PROCEEDINGS OF THE CRACK

PROPAGATION SYMPOSIUM

CRANFIELD

SEPTEMBER 1961

Vol. 1
 - 76 -

.A Two Stage Process of Fatigue Crack Growth

-by-

P.J.E. Forsyth
Royal Aircraft Establishment

SUMMARY

A study of the initiation and growth of fatigue cracks has indicated that several
complex processes are involved. The observations suggest why the present
mathematical theories based on simple models are likely to be only of very limited
application to the practical state of affairs.

The extension of a fatigue crack can occur in several ways. Initially it may grow
by the process of its formation, that is, by an 'unslipping' or reverse glide mechanism
forming a surface crevice which deepens with time. This unslipping mechanism has.
been designated Stage I fatigue growth. The circunistances favourable to its active
continuance will be outlined in the paper. It is characterised by crystallographic
fracture facets changing angle with orientation at grain boundaries. Because it is a
shear stress dependent process these facets lie .on or near the planes .of maximum
shear stress. In the general case, Stage I fatigue is superseded by a second stage.
This second stage cracking occurs in a plane perpendicular to the maximum tensile
stress. In most light alloys the fracture shows characteristic striations which are of
diagnostic value. Becuase it can be demonstrated that these striations delineate
successive positions of the crack tip, and their spacing represents the crack growth
for each cycle of stress, they are additionally useful in determining crack growth
history in this stage. By various observational techniques it has been possible to
analyse the nature of these striations, and present a tentative explanation of their
formation. Their appearance, and possibly the detail of the mechanism whereby
they are formed, differs for various materials, and with both environment and
frequency of cyclic stress. Ih the technologically important materials such as the
high strength aluminium alloys the extension of the crack tip in one cycle of stress
involves both a brittle and plastic growth period. The conditions favouring each .of
these components are considered.
 .- 77 -

Introduction

The prime evidence of fatigue failure is a crack, and the fatigue endurance
. is commonly divided into (a) the crack initiation period and (b) the crack growth
period. Evidence for the early formation of fatigue cracks continues to accum-
ulate and therefore this arbitrary division is being forced back further and fur-
ther in time towards the origin of the process. 1f WI> accept that the common
mode of crack formation is by the deepening of a slip band groove by a,n atomic
process of dislocation movement, then the continued use of this division presents
certain difficulties, and leads to the absurdity of trying to specify a minimum
crack size in a process starting on an atomic scale. The test applied in these
circumstances is to try to prise open the questionable slip band· with a tensile
stress, and if it gapes open then the band must have contained a fatigue crack.

This paper presents evidence for what is considered a rational division which
has some physical meaning based on the mode of crack growth. It will be
shown that there are basically two modes of crack growth, and that the change
from one to the other is a natural division in behaviour. As we are not concern-
ed in ·this context with safe fatigue conditions, although it is known that non-prop-
a,gating cracks sometimes exist in this state, we can label the two behaviours
stage I and stage II crack growth. Precracking damage will be discussed only
in context With the process of growth of an existing crack.
Fatigue cracks, as a general rule, originate in slip bands, although both
grain boundary and sub grain boundary initiation are not uncommon. These ex-
ceptions to the rule are fairly well understood and are explicable in terms of
known grain boundary behaviour and environmental effects. Their importance
is not denied, but in order to make a general classifi.cation of behaviour they
will be discussed separately as subsidiary effects.

Stage I fatigue crack grow~

The criterion for slip band cracking is the range of reso!ved shear stress
on the slip plane. Thus cracks will form on those planes most closely aligned
with the maximum shear stress directions in the component or fatigue specimen.
For various reasons the surface grains behave in a soft mannerJ and in most
cases the stress is a maximum at the surface. Therefore it is not surprising
that the surface grains deform more than those in the interior of the metal.
Furthermore, fatigue cracks originate at the free surface of these grains. The
surface is not only a favoured place for dislocation movement, but also for the
irreversible action of forming slip band grooves. The most important differ-.
ence between slip bands formed by steady stress and those formed by cyclic
stress is that the cyclic stress produces a slip band groove. The groove deep-
ens with continued cyclic stress to form a crevice. or intrusion. There is no
evidence that the mechanism of crack extension changes while it follows the
active slip plane. This mechanism of crack growth can persist for an apprec-
iable proportion of the endurance, but whether this proportion is 'large or small
it will be designated as stage I fatigue crack growth because it differs 'in an
essential manner from later behaviour.
The crack can be studied in this early stage by means of various metall-
ographic techniques such as taper sectioning,. replication methods, or by study-
ing the fracture surface itself. The intrusion or crevice forming process is
often accompanied by extrusion of thin metallic slivers from the slip bands,
sometimes in a closely associated manner as though some conjoint actiOn was
involved. There are many clear indications that the intrusion or crevice is
formed by reverse slip, i. e. slip occurring in a preferred direction in neighbour-
ing packets of slip planes. The behaviour envisa,ged is similar to that suggest-
ed by Shanley in his so-called 'unbonding 1 process of crack formation and
 - 78 -

demonstrated by. various other workers (1) (2) (3). We know by direct obser-
vation that extrusion occurs in the direction of slip, and it has been assumed
that the crevice or crack also grows in the slip direction. Direct evidence
has now been obtained from the study of the crack surface which substantiates
this point. Fig. 1 shows an extrusion in an aluminium-zinc alloy where the
surface markings indicate the direction of slip and extrusion, and Fig. 2 shows
features on the surface of an aluminium-zinc-magnesium alloy in the first stage
region. These features which seem to be the traces of jogged slip again in-
dicate the direction of slip, and consequently the direction of crack growth.
The shape of the crack front in this stage of growth is difficult to ascertain,
but the evidence from a transparent non-metallic material such as silverchloricfe ..,
shows thst it can be sharply serrated, pushing forward at many places in a
spear-head attack. Fig. 3 illustrates in a diagrammatic form, this type of
cracking. Although extrusions rarely exceed 20 microns in length in the direc-
tion of growth, the crevices may extend crystallographically across many grains
with tilt of the fracture path as the crack crosses a boundary from. one grain to
its neighbour. This produces a facetted texture, the general fracture plane
being one of maximum shear. These facets may be on!y slightly disoriented
in aluminium alloys which is probably related to the fact that fatigue slip can
occur in these materials not only on the (111) but also on (11 0) and (1 00) planes
(4). With this selection of planes the final crack path may be relatively planar.

If the two dislocation processes of extrusion and intrusion are similar, but
working in. opposite directions, it may be questioned why the former process al-
ways quickly ceases, whereas the latter may continue for a relatively long "time.
We know that in aluminium-4o/o copper alloy the extrusion process starts sudden-
ly, and stops after only a few cycles, in most cases less than 100. This
effect has been observed directly, and there are certain features in the struct-
ure of the extrusion which seem to confirm this. It seems thst while the
crystal surface is essentially plain, extrusion and intrusion have similar opport-
unities for operation, but the formation of an extrusion effectively notches the
material and 'unlinks' the extruded region from the stress. The tip of the cre-
vice becomes a very active dislocation source, and these sources feeding the ex-
trusion process cease to operate. It is interesting to note that in the case of
torsion, filamentary extrusions can occur many hundreds of microns in length.
In torsion the shear can be transmitted during both halves of the stress cycle
by frictional forces developed between the crack faces, there being practically
no tension component to part them. Similarly a directly stressed specimen, if
given a compressive mean stress will produce more marked extrusion than one
loaded with zero mean stress. The newly formed crevices now gr<>W by the
operation of dislocations at the tip. This is. probably enhanced by the cessation
of operation of the other sources in the vicinity. As there is a valid reason
for the cessation of extrusion, and an enhancement of the intrusion process, there
seems to be no reason why the latter should not continue indefinitely. It will
be shown that it eventually ceases in favour of a second stage growth process,
usually when it meets a slip obstacle such as a grain boundary. In single cry-
stals or in cold worked pure metals whose preferred orientation makes the grain
boundaries less formidable dislocation barriers, this Stage I process may continue
to final fracture. These crack surfaces are often remarkably free from fretting
which suggests that the crack walls gape and the stress concentration operates
quite freely in compression as well as in tension. This may be why it is the
resolved shear stress range that is the important criterion in this stage.

The effectiveness of reverse slip processes in extending the crevice .will de-
pend on the operation of some mechanism whereby reversal of .dislocation move-
ment is blocked. This might happen, as has been suggested by several people;
by surface contamination by gas atoms, the formation of oxide, and subsequent
locking of dislocation sources. As this surface locking process would be expect-
ed to work equally over the whole crack tip surface, it would seem more likely
that it is achieved by the interaction of dislocations themselves.
 - 79 -

A mechanism similar to that suggested by Cottrell & Hull (5) for the extru-
sion effect, seems to operate at the crack tip of pure aluminium growing by the
Stage I process. This is illustrated in Figs. 4 and 5, There is no doubt
that the crack tip is an active dislocation source, and it can be demonstrated
that slip operates in two sets of planes originating from the crack tip. One
set operates under the tension half cycle. The other set operates under. the
compressed half cycle i. e. when the shear directions are reversed.

Conditions favouring stage I crack growth

Certain conditions of stressing and environment favour this form of crack

growth sometimes to the complete exclusion of the 2nd stage. Low stresses
resulting in slow growth favour stage I behaviour. · Mean tension stress will
encourage the changeover to stage II which suggests that there is· a minimum
tensile stress necessary for stage II growth. Corrosion-fatigue conditions at
stresses below the air fatigue limit may cause almost complete failure by slip
plane cracking as shown in Fig. 6. This suggests that the effect of corrosion
is to aid the unbending process probably by film removal, or by the dissolution
of crack tip material containing dislocation obstacles such as locked groups of
dislocations. In the aluminium alloys considered in this context, the active
slip may be accompanied by local chemical changes such as precipitation des-
truction, and solute disordering. The slip band crack usually changes from
the slip plane to stage II growth i. e. in a plane perpendicular to the maximum
tensile stress, when it reaches a grain boundary. The slip band cracks grow
essentially on the single slip plane or planes containing a common direction,
the slip direction. Occasionally it deviates in an alternating manner between
two favourable oriented planes as shown in Fig. 7. Although a second plane
is shown to be operative in the dislocation bloc!dng process, it would seem where
easy glide can occur the cracks grow a fair distance along a plane of one part-
icular index. If the material and the conditions of stressing favour duplex slip
then the stage I process gives way to the stage II process. If the material it-
self particularly favours duplex slip deformation e. g. an averaged aluminium
alloy containing microscopically visible precipitate, then stage I may never de-
velop in detectable form even under corrosion fatigue conditions. Stage II
crac!dng will occur if the stress conditions favour duplex slip. These conditions
are periodic high stress cycles, notched conditions, obstacles to easy glide block-
ing the crack tip, and the growth of the crack into the depths of the specimen
where conditions of constraint exist~ and the ratio of shear stress to tensile
stress is low. The changeover is not necessarily sudden~ the slip band crack-
ing sometimes decreasing in a gradual manner. Fig. 8 shows the changeover
(arrow A). It can be seen that although a weakened path has been produced by
the slip process the crack still deviates to the normal plane.

Stage II fatigue crack growth

Whereas stage I growth is governed by the local shear stresses and the free-
dom of the dislocations, the criterion for growth in the second stage is the value
of the maximum principal tensile stress operating in the component or specimen
ln the region of the crack tip. Fig. 8 showed that the change over could occur
even when an apparently easy path for slip plane cracking existed. This appar-
ent anomaly may be because of the changing value of (3 the shear stress/tensile
stress ratio with depth ih the specimen. This indicates that the peak tensile
stress developed across the crack tip causes deviation of the crack from its slip
path. If this is so it suggests that stage II cracking probably contains an ele-
ment of cleavage. It will be shown later that certain observations have been
made which confirm this view. The stage II mode of crack growth is charact-
erised by microscopic features, the fracture striations, which may become visi-
ble as soon as the crack starts to deviate from the crystallographic path as
 - 80 -

shown diagramatically in Fig. 3. These features have proved to be of consid-

erable diagnostic use in investigating service failures, and have been described
elsewhere (6). Their appearance varies in detail from one material to another
and these variations can be related in a qualitative wa:y to the plastic behaviour
of the material. Because the crack in that stage is governed by the direction·
of the maximum tensile stress, and there is an element of cleavage, this pRrt
of the fatigue fracture will be fairly smooth compared with the final tensile part of
the fracture. This smoothness and reflectivity is the result of the crack grow-
ing across many small plateaux, and although the levels of the plateaux may dif-
fer they are substantially co-planar. Because their levels differ, a consider- '
able area of the fracture may consist of fairly steep .cliff edges joining up the
plateaux. The nature of the surface is illustrated in a schematic form in Fig. 9·.
The striations are ridges in the surface of the fracture, and it has been
proved that each striation results from one stress cycle and the spacing of the
striations represents the local crack growth during a stress cycle. Crack
fronts do not extend at a constant rate over their whole length, local differences
in rate are caused by straightening and local advancement. The evidence for
this 'one for one' relationship between striations and stress cycles, is best seen
in programme loaded specimens where the discrete cycles of the programme can
often be indentified directly from· the fracture surface, as shown in Figure 10.
Because the striations are ridges on the surface, and delineate the successive
positions of the crack tip, it has been assumed that they are the result of local
plastic deformation at the crack tip with a resultant periodic deviation of the
crack path. This being so it is not surprising that they are most clearly re-
vealed in ductile materials. Again, because their formation depends on local
plastic deformation the crack tip stress will govern the amount of deformation.
It has been mentioned that there is a component of cleavage present in the stria.-
tion, and the proportion of cleavage to ductile growth will affect the appearance
of the striation as illustrated in Fig. 11. A high tensile mean stress super-
imposed on the peak positive half of the alternating stress will encourage the
cleavage component, whereas a large alternating stress range Will favour the
plastic deformation. component. The two types of behaviour are shown in Figs.
12 and 13. ·

There is propably no basic difference in behaviour between stage I cracking

and the ductile component of the stage ll striations. The important difference
is the presence of the cleavage component which alters and governs the crack
path. Not only does stress level contr<>l the amount of deformation at the crack
tip, but it also controls the distance the cleavage crack grows each cycle.
However, these two components can be made to operate independently by changing
stressing and environmental conditions. Thus a direct linear relationship bet-
ween spacing and stress has been obtained based on ~tatistical counts taken on
appr.opriate positions on fracture surfaces produced at different nominal stress
levels (7). Local variations in spacing will occur depending on local stress con-
centrations resulting from the irregular crack front, thus the local stress concen-
tration could be deduced from the striation spacing in the region. At low stress-
es nearet- the crack origin or the change from stage I growth, the striations ma:y
not be resolved, either because ·of their close spacing, or because of the virtual
absence of deformation at the crack root, and therefore the absence of a fracture
surface ridge. During this stage· of growth the fracture surface may appear
featureless except for river patterns characteristic of cleavage. It therefore
seems that at these early stages the fracture has all the characteristics of cleav-
age Which can again be identified in the cleavage component observed as a separ-
ate part of the striations formed in the later stages of crack growth. The gen-
et-al conclusion from thi.s must be that slip is virtually absent at the stage that
cleavage sets in, dying out as the crack deepens along the slip plaJ1e. The
cleavage component then predominates and causes growth for a period with the
later appearance of a ductile component which continues to increase t.,o final frac-
ture. These changes in crack behaviour could be related entirely to changes in
 - 81 -

the shear stress to tensile stress ratio /3. The tensile stress ,will increase to the
fracture stress but the effective shear stress value will
. be amax -qy>e
. ld' The crack
starts from the surface where the grains will have a lower than average shear str~ngth
then grows into the constrained region of the bulk of the specimen where brittle cleavage
conditions are favoured. The crack then grows to a stage where the stress has risen
to a value at which deformation again occurs effectively, plastic flow occurs at the,
crack tip, and the conditions of constraint continue to decrease' In materials where
brittle boundary fracture is common this cleavage period.is replaced by a brittle
boundary fracture period with stage II striations recurring at a later stage. The
brittle boundary facets themselves may contain faintly marked striations indicating
that a slight aniount of plastic deformation is occurring at the crack tip even during this
stage. This is discussed in the paper.

Frequency effects in stage II cracking

Reverse plane. bend fatigue tests on an aluminium-7. So/o zinc-2. So/o magnesium
alloy have showh that if the test frequency is changed from 1800 cycles per min-
ute to 20 cycles per minute, the striation spacing increases by about a factor of
10. This change in spacing represents an increase in crack tip speed of the same
order. A comparison of the two sets of striations showed that the low frequency
ones contained more olea vage component including the characteristic river patterns
which were absent on the high, frequency striations. These features are all ill-
ustrated in Fig. 14 which shows a specimen that had been alternately fatigued at
the two frequencies. It is well known that there is no such marked frequency
effect on endurance, a decrease of life of 30% being reported by Schijve (8) for
a change in frequency of about. the same ratio. However, the fractographic
effects were observed at an advanced crack stage, and are likely to be most ac-
tive under conditions of triaxial stress where cleavage would be encouraged.
Thus the frequency effect might only operate over a limited part of the life his-
tory of a specimen, and may have a minimal effect in sheet materials. The
importance of this observation on the effect of frequency seems to lie in the fact
that a few low frequency cycles at an advanced crack stage may be particularly
damaging because of the initiation pf the cleavage .type striations.

Corrosion fatigue
The importance of corrosion fatigue lies in the f;>ct that failures occur, under
the conjoint action of corrosion and fatigue , at stresses very much lower than the
air fatigue limit. A comparison of fracture behaviour of air and corrosion fat-
igue specimens showed that a continuity of behaviour existed whi.ch was related to
the stress level of the test. For equivalent stresses. the air and the corrosion
fatigue fractures showed a similar mode of fracture although the microscopic
features were somewhat different. At very low corrosion fatigue stresses stage
I cracking predominated, and as the stress was raised stage II cracking set in
earlier. Thus a much greater proportion of stage I cracking can be obtained
under corrosion-fatigue ·because it allows cracking to occur at very low stresses
which are unfavourable for the stage II process.
At the higher stresses· where the air and the corrosion fatigue S/N cur.ves
approach one another the .. fracture modes are similar but the striations formed
in stage II are noticeably different. The striations produced under air are
much more regular than tll.ose produced under corrosion the corrosion fatigue
striations i're characteristically brittle in nature with many river markings a<l
previously described. They are also remarkably similar to the low frequency
fracture striations. Thus it would seem that the action of the corrodant and
the slow application of stress similarly encourage cleavage fracture in these al-
uminium alloys. These cleavage facets have :been identified as (100) planes
both by back r.eflection Xray and etch pit techniques. Fig. 15 shows local
 - 82 -

changes occurring on the fracture surface as the crack deviates from the normal
fracture plane to the cleavage plane. This is also illustrated schematically in
Fig. 16. Fig. 17 shows the cleavage striations in more detail.

Fatigue crack growth in sheet materials

The growth of cracks in sheet materials is a subject of particular interest

to the designers of aircraft structures. The conditions of working are those of
relatively high mean stress with a superimposed small cyclic stress. The test
pieces used to simulate working conditions are usually centrally slotted panels
With the loads applied directly across the slot. Under the conditions of loading
commonly used the crack starts very early in the fatigue life, most of which is
taken up with its growth. The loads used may be applied at a constant level or
applied as a repeated programme of varying loads. The combined effect of a
notch and a high mean stress usually encourages the immediate onset of stage II
growth with no indication of initial slip plane growth, thus the fracture striations
may often be traced almost to the origin of the fracture. Most of the materials
of interest to the aircraft designer are clad with pure aluminium or an alumin-
. ium -zinc alloy and the striations are particularly well defined in this layer. It
has been possible to count the striations and construct a crack growth curve
which agrees well with the curve constructed from surface observations. Apart
from the local jumps in crack growth resulting from local stress concentrations
or irregularities in tile crack front, larger scale jumps are sometimes observed,
particularly in the very strong aluminium-alloy to specification D'fD687, This
intermitt.lnt crack growth is the result of alternating tensile fracture jumps and
periods of more gradual stage II fatigue growth. These separate periods of
growth are revealed on the fracture as differences of texture as shown in Fig.
18. It can be seen that as the crack grows the front bulges forward advancing
more rapidly in the central zone of the sheet. The bright textured areas are
fatigue growth, and the dull areas are tensile fracture. . The tensile crack
growth has a tendency to tunnel down the centre of the sheet whereas the fatigue
crack growth favours the surface layers. This is illustrated schematically in
Fig. 19. VFor a considerable period the fatigue crack continues to run in a
plane at right angles to the plane of the sheet, as the crack tunnels, the front
becoming an acute arrow h"ad form, a double 45° fracture surface appears which
eventually changes over to single 45° fracture. Stage II striations can be de-
tected on the early part of the 45° face, which is in agreement with surface ob-
servations of the crack growth rate. The crack may sti).l grow in a non cata-
strophic manner, the fracture showing no clear change in appearance on attaining
the critical crack length. Less ductile materials may grow fatigue cracks in
the normal mode, and fail catastrophically with no change of ·fracture angle, ex-
cept perhaps in the surface layers of tlie sheet. In the materials with less
ductile behaviour where there is a particular tendency for the intermittent behav-
iour with periodic changes from fatigue to tensile fracture, advanced damage in
the form of cracked particles is often detected in the boundary zones Where the
fatigue crack is becoming unstable and changing to the tensile mode. This pre-
crack damage is naturally enough associated with the central zone of the materlal
where the tensile crack is tunnelling.· However, the tensile crack growth once
started does not necessarily extend catastrophically, apparently re-establishing a
stable crack front shape from which only the fatigue stress can cause further
growth. The explanation for this may be that as the tensile crack tunnels for-·
ward it is retarded by the plastic deformation it has to achieve in the surface
layers, the more acutely shaped it becomes the greater the length of crack fi'ont
involved in this plastically deforming layer. 'I:hus an energy balance is achieved

Pl'ecFacking damage
Various forms of metallurgical change have been observed to occqr, usually
 - 83 -

in the vicinity of the fatigue striations or particularly active slip planes. The
more plastic deformation induced by the fatigue stress the more widespread the
precracking damage is likely to be. Experiments on the effect of removing
surface layers on residual fatigue strength, confirmed by direct metallographic
Observations, show that most of the damage. even in homogeneously stressed
specimens, occurs in the surface layers of uncracked specimens. When slip
pla:qe cracks form then the damage may continue to develop in a narrow zone
ahead of the crack tip. This zone may sometimes grow deeper than its asso-
ciated slip plane crack, the crack deviating from its damage plane when influ-
enced by changing stress conditions. No widespread zOne of damage has ever
been observed in stage II cracking which is in any way comparable with that
found in stage I. However, stage II produces damage not ahead of the crack,
but at the tip of the crack associated with the plastic deformation, and this
damage does not appear to differ in any way from that which can be observed
at the point of fracture of a tensile specimen in the same material. These
observations are reconcilable with the fact that in stage I which may be a re-
latively long period. dislocation movement is more int~nse and continuously act-
ing on a few small slip zones. In the stage II growth process the crack is
continually growing into unaffected material, , and each increment of growth is a
small tensile test on practically unaffected material. Figs. 20 and 21 show
stage II damage revealed in two different materials with different techniques.
Fig. 20 illustrates sub-structure revealed on the fracture surface of a pure
aluminium fatigue specimen by a very light electropolish followed by etching.
This layer was only several microns deep and microsections prepared to show
the material ahead of the crack tip revealed no advanced damage. The fact
that the sub- structure delineates the stage II striations confirms that the two
features are closely aSsociated with one another. Similarly Fig. 21 shows the
localised nature of the deformation in an aluminium alloy. The slip deform-
ation has been revealed by a subsequent ageing treatment which causes slip band
precipitation. It can be seen that the only appreciable slip is that associated
with the striations themselves.

In most complex commercial alloys such as those used for ~ircraft skinning,
the larger intermetallic particles may fracture in the vicinity of the crack tip.
This form of damage is associated with high mean stresses, and large amounts
of crack tip deformation. The stress raisers so formed may initiate fatigue
cracks in advance of the main crack, but there is as yet no evidence that this
behaviour occurs at a greater depth than a few tens of microns ahead of the
main crack tip. Fig. 22 shows evidence for advance cracks in DTD683 alloy.
As final fracture is approached the advanced damage becomes more extensive
and the process of void formation sets in. Fig. 2~ illustrates an aluminium-·
4'l'o copper alloy tested in the solution treated condition, the observed features
being near the point of final fracture. Arrow A indicates deep voids formed
from cracked particles. The stage II crack growth striations show geometric
growth which would appear to be related to the brittle fracture component.

Non -propagating cracks

It is of some interest to see if these observations on two stage crack prop-

agation assist in any way to explain the phenomenon of the non-propagating
crack. This' form of crack occurring as it does in practice from the root of a
notch, would be expected to start almost immediately as a stage II crack. An
inspection of the various micrographs presented in various papers by Frost, (9)
(1 0) suggests that this is so; it also reveals that these cracks usually stop on
turning into the 45° mode i. e. they change to growth in the stage I mode, and
then effectively become non-propagating. Orowan (11) has pointed out the im-
portance of the straightness of the crack along its tip as regards its effective-
ness as a stress raiser, and the fractograpbic evidence shows that the growth
mode may change from place to place along the crack front. The behaviour of
 - 84 -

the crack is therefore governed by the state of the majority of the crack front."'
If it can be established that a crack becomes non-propagating when most of the
lengthofits front has turned into the stage I mode, then we have a useful in-
dication of the local conditions existing in the material. It is very probable
that service fractures continuously change their fracture mode with the various
stress levels encountered. because it has often been noticed that the fracture
surface of programmed specimens changes its angle with stress level taking up
a shear type path at the lowest stress levels of the programme. In both the
programme and the service case the recurrence of sufficiently high stresses will
restart the crack growing in the stage II mode.

It has been stated elsewhere that the changeover from stage I to stage II
occurs when the shear stress/tensile stress ratio reduces sufficiently to discour-
age the effective dislocation movement required for slip plane cracking and at the
same time encouraging cleavage conditions. For this reason a crack growing in
an un-notched specimen having once started to grow in the stage II mode is un-
likely to revert to stage I again, although as the section is reduced a rapid shear
type growth may occur. We are faced with not only variations in the ratio, but
in the value of the shear stress, and it seems that the mode of growth the crack
will follow depends on both factors. When the crack, growing initially from a
notch, reaches a certain depth, the tensile component which had encouraged it to
start as a stage II crack is now reduced and this form of growth cannoi be main-
tained. However, because conditions of constraint are such that easy slip plane
growth is impossible the crack becomes virtually non-propagating. It has been
suggested that these cracks do in fact grow very slowly although for all practical
purposes they have stopped. This is very likely to be true because they do seem
to turn over to the shear mode where very small dislocation movements could
maintain growth, although this is less likely with a strain ageing material. Again,
any atmospheric corrosion would encourage the slip plane growth as a corrosion-
fatigue crack.

Fatigue fracture through grain boundary paths

In certain circumstances grain boundaries may behave in a particularly weak

manner, and cracks may originate in them. This form of boundary cracking has
been studied in an aluminium-zinc-magnesium ter'nary alloy heat treated to a
particularly susceptible grain boundary condition. Even at room temperature
boundary sliding could be induced in this alloy which developed into complete boun-
dary fracture with time under load. Similarly fatigue cracks could, if the stress
was high enough, start in the grain boundaries, and extend to almost complete in-
tercrystalline fracture. Stage I fracture has been defined as that resulting from
the movement of dislocations, commonly along a slip plane. It seems very likely
that a stage I type of cracking can occur at an obliquely inclined grain boundary
by reverse boundary sliding. This idea is substantiated by the fact that grain
boundary extrusion has been observed in an alwninium-4o/o copper fatigued at an
elevated temperature. Conditions in these materials favour stage II cracking because
local high tensile stress concentrations are likely to result from boundary sliding.

* Microsectionssuggest that most cracks occasionally travel for short distances in

the stage I mode, and it might he argued that such occasions existed in Frost's
specimens several times before they eventually become non-propag8.ting. How-
ever, it must be remem that a microsection only reveals two dimensions.
and we are only observing the nature of the crack tip at one particular position.
 - 85 -

Therefore it is unlikely that stage I cracking will persist long before stage II
takes over. . It has already been stated that striations have been observed on
boundary facets, the boundary fracture now taking the place o! the cleavage
component. A certain amount of plastic deformation is still occurri.ng at the·
root of the growing crack, the amount depending on the local ductility i. e. on
the state of the grain boundary resulting from ageing. Under these conditions
it would seem that th& rate o! crack growth is particularly dependent on the
metallurgical state of the grain boundary rather than that of the bulk material.
There should be interesting differences in behaviour between a material cracking
by cleavage, and another by an intercrystalline path, when fatigued in various
active environments. The two types of behaviour could probably be achieved in
the same material by varying the heat treatment.

Discussion

The relative importance of the two fatigue stages depends on the geometry
of the specimen or component, the conditions of stressing, and the environment.
The stage I process is important at low stresses with aggravating conditions such
as corrosion, because it prepares the material for stage II cracldng producing
the right sort of stress concentration to the required depth in the material.

In steels that show air fatigue limits, it is reported that slip band cracks
develop at stresses in the safe range, but they do not grow. Thus the fatigue
limit appears to be that stress below which fatigue cracks do not propagate.
However, in the presence of a corrodant these fatigue cracks will grow at ex-
tremely low stress.es, and the material will show no true fatigue limit. This
suggests that stage I cracking can be maintained in the presence of a corrodant,
whereas in air dislocation movement ceases. and the cracks become non..:propa-
gating. This fact is substantiated by the obserVation that in some aluminium
alloys a corrodant encourages the advancement of stage I cracks, at low stresses
to the complete exclusion of the stage II growth. Stage II growth contains a
component of plastic deformation, and therefore conditions whlch affect stage I
might produce a similar if limited effect in the second stage. The limitation
. will be that the plastically deformed zone· is formed in one cycle so that any
damaging process must, to be effective, act very rapidly, because after the in-
crement of craclting that element is no longer concerned with the growth process.
However, the brittle cleavage component is affected in a marked manner by .the
environment. Observations on an aluminium-zinc-magnesium alloy fatigued in
3o/o NaC1 solution, sh<;>Wed that the cleavage component predominated and dictated
the path of fracture, along the (100) planes, with occasional excursions back to ·
the normal fracture plane. Fatigue at low frequencies produced a similar effect,
the fracture path often followed (1 00) planes along which the growth for each
cycle of stress increased considerably. The association of. cleavage with high
tensile stresses and with triaxial stress conditions is not surprising. However,
. .the fact that for similar stressing conditions it occurs only in the presence of a
corrodant, or in air with slowly applied stresses, suggests the action of some
time dependent phenomenon that encourages cleavage. There is no indication
from electron micrographs of any electro-chemical removal of material from the
crack tip, and it seems probable that cleavage is being encouraged by the form-
ation of hydrogen ions, and the diffusion of these or others already existing in
the material, to the crack tip. There is a strong analogy here with a hydrogen
charged steel which behaved in the same way when fatigued in air at a frequency
of 1800 cycles per minute. (12)
It is of interest to apply the ideas presented in this paper to the important
practical effect of occasional overloads. It is generally agreed that occasional
high positive overloads are beneficial to the fatigue life of a cracked component
because they induce residual compressive stresses around the crack tip. Oc-
casional high negative loads are fortunately not particularly detrimental because
 - 86 -

the crack tip is a less effective stress raiser in compressioiL It would appear
that any residual stress would affect the cleavage component of crack growth,
and a high residual compressive stress might inhibit it altogether. If this hap-
pened growth would continue by the stage I process. Eventually the residual
stress field Would be relaxed and stage 11 growth would continue again. In this
respect we cannot wholly ignore the Bauschinger effect which may well encourage
stage I growth. Stage I growth which involves the movement of dislocations will
always tend to relax residual stresses of either sign, but because the induced
stresses are usually beneficial, the stage I process is acting in a detrimental
manner in preparing the crack tip for continuance of stage 11 growth.

Conclusions
A study of various aluminium alloy fracture surfaces shows that a general
two stage classification can be made.
Stage I cracking is primartly the result of dislocation movement and fracture
occurs along the slip plane.
Stage 11 cracking contains a component of cleavage.
Stage I cracking will proceed under conditions favouring single slip, whereas
Stage 11 cracking occurs under conditions where slip becomes more complex and
appreciable triaxial stresses exist.
The component of cleavage in stage 11 cracking varies in importance. The
basic increment of fracture, the striation, will show a greater or less degree of
brittle behaviour depending on the mechanical properties of the material, the en-
vironment, and the stressing conditions.
Cleavage is encouraged by a corrodant or by slow application of siress. It
has also been reported that it is encouraged in a mild steel by electrolytically
charging with hydrogen before fatiguing.
The phenomenon of non propagating cracks can be explained with refer.ence to
the two stages of crack growth. The cracks show features that sUggest that they
start almost immediately as stage 11 cracks, then grow to a length at which the
crack tip is subject to too small a tensile stress to continue to grow in this mode.
The cracking mode may change to stage I, but the rate of growth will be very
small because of the constrained conditions at the crack tip.

Acknowledgements
The author gratefully acknowledges useful discussions with Mr. D. A. Ryder
and Mr. C. A. Stubbington who have also worked on this investigation.
Published by permission of H. M. Stationery Office - Crown Copyright Reserved.
 - 87 -

References

1. Shanley, R. F, Proc. lnter. Union Theoretical and Applied

Mech. Colloquim on Fatigue,
Stockholm (1955) Springer Berlin, 251 (1956)
2. Forsyth, P. J. E. Fatigue in Aircraft Structures,
Columbia University Conference -
· 19 56 - Academic Press lnc. N, Y.
3. Wood, W.A. ---ibia__ _

4. Broom, T, and Nature, 177, 486-487

Whittaker, V. N. March, 1956,
5. Cottrell, A." H. and Proc. Roy. Soc. (London) A, 242,
Hull, D. (1957).
11
6. Forsyth, P. J. E., and Metallurgia 11 Vol. 63, No. 377, p. 117,
Ryder, D. A. March, 19 61.

7. Forsyth, P. J. E. and "Aircraft Engineering"

Ryder, D, A. April 1960.

8. Schijve, J. 2nd Congress of the International Council of

Aeronautical Sciences,
September 1960 (Zurich).

9. Frost, N. E. Proc, lnst. Mech. Eng. London, Vol. 173,

No. 35.

10. Frost, N.E. J. Mech. Eng. Science,

June, 1960.

11. Orowan, E. Fatigue in Aircraft Structures,

Columbia University Conference 1956 (Discus-
sion on C. E. Phillips paper, p. 121).
Academic Press, Inc. N. Y.

12. Ryder, D. A. Private Communication.

 - 88 -

FIG. 1. EXTRUSION FROM SLIP BAND IN AN ALUMINIUM - 10~ ZINC ALLOY .

(x 2,500) [Arrow A indicates markings on extrusion surface revealing slip direction)

FIG . 2. TORSIONAL FRACTURE SURFACE OF AN ALUMINIUM 7 .5'1o ZINC

2 .5.,. MAGNESIUM ALLOY . (x 240)

Inif"iat,_, o1 Sbtt-I!
wiff! /tattllre ~trfitl•ns
CDrt~/>lrte chuge over
to Su~ II er«k.

FIG. 3 . SCHEMATIC DIAGRAM SHOWING SLIP PLANE CRACK, STAGE I, THE

ONSET OF STAGE 11 CRACKING, AND THE NEW FRACTURE PLANE
 - 89 -

FIG. 4 . A FATIGUE CRACK SUBSEQUENTLY SECTIONED, POLISHED AND ETCHED

The material wae cold rolled pure aluminium A zone of eubgraln growth surround& the
crack and extends ahead of the crack Up . (x 1500) [sup bands indicated by arrow& A and
B were produced by a subsequent stress cycle]

FlO 5 . THE SAME SPECIMEN AS ILLUSTRATED IN FIG 4 . Heavy localized sllp

baa been produced in recrystallized zones by subsequent tensile deformation . (x 1500) .
 - 90 -

FlO. 8 . SHADOWGRAPH OF FRACTURED FIG. 7 . FATIGUE CRACKS IN AN

ROLLS ROYCE CORROSION FATIGUE ALUMINIUM 7 .5'1o ZINC 2 .5'1o
SPECIMEN (x 12) MAGNESIUM ALLOY . (x 750)

FlO. 8 . SPECIMEN AS FOR FIG. 4, SHOWING CHANGEOVER OF MODE OF

CRACKING, ARROW A. (x 200) .

FlO. 9 . A SCHEMATIC ILLUSTRATION OF STAGE II I-'RACTU)tE SURFACE .

 - 91 -

'•

FlO. 10. FATIGUE FRACTURE SURFACE OF PROGRAMME LOADED SPECIMEN.

(x 2,000) 1, 2 , 3 and 4 •how the repeated groupe of 7 cycles of a particular load level.
A, B and C are area• of low load level growth , where the fracture mode has changed
to Stage 1 with a tilt of the fracture surface .

1
dit'tctto-t
of ~rowth

WY#
contour ~
Brittle fr.advr"e s~riations
Ductile Jr.Kba,.. striatioru

FIO. ll . A SCHEMATIC DIAGRAM ILLUSTRATING THE DIFFERENCES BETWEEN

DUCTILE AND BRITTLE STRIATIONS.

Cll
v
4 ......
()
>.
u
3 Cll
c
..
..
0

...,..
2 ..
1'6

Cll

FIO. U. DUCTILE STRIATIONS. (x 1500) FJG. 13 . BRITTLE CLEAVAGE

STRIATIONS. A lndicatea crack
front, B , river markln&• . (x 1500)
 - 92 -

FlO. 14 . ALUMINIUM 7 .5,. ZINC 2 .5,. MAGNESIUM ALLOY FATIGUED AT

1800 CYCLES PER MINUTE AND 20 CYCLES PER MINUTE (x 600)

FIG . IS . LOCAL CHANGES IN MODE OF CRACKING UNDER CORROSION

FATIGUE~ (x 600)

ductile Sfr;.t-;'o, s
direct/on of cr~k
growth

-
FIG. 16 . A SCHEMATIC ILLUSTRATION OF FIG.15 .
 - 93 -

FlO. 17 . BRITTLE CLEAVAOE STRIATIONS ON A CORROSION FATIGUE

FRACTURE. (x 2,000)

FIG . 18 FRACTURE GROWTH PATTERN ON DTD 607 SHEET SPEClMt.N. ,,. l1

e,.,, ..T AUA

s~.... r--a-,....

FIG. 19 . SCHEMATIC ILLUSTRATION OF THE

FEATURES OF THE GROWTH PATTERN S HOWN
IN FIG. 18 x represents fatigue growth at aur!nce
and y represents fatigue growth In centre of sheet.
The changes In crack front angle are repreoented
by 0, and 01 •

FIG . 20. THE REPOLISHED AND ETCHED FRACTURE

SURFACE OF A PURE ALUMINIUM FATIGUE SPECIMEN
STAOE 11 CRACKING The eubotructure reveals the traces
of the strlahona . (x 1500)
 - 94 -

FIG . 21 . A TAPEH SECTION THHOUGII A STAGE II

FATIGUE CRACK IN AN ALUMINIUM-ZINC- MAGNESIUM
ALLOY . The slip markings have been revealed by sub-
sequent ageing after fatigue . (x 200)

FIG 22 . A FATIGUE FRACTURE SURFACF: IN DTD . 683

ALLOY SHOWING PJTS (A) WHERE PAJlTIC LFS IIA VE
CRACKED AND GROWN INTO FATIGUE CRACKS AHEAD
OF THE MAIN CRACK . (Programme loaded IIIJI!clmen)
(x 2000)

A A

FIG. 23 . FINAL I<"RACTURE ZONE o ..~ A FATIGUED

ALUMINIUM 4'1o COPPER ALLOY SHOWING PITS (AI
AND GROWTH STRIATIONS . (x 1000)