mechanics and adhesion. He is listed in American Men and
Women of Science, Who's Who in the East, and Who's Who of Sino-
Americans.

T. T. Todosiyahuk is with the Institute of Macromolecular

Chemistry, The Ukrainian SSR Academy of Sciences, Kiev, USSR.

P. K. Tsapev is with the Institute of Macromolecular

Chemistry, The Ukrainian SSR Academy of Sciences, Kiev, USSR.

A. M. Usmani is a Senior Research Staff, Research Institute,

at the University of Petroleum and Minerals, Dharan, Saudi Arabia.
Prior to his current position, he was Research Professor and
Associate Director of Polymer Research at the University of Dayton.
He has more than 20 years of experience in polymer science and
technology and has been associated with a number of industrial
concerns. He received his Ph.D. degree in Chemistry from North
Dakota State University and has more than 100 publications and many
issued U.S. patents to his credit.

H. A. van Hoof is with Fokker B.V., Technological Centre,

Schiphol-Oost, The Netherlands.

John D. VenabLes is the Manager of the Materials and Surface

Science Department at Martin Marietta Laboratories, Baltimore, MD.
He has been with the Laboratories since 1964, and is responsible
for studying properties of the transition metal carbides, nitrogen
ceramics, and adhesive bonded structures. He was a Research Fellow
at the University of Warwick in 1968-1969 and obtained his Ph.D.
degree there in 1971. He has been a consultant on the Viking Mars
Lander program and the Space Shuttle program.

Geopges Vepahepy is Professor of Materials Science at Ecole

des Mines, Saint-Etienne, France. He graduated in Engineering from
Ecole poly technique, Paris in 1967, and obtained a Science
Doctorate from the University of Paris in 1973. He has worked in
the field of mechanics of composite materials and interfaces since
1967. He is a consultant in Mechanical Engineering and Composite
Materials.

S. S. Wang is Associate Professor at the University of

Illinois which he joined in 1977. He received his Sc.D. degree from
MIT in 1975 and stayed there for an additional two years as a
Research Associate and Lecturer. His work involves experimental and
analytical studies of adhesives and composites with special empha-
sis on advanced analytical and finite element analysis techniques.
ABOUT THE CONTRIBUTORS 921

In 1982, he received the Xerox Award for Outstanding Faculty

Research.

Thomas C. Ward is Professor of Chemistry at VPI&SU, Blacks-

burg, VA which he joined in 1968. He received his Ph.D. degree in
Physical Chemistry in 1966 from Princeton University and spent two
years in England with Prof. Manfred Gordon. He has received several
teaching honors, including the Wine Award and is a member of the
Academy of Teaching Excellence at VPI&SU. He has also been a major
contributor to the ACS polymer short course program since its
inception in 1976. His publications have been principally in the
areas of polymer characterization, testing, adhesion and long term
property evaluation of polymeric materials.

M. L. Wittiams is presently Dean of the School of Engineer-

ing, University of Pittsburgh having come there in 1973 from the
University of Utah where he was Dean of the College of Engineering
and Distinguished Professor of Engineering. Prior to these 8 years
in Salt Lake City, he had been at Caltech as Professor of Aero-
nautics since receiving his doctorate there in 1950. His main pro-
fessional research activities have been concentrated in the field
of general materials behavior and structural mechanics and design,
with a specific interest in both cohesive and adhesive fracture
which is reflected in his being Editor-in-Chief of the Internation-
al Journal of Fracture. He has lectured nationally and internation-
ally, for example as a Sigma Xi National Lecturer and as a Visiting
Professor at the Imperial College. He has served on many govern-
mental technical advisory committees such as the National Institute
of Dental Research, NSF, NASA and National Materials Advisory
Board.

Richard s. Wittiams is a Senior Research Engineer at the

United Technologies Research Center, E. Hartford, CT. He received
his Ph.D. degree in Materials Engineering Science from Virginia
Tech in 1975. His research has spanned all areas of NDT and experi-
mental mechanics, with special emphasis on ultrasonics and acoustic
emission, and has over 60 publications and presentations. He most
recently served as Editor of the ASNT Symposium on Ultrasonic
Reliability. He is an active member of ASNT. In addition to his
Associate Technical Editor position with Materials Evaluation, he
has served or is serving on a number of committees.
INDEX

Accelerated Procedures for Glue- Adhesive Bonded Joints (cont.)

Wood Bonds, 395-417 role of surface and bulk
Acoustic Method for Adhesion characterization, 3-17
Estimation in Filled Systems, fracture aspects of, 689-702,
440-445 703-737, 739-753, 755-787,
Adhering Systems (Single Adher- 789-807, 809-827, 839-853
end/Adhesive), 19-30, 85-102 influence of aging of film
effect of surface conditioning adhesive on, 231-246
of the adherend on the mech- influence of surface
anical behavior of, 85-102 roughness on, 69-84
practical adhesion measurement life prediction methodology
in, 19-30 for, 567-586
Adhesion Estimation Methods in model of the interfaces
Filled Polymer Systems, 433-450 formed in, 6
acoustic method, 440-445 nondestructive evaluation of,
compensation method, 435-438 347-367
volumetric method, 438-440 stress analysis of, 587-610
Adhesive three-point bend test for,
curing and viscoelastic pro- 381-394
perties by interface waves, Adhesive Bonded Thin Gauge
328-337 Sheet Metal Box-Section
Adhesive Bonded Joints (see also Beams
Adhesive Joints of) performance of, 627-637
characterization of Adhesive Bond Lines (Shear
by ultrasonic surface and Stressed)
interface waves, 307-345 effects of low cycle loading,
dimensional stability of, 137- 659-677
150, 151-164 Adhesive Bonding
effect of moisture on, 137- of contaminated surfaces,
150 259-266
durability of, 453-467, 469- of oily surfaces, 259-26~
484, 485-501, 503-521, 523- vibrational, 262-266
540, 541-553, 555-564, 694- Adhesive Bonding Process
701 integrating surface treatment
effect of scrim cloth on, 679- of bonding parts in, 257-
685 266
evaluation of Adhesive Failure Energy,748-750

923
924 INDEX

Adhesive Failure in Dentistry, Adhesive Test Methods, Evalua-

295-300 tion of, (cont.)
Adhesive Fracture rod pull-out, 707
continuum mechanics factors in, tensile bond, 283-286
703-737 torsional shear stress, 706
structural precursors to, T-pee1, 706
809-827 Adhesive Test Specimens,706-707
Adhesive Fracture Energy of Adhesively Bonded Composites,
Epoxy Adhesive Joints, 799- 639-658, 755-787
803 Adhesively Bonded Honeycomb
effect of bond thickness on, Structures
800-803 ultrasonic assessment of cure
Adhesive Joints of rate effects in, 369-380
aluminum (alloy), 195-212, 233, Ageing of Structural Film Adhe-
240, 245, 383, 453-467, 479, sives
485-501, 503-521, 616, 793, and effect on joint strength,
817 231-246
box-section beams, 627-637 Aliphatic Po1yamides-Meta1
composites, 639-658, 755-787 Bonds, 195-212
copper, 523-540 Alumina
glass, 694-701 fracture surface energies of,
graphite/epoxy, 639-658, 757 879
honeycomb structures, 369-380 Aluminum Adhesive Bonding (see
iron, 549 Adhesive Joint of)
microscope cover slip, 152 Aluminum Alloys, Oxides on
oily surfaces, 259- 266 and characterization nomo-
polyamide 11, 133 grams, 113-115
polyamide 6.6, 133 morphology of, 104-106, 107-
polyethylene (low density), 130 113, 455-456, 472
polyethylene (high density), polymer identification in,
130 473
polymers, 121-136, 692-694 surface characterization of,
po1yoxymethy1ene, 132 103-120
polypropylene, 132 surface impedance measure-
po1ytetraf1uoroethy1ene, 129 ments of, 107
steel, 7, 15, 69, 85-102, 165- surface potential difference
193, 195-212, 351, 429, 564, of, 106-113
793 ternary phase diagram of, 462
titanium, 8, 138, 158, 453-467, Aluminum Hydroxide
541-553 morphology of, 460
Adhesive Lap Joints Aluminum Polymer Bonds
impact strength of, 611-626 durability Qf, 453-467, 469-
Adhesive Test Methods, Evalua- 484, 485-501, 503-521
tion of, 269-287 Aminosi1ane Treatment
blister, 296, 304, 707 of titanium, effect on Ti/
cone, 706 epoxy adhesive joints,
double cantilever cleavage, 541-553.
706, 762, 794 of iron, effect on iron/epoxy
lap shear, 271-283 adhesive joints, 549
double lap shear, 706 Auger Electron Spectroscopy, 9
single lap shear, 706
INDEX 925

Benzotriazole (BTA) Coupling Copper Adhesion to Epoxy,

Agents for Copper, 523-540 523-540
Blister Technique, 296, 304, Copper-Alumina Joints
707 critical energy release rates
Bond Length of, 879
effect on monowire joints, Copper-BTA Complex, 524
178-179 Crazing in Polymers, 821-824
effect on multiwire joints, Creep of Epoxy Adhesives, 140-146
183-190 Critical Energy Release Rates
Bondability of Low Surface Energy in Metal-Ceramic Joints, 855-
Materials 869, 871-882
effect of surface treatment on, Critical Fracture Energy, 62
121-236 Critical Surface Tension of
Bonded Honeycomb Structures Wetting, 122
ultrasonic assessment of cure for various polymers, table,
rate effects in, 369-380 123
Bonding Parts Surface Treatment Cure Rate Effects, Ultrasonic
integration of, in adhesive Assessment of, 369-380
bonding process, 257-266 Curing of Adhesives
Bondline Stiffness, 387 investigation by interface
Box-Section Beams (Adhesively waves, 328-337
Bonded) Cyanamid FM73 Adhesive
performance of, 627-637 shear stress-shear strain
Bulk Characterization diagram of, 663
and evaluation of adhesive Cycle (low) Loading
joints, 3-17 effect of, on shear stressed
Butt Joint with a Flaw adhesive bond lines, 659-677
three dimensional analysis of, Cyclic Debonding of Adhesively
839-853 Bonded Composites, 639-658

Ceramic-Metal Crown, 290 Deformation of Adhesive Joint

Ceramic-Metal Joints, 855-869, study of, by Moire patterns,
871-882 151
Characterization Methods for Dental Adhesion Testing
Locus of Failure, 7-15 a critical appraisal of, 289-
Characterization Nomograms for 305
Representing Oxide Film Dentistry
Properties, 113-115 adhesive failure in, 295-300
Chromic-Sulfuric Treatment of Diallyl Phthalate (DAP)
Polymers, 124 epoxy IPN adhesives, 252-254
Composite-Composite Bond Joints structural adhesives based
cycling debonding of, 639-658 on, 247-256
fracture of, 755-787 Dimensional Stability of Adhe-
Contact Angles, Measurement of, sively Bonded Joints, 137-
125 150, 151-164
effect of surface treatment of effect of moisture on, 137-
polymer substrates on, table, 150
128 Double Cantilever Beam Specimen
Continuum Mechanics Factors in fractographs of, 765
Adhesive Fracture, 703-737 geometry of, 706, 762, 794
926 INDEX

Ductile Fracture Energy, 747-748 Electron Spectroscopy for Chem-

Durability Factor, 505 ical Analysis (ESCA), 10
Durability of Metal-Polymer investigation of adhesively
Bonds, 453-467, 469-484, bonded Al joints by, 479
485-501, 503-521, 523-540, Enamel-Composite Failures, 300
541-553, 555-564, 694-701 Enamel-Resin System, SEM of,293
in industrial atmosphere, 495- Energy Dissipation
497 effect on peel strength,
in salt fog cycle environment, 419-432
510, 513-514 Epoxide Resins (Filled)
in seacoast atmosphere, 497-500 interfacial properties of,
Dynamic Mechanical Analysis 883-904
and polyurethane coating Epoxy Adhesion to Copper, 523-
curing, 46-48 540
of aged epoxy adhesive, 239 Epoxy Adhesive
Dynamic Mechanical Relative Shear aging of, effect on joint
Modulus and Damping strength, 231-246
for epoxy bonded joint, 356 creep of, 140-146
for poly(allyl 2-cyanoacrylate) dynamic mechanical analysis
bonded joint, 359 of aged, 239
for poly(ethyl 2-cyanoacrylate) effective surface energy,
bonded joint, 358 fracture energy and frac-
for polyester/polysulfone block ture toughness of, 742
copolymer bonded joint, 357 HPLC analysis of fresh and
Dynamic Mechanical Storage aged, 237
Modulus and Damping modified (Epasol FV/ZIS939),
for epoxy free film, 355 259-262 .
for polyester/polysulfone block and bonding of oily sur-
copolymer film, 357 faces, 259-260
for poly(ethyl 2-cyanoacrylate) moisture absorption in, 148
film, 358 Epoxy Adhesive {Rubber or
Elastomer Modified}
Effective Fracture Surface Energy effect of loading rate on,
of epoxy adhesive, 742, 745 613, 618-619
Effective Shear Modulus, 337 effective fracture surface
Elastic Interface Waves, 321-328 energies and fracture
Elastomer-Metal Bonding, 555-564 energies of, 745
Electrochemical Pretreament of fracture energy of, 796
Copper fracture toughness of, 829-
and its effect on joint 838
strength, 523-540 replica TEM of fracture
E-Glass Single Filaments surface of, 791
epoxy resin wetting of, 51-65 Epoxy Adhesive or Bonded Joints
silane treatment of, 51, 54-65 (see also Adhesive Bonded
surface free energy of, 62 Joints), 8, 69, 121-136, 138,
work of adhesion on, 58 151-164, 165-193, 231-246,
Electron Energy Loss Spectroscopy 259-260, 356, 383, 485-501,
and polymer identification in 503-521, 523-540, 541-553,
aluminum oxide, 473 549, 639-655, 694-701,
799-803, 817
INDEX 927

Epoxy Bonded Joints of Fracture of Metal-Ceramic

aluminum, 233, 240, 245, 383, Joints, 855-869, 871-882
616, 793, 817 Fracture Properties of Filled
copper, 523-540 Epoxide Resins, 883-904
glass, 694-701 Fracture Toughness, 740
graphite/epoxy, 639-658, 757 of epoxy adhesive, 742
iron, 549 of epoxy (rubber modified),
microscope cover slip, 152 829-838
polymers, 121-136
steel, 69, 85-102, 165-193, Generalized Theory of Fracture
351, 429, 793 Mechanics, 689
titanium, 8, 138, 158, 459, Glass-Epoxy Bonds
541-553 effects of silanes on,
Epoxy Free Film 695-701
dynamic mechanical storage effects of water on, 694-701
modulus and damping for, 355 Glue Line Thickness
Epoxy Resin Wetting of E-Glass effect of, on monowire
Single Filaments, 51-65 joints, 176-177, 181
Glue-Wood Bonds
Failure, Locus of, 4-7, 10-15 accelerated aging procedures
Filled Polymer 9ystems for, 395-417
estimation of adhesion in, Gold-Aluminum Diffusion Couple
433-450 surface acoustic wave phase
Film Adhesives (Structural) shift in, 315
aging of, and effect on joint GPC Analysis of Nitrile Epoxy
strength, 231-246 Adhesive, 242
Finger Joints (Exterior Grade), Griffith Fracture Energy for
213-230 Glass-Epoxy Composite, 62
Finite Element Analysis of Groups in a Resin and Property
adhesive bonded joints, 648-652 Obtained, Table, 249
box-section structure, 635-637
butt joint with a flaw, 839-853 Hand Pulled Peel Test, 44, 49
single lap joint, 617-618, Honeycomb Structures
622-625 ultrasonic assessment of cure
thick adherend model jOint, rate effects in, 369-380
682-684 "Honeymoon" Adhesives, 213-230
Flexibilized Diallyl Phthalate Hot-Pressed Silicon Nitride
Adhesives, 250-251 Joints, 856
Fracture Aspects of Adhesive HPLC Analysis of
Joints, 689-702, 703-737, fresh and aged epoxy
739-753, 759-787, 789-807, adhesive, 237
809-827, 839-853 nitrile-epoxy adhesive, 244
Fracture Energy Hydrothermal Stability of
measurement of, 740-741 Titanium-Epoxy Adhesive
of adhesive joints, 739-753 Joints, 541-553
of epoxy adhesives, 742, 745
of glass-epoxy composite, 62 Impact Strength of Adhesive Lap
of thermoplastics and thermo- Joints, 611-627
sets, 743 Impact Torsion Test Rig, 615
928 INDEX

Industrial Atmosphere Low Energy Electron-Induced

bond durability in, 495-497 X-ray Spectroscopy (LEEIXS)
Infra-red Spectrum of Amino- and surface characterization,
silane on Titanium, 544-547 85
Inhibitors and the Durability of schematic diagram of, 87
Adhesive Bonds to AI, 463-464 Low Surface Energy Materials
Integrated Methodology for effect of surface treatment
Adhesive Bonded Joint Life on the wettability and
Predictions, 567-586 bondability of, 121-136
Interatomic Bonding Energies at suitable surface treatments
Interfaces and adhesives for, 134
mechanical measurement of,
689-702 Mechanical Measurement of
Interface in a Single Two Com- Interatomic Bonding Energies
ponent System, 5 at Interfaces, 689-702
Interface Waves, Application of Mechanical Properties of Joints
in investigating curing and influence of surface rough-
viscoelastic properties of ness on, 69-84
adhesives, 328-337 Metal-Ceramic Crown, 290
Interfaces in an Adhesive Bond- Metal-Ceramic Joints, 855-869,
ing System, 6 871-882
Interferometric Technique influence of layer thickness
and study of deformation of and internal stress on,
adhesive joint, 151 855-869
Intermediate Adhesive Layers, 196 Metal-Polymer Bonds (see
effect on adhesive strength, Adhesive Joints of)
197-212 durability of, 453-467, 469-
Interpenetrating Polymer Network, 484, 485-501, 503-521, 523-
249, 252- 255 540, 541-553, 555-564, 694-
DAP/epoxy, 252-255 701
Interphase, 5 Metal Surfaces
Intrinsic Fracture Energy, 741 interaction with phenol for-
Ion Scattering Spectrometry, 9 maldehyde resin, 35
Iron-Epoxy Adhesive Bonds and Modified Epoxy Adhesive (Epasol
Effect of Aminosilane, 549 FV/ZIS939)
and bonding to oily surfaces,
J-Integral, 746-747 259-260
Moire Fringes, 154
Lap Shear Test and study of deformation of
for adhesive joints, 271-283, an adhesive joint, 151
706 Moisture (see also Durability
stress analysis in, 273-283 of Metal-Polymer Bonds)
stress intensity factors for, effect of, on the dimensional
277 stability of adhesive bond-
Leaky Surface Waves, 316-317 ed joints, 137-150
Life Prediction Methodology for Mono and Multiple-Wire Steel to
Adhesive Bonded Joints, 567-586 Steel Joints, 165-193
Locus of Failure, 4-7, 10-15
characterization methods for, Naphthalene-Sodium Etching of
7-15 Polymers, 125
INDEX 929

Nitrile Rubber Modified Epoxy Polymers (cont.)

aging of, 232, 240-245 critical surface tension of
and tensile shear strength of wetting of, table, 123
AI-AI lap joints, 245 work of adhesion for SBR
GPC analysis of, 242 bonded to, 693-694
HPLC analysis of, 244 Polysulfide Sealant
Nondestructive Evaluation of adhesion of, 31-40
Bonded Joints, 347-367 effect of phenol formal-
dehyde resin, 31-40
Oily Surfaces crosslink density of, 33,
bonding by modified epoxy ad- 37-38
hesive (Epasol FV/ZIS939), cure rate of, 34
259-260 polyurethane coating
Optical Micrographs of Fracture interfacial aspects, 41-50
Surfaces in Ti/Epoxy Bonds, SEM study of, 44-46
551-552 pot life of, 34, 38
water absorption in, 34, 39
Particle Filled Epoxide Resin Polyurethane
fracture properties of, 883- chemistry, 42-43
904 coating curing by DMA, 46-49
Peel Strength and Energy Dis- coating-polysulfide sealant
sipation, 419-432 interfacial aspects, 41-50
Peel Test, Hand Pulled, 44, 49 Porcelain-Metal
Phase Boundary Sensitive Test. 19 failures, 299
Phenol Formaldehyde Resin interface, test of, 302
and interactions with metal specimens, 296
surfaces, 35 Practical Adhesion, 20, 32
effect of, on the adhesion of
polysulfide sealant, 31-40 Quality Control (In-Plant) of
Phenolic ("Honeymoon") Fast Glue-Wood Bonds
Setting Adhesives accelerated aging procedures
and exterior grade finger for, 395-417
joints, 213-230
Plasma Treatment of Polymers, 125 Rayleigh-Angle Method, 317-321
Polyamides Recursive Structural Identifi-
adhesive joint strength with cation Method, 370
metals (AI, steel), 195-212 Resin-Enamel/Dentin Interface
Polybutadiene-AI Powder System test of, 302
adhesion estimation in, 444 Resin Properties vs. Groups in,
Polymer-Metal Bonds (see Metal- table, 249
Polymer Bonds) Rod Pull-out Test, 707, 716,
Polymeric Films (Thin) 719-721
and metal-polyamide joints, Roughness (Surface)
195-212 influence of, on mechanical
Polymers properties of joints, 69-84
bondability of, 121-136, parameters, 72
692-694 Rubber-Metal Adhesion, 555-564
contact angles of, effect of
surface treatment on, table, Salt Fog Cycle Environment
128 bondability in, 510, 513-514
930 INDEX

Sapphire/Copper Joints, 871-882 Structural Adhesives

Scrim Cloth based on diallyl phthalate,
effect of, on adhesively bonded 247-256
joints, 679-685 form and composition of, 248
Seacoast Atmosphere Structural Film Adhesives
bond durability in, 497-500 aging of, and effect on joint
Secondary Ion Mass Spectro- strength, 231-246
metry, 9 Styrene/Isoprene/Styrene Tri-
SEM of block Copolymers, 353
Cr0 3 anodized Al surface, 105 dynamic mechanical damping
enamel-resin system, 293 of, free films and bonded
fracture of sapphire joint, 880 joints, 362-364
fracture surfaces of filled Surface Acoustic Waves, 310-316
epoxy resin, 890-894 phase shift in the Au-AI
fracture surfaces of FM-73 diffusion couple, 315
adhesive, 681 Surface Behavior Diagram, 461-
fracture surfaces of Ti/epoxy 462
lap joints, 551 Surface Characterization
sealant-coating interface, and evaluation of adhesive
44-46 bonds, 3-17
Shear Strength of Glass Fiber/ methods for locus of failure,
Epoxy Resin, 51 5-15
Short Beam Shear Test, 384_ of anodic oxides on Al
Silane Treatment of alloys, 103-120
alumina particles, 884 Surface Free Energy of Silane
E-glass filaments, 51, 54-65 Coated Fibers, 62
wetting of, 58 Surface Impedance Measurements
glass, 695-701 of anodic oxides, 107, 116-
silica particles, 884 117
titanium adherend, 541-553 Surface Potential Diffeerence
Stainless Steel (AISI-304) of anodic oxides, 106-113
and epoxy joint, 85-102 Surface Roughness
effect of surface conditioning influence of, on mechanical
upon, 85-102 properties of joints, 69-
effect of surface conditioning 84
on oxide thickness of, 93 parameters, 72
LEEIXS spectra of, 95 Surface Treatment of Bonding
Steel, Adhesive Bonding of, 7, Parts
15, 69, 85-102, 165-193, 351, integration in the adhesive
429, 793 bonding process, 257-266
Steel to Steel (Mono and Multi- Surface Treatments of Low Sur-
ple-Wire) Joints, 165-193 face Energy Materials, 124-
Stone ley Wave, 321 125
Stress Analysis of Adhesive effect of, on contact angles,
Joints, 587-610 table, 128
Stress Energy Release Rates
and cyclic debonding, 639-658 Tapered Double Cantilever Beam
Stress Intensity Factors for Lap Adhesive Joint Specimen, 794
Shear Tests, 277 Tensile Bond Tests, 283-285
stresses in, 284
INDEX 931

Theory of Attachment Site, 35 Viscoelastic Properties of

Thermoplastics Adhesives
ductile fracture energy of, 748 investigation by interface
fracture energies of, 743 waves, 328-337
Thermosets
fracture energies of, 743 Weak Boundary Layer, 4, 6, 11,
Thin Films 35, 337, 448, 480
application of surface wave
method to, 315 Wedge Test, 12-14, 459, 463-464,
Thin-Gauge Sheet Metal Box- 473
Section Beams Wettability of Low Energy Surface
adhesive bonding of, 627-637 Materials
Three Point Bend Test for effect of surface treatment on,
Adhesive Joints, 381-394 121-136
Three Point Flexural Test for Width Tapered Beam Specimen
an Adhering System, 19-30, 85 fractograph of, 771
Titanium-Adhesive Bonding, 8, geometry of, 767
138, 158, 453-467, 541-553 Wire (Steel Mono and Multiple)
Titanium Pretreatment Pro- Joints, 165-193
cesses, 458 Wood-Glue Bonds
accelerated aging procedures
Ultrasonic Assessment of Cure for, 395-417
Rate Effects in Bonded Honey- Work of Adhesion
comb Structures, 369-380 for SBR bonded to various
Ultrasonic Surface and Interface plastic substrates, 693
Waves on glass filaments, 58
and adhesive joint character-
ization, 307-345 X-ray Photoelectron Spectro-
Underwater Sonar Systems scopy (see ESCA)
elastomer/metal bonds appli-
cable in, 555 Yield Strength of the Inter-
face, 387
Vibrational Adhesive Bonding,
262-266