Prognostics Modeling of Solder Joints

in Electronic Components 12

Jeffrey W. Simons and Donald A. Shockey

SRI International
333 Ravenswood Ave
Menlo Park CA 94025
650-859-4495 650-859-2587
[email protected] [email protected]

Abstract—We describe a methodology for performing The component considered here, Q10, shown in Figure 1, is
materials-based prognostics for failure in electronic an 8 gull-wing leaded power supply chip on CD
components. The methodology is based on results of a Technologies’ WPA50 DC/DC voltage converter PWB
previous research program involving a substantial number assembly. The WPA50 Series is a 50 Watt single output,
of mechanical fatigue experiments and FE analyses low-profile DC/DC converter in an industry standard
investigating solder joint integrity. The component package of 2.3” x 0.9” x 0.35” (58.42mm x 22.86mm x
considered here is a gull-wing lead power supply chip on a 9.0mm). The WPA50 has an input voltage range of 36-75
DC/DC voltage converter PWB assembly. As a first step, VDC and output voltages between 1.2VDC and 5.0VDC.
three-dimensional finite element analyses are performed to
determine macrostrains in the solder joint due to thermal or
mechanical cycling of the component. We consider strains
due to lead bending resulting from the thermal mismatch of
the board and chip and those resulting from local thermal
mismatch between the lead and the solder as well as
between the board and the solder. In the second step, the
macrostrains are used to set boundary conditions for a
probabilistic micromodel that can explicitly simulate
initiation and growth of cracks in the microstructure of the
solder joint. Based on the growth rate of the cracks in the
solder joint, estimates can be made of the cycles to failure a) Side A
for the electronic component. Here, we describe the first
step of the methodology.

TABLE OF CONTENTS
1. INTRODUCTION ......................................................1
2. APPROACH .............................................................2
3. ANALYSIS METHOD ...............................................3
4. RESULTS ................................................................3
5.CONCLUSIONS ........................................................5
REFERENCES .............................................................5 b) Side B
BIOGRAPHY ...............................................................6 Figure 1. WPA50 Images
The objectives of these analyses were to (1) establish a
1. INTRODUCTION mechanical test cycle that would produce strains and strains
in the solder joint equivalent to those experienced by the
We performed finite element analyses to determine component during thermal cycling in operation, and (2)
macrostrains in the solder joints of an electronic component reach a level of strain that would lead to crack growth under
under thermal and mechanical loading. a reasonable number of cycles.

1
1
0-7803-9546-8/06/$20.00© 2006 IEEE
2
Revision 3 - final

1
In the thermal analyses, we considered strains due to lead
bending resulting from the thermal mismatch between the
board and chip and those resulting from local thermal
mismatch between the lead and the solder as well as
between the board and the solder.

These simulations are part of an effort to develop and

validate a prognostics algorithm for electronics components.
Calculated strains in the elements (here considered as
macrostrains) will be used to set boundary conditions for a
probabilistic model that can explicitly simulate initiation
and growth of cracks in the microstructure of the solder
joint. Based on the calculated growth rate of the cracks in \
the solder joint, estimates can be made of the cycles to Figure 3. Dimensions for chip and board section
failure for the electronic component.

To validate the prognostics approach mechanical fatigue

tests will be performed on the component and the location
and rate of growth of cracks in the solder will be measured
and compared with simulations.

2. APPROACH

Geometry
Base on high-resolution photographs provided by Vextec,
Inc such as the one shown in Figure 2, we developed a finite
element model of the Q10 component. a) elevation

Figure 2. High resolution micrograph of component.

The geometry of the finite element model configuration
along with some critical dimensions are shown in Figure 3
for the board and chip and in Figure 4 for the lead, solder
and pad. For our finite element analysis, we used three b) plan
planes of symmetry, two vertical planes bisecting the length
and width of the chip and, because of the components on Figure 4. Dimensions for lead, solder and pad
both side A and side B of the board, as shown in Figure 1, A representative portion of the finite element mesh is shown
one horizontal plane bisecting the centerline of the board. in Figure 5. The element size varied from a minimum
Thus, the computational model included the chip and a dimension of about 0.01 mm for high resolution areas in the
section of the board as well as two leads, solder joints and solder to a maximum dimension of about 0.3 mm in parts of
pads. the chip and board away from the leads and solder. The
model had a total of about 44,000 nodes and 39,000
elements. As seen in Figure 5, all the elements used were 8-
node brick elements with single-point integration.

2
 was assumed to be uniform and was raised 100C.
3) For the mechanical analysis, as shown in Figure 6, the
board was split along the plane parallel to the chip edges
containing the leads. The two outer edges of the board were
then displaced 0.005 mm away from the chip (a total
relative displacement of 0.10 mm) while being prevented
from rotating.

Figure 6. Mechanical cycling method.

3. ANALYSIS METHOD

The analyses were performed using LSDYNA3D [4], a

Figure 5. Representative mesh for component commercial code for three-dimensional finite element
Materials analysis of structural response. We used explicit time
integration, i.e., performed a dynamic analysis but at a slow
The materials were all modeled as thermal elastic, that is,
enough rate to keep dynamic effects less than about 5% of
linear elastic mechanical properties with a coefficient of
the response, as seen in Figure 7 for a solder element
thermal expansion (CTE). Material constants were taken by
located along the lower edge where the lead exits the solder
averaging values from several sources [1-3] and are listed in
joint.
Table 1. We recognize that to accurately model the material
response would require including more than elastic
120 60
response, particularly for the solder, but for our primary Temperature
objective here, establishing equivalence between thermal Stress

EFFECTIVE STRESS (MPa)

100 50
and mechanical cycling, elastic analysis is adequate. Note
TEMPERATURE (C)

80 40
the large difference in CTE between the chip and board; that
is the primary source of stress and strain for thermal 60 30
loading; the lead must accommodate the relative changes in
dimensions of the chip and board under thermal loading. A 40 20

secondary source of stress is the “local mismatch” in CTE 20 10

between the solder and lead and the solder and pad.
0 0
Table 1. Material properties for simulations 0.000 0.002 0.004 0.006 0.008 0.010
Part Material Young’s Poisson’s CTE TIME (s)
Modulus ratio (ppm/C)
(MPa) Figure 7. History of applied temperature and resulting
Chip Silicon 190 0.28 3 element stress.
Lead Copper 120 0.35 17
Solder Solder 19 0.28 25 4. RESULTS
joint
Pad Copper 120 0.35 17
Thermal Analysis
Board 15 0.4 14.5
Calculated contours of effective strain in the solder joints
Boundary conditions for 100°C thermal loading are shown in Figure 8. In the
outer joint the peak strain (4.0x10-3) is at the upper outside
The assumed boundary conditions for the analyses were as corner where the lead exits the solder. There are also
follows: secondary concentrations along the outer lower edge of the
1) Three planes of symmetry for both the thermal and foot of the lead. Because the thermal expansion is in all
mechanical analyses directions, the strains in the outer joint are higher than in the
inner joint and also the outer edges of each joint have higher
2) For the thermal analysis, the temperature of all the parts strains than the inner edges. For the inner joint, peak strains
3
between (2.0x10-3 and 2.4x10-3) are found at the upper simulations are linear elastic, the results can be scaled to
corner and lower edge of where the lead exits the solder and define a mechanical testing cycling that produces equivalent
the bottom edge outer corner of the toe of the lead. peak effective strains to a thermal cycle, for example, the
0.010 mm mechanical cycling would be equivalent to a
138°C thermal cycling in terms of when primary cracking
would first be expected. In simple terms, 0.001 mm of
displacement is equivalent to about 13.8C.

Estimate of cycles to cracking

To estimate the number of cycles to initiate cracking in the
mechanical test we looked at the calculated effective strain
in the solder for the 0.010 mm mechanical cycle and
compared it with data from previous tests and analyses.

Figure 10 shows results from Solomon [7] for isothermal

shear strain range controlled fatigue of 60Sn/40Pb solder.
Solomon developed fatigue failure criteria for mechanical
Figure 8. Effective strain contours in solder for 100°C
tests based on reduction in the load range by a specified
thermal loading
percentage of the initial load range. For our mechanical test
Mechanical Analysis
design we selected Solomon’s correlation for a load range
The calculated effective strain contours for a 0.010 mm reduction of 50%. Thus for a number of cycles between
mechanical displacement are shown in Figure 9. Unlike the 1000-3000, we should design the test to produce a plastic
thermal loading, the strain distributions in the mechanical strain per cycle between 0.02 and 0.04.
loading look the same for both the outer and inner joint.
For both joints the peak strain (5.5x10-3) for the outer joint)
1.00
is at the upper outside corner of where the lead exits the
solder. The strains along both the upper and lower edges of 90% Load drop
where the lead exits the solder have comparable strain
PLASTIC STRAIN PER CYCLE

50%
(above about 5.0x10-3). For the mechanical loading, the 25%

strains are more concentrated where the lead exits because 10%

there are no effective strains resulting from the local

mismatch of thermal expansion between the solder and lead 0.10

or between the solder and pad.

design
space

0.01
10 100 1000 10000
CYCLES TO LOAD REDUCTION

Figure 10. Solomon’s data for mechanical fatigue of solder

Because the plastic strain accumulates equally during
unloading as well as loading, we can use half these values
(i.e., 0.01 to 0.02) as our target design values for effective
plastic strain developed for a single half-cycle of
mechanical loading. As shown in Figure 9.

Because our analyses were elastic, the calculated strains are

Figure 9. Effective strain contours in solder for 0.010 mm elastic strains. During a test, the solder will relax and the
mechanical loading elastic strains will transform to creep strains. Thus, the
Comparing the strain distributions between the thermal and amount of plastic strain accumulated in the solder during
mechanical loading, we see that the peak effective strains cycling depends not only on the displacement, but also on
are about 38% higher for the 0.010 mechanical cycling than the hold time of the cycle.
for the 100°C thermal cycling, but because both these

4
To estimate the creep strain accumulation, we assumed that,
0.020
because the tests are displacement controlled, the total strain
remains constant and we calculated creep strain 0.018
accumulation with hold time. To model the solder creep, 0.016
we used the formulation given by Schmidt [3],
0.014

PLASTIC STRAIN
0.012
ε t = ε e + ε c (1)
0.010
0.008 0.035 mm
Where ε t is the total strain rate, ε e is the elastic strain rate
0.006 0.030 mm
and ε c is the creep strain rate The elastic strain rate is given
by, 0.004
0.002
σ 0.000
ε e = 0 20 40 60 80 100
3G
TIME (s)
And the creep strain rate by,
n
⎡ ⎛σ ⎞⎤
ε c = A⎢sinh ⎜⎜ ⎟⎥ (2)
⎟ Figure 11. Accumulated plastic strain for increased
⎣⎢ ⎝σ0 ⎠⎦⎥ mechanical displacements.
⎛ T ⎞ Thus, for the displacement-controlled mechanical tests to be
A = A0 exp⎜⎜ − 0 ⎟⎟
⎝ T ⎠ used to validate the prognostics model, we recommend a
Where σ is the effective stress, T is the temperature in displacement cycle 0.035 mm. This level of displacement
should produce cracking in the solder after a few thousand
degrees K, and A0 , T0 , G , σ 0 , and n are materials cycles that significantly reduces the load.
parameters that are sensitive to grain size and temperature.
We used the parameters shown in Table 2 derived by 5.CONCLUSIONS
Schmidt from analyses of thermal and mechanical fatigue of
solder joints.
Based on our analyses we made the following conclusions:
Table 2. Parameter Values for Solder Creep Model (1) By performing elastic analysis of thermal and
A0 1.3 x106/s mechanical cycling of the Q10 component, we
established an equivalence of 13.8C/0.001 mm for
T0 9120° K
peak effective strains.
G 7.5 GPa
σ0 9.8 MPa (2) Recommended design parameters for mechanical
n 2.4 fatigue tests are a peak displacement of 0.035 in and a
hold time of at least 20 s. This should produce
Creep strain can be calculated by integrating eq. (2) using significant cracking in the solder after a few thousand
eq. (1) to update the stress. Assuming that the mechanical cycles. Note that these strains are considerably higher
cycling would take place at 25C, and using a rise time of 5 s than expected for thermal cycling in service.
for the cycling, we calculated plastic strain accumulation as
(3) The calculated strains in the solder can be used as
a function of time for two increased levels of mechanical
boundary conditions for a probabilistic model that can
displacement: 0.030 mm and 0.035 mm as shown in Figure
explicitly simulate initiation and growth of cracks in
11. The plastic strain increases rapidly during the 5 s rise
the microstructure of the solder joint.
time and by 20 s begins to level off. The levels reached after
20 s, 0.015 and 0.018 correspond to plastic strain values of
0.030 and 0.036 in Figure 10, and are within the design
space.
REFERENCES

1. M. Amagai, ”Characterization of chip scale packaging

materials,” Microelectronics Reliability 39 (1999) 1365-
1377
2. Intel Performance Characteristics of IC Packages, 1998
Packaging Databook
3. C.G. Schmidt, J.W. Simons, C.H. Kanazawa, and D.
Erlich, "Thermal Fatigue Behavior of J-lead Solder Joints,"
5
IEEE Transactions on Components, Hybrids, and
Manufacturing Technology Part A, Vol. 18, No. 3
(September 1995).
4. LSTC website: www.lstc.com
5. H. D. Solomon,”Fatigue of 60/40 Solder,” IEEE Trans.,
CHMT, 9, 423-433 (1986).

BIOGRAPHY

Jeffrey Simons is a senior research engineer in the Center

for Fracture Physics, Materials Research Laboratory, at
SRI International. Dr. Simons has
extensive experience in developing
and using finite element analysis
codes to investigate the response of
materials and structures with a
concentration on model damage and
failure. He has a BA in applied
mechanics from UCSD and a MS and
PhD in Civil Engineering from UCB.

Donald Shockey, Director of SRI's Center for Fracture

Physics, is an internationally recognized expert in the
fracture behavior of materials. He earned a Ph.D. in
metallurgy and materials
science at Carnegie-Mellon
University. An authority on
material failure under impact
and explosive loads, Dr.
Shockey has published over
120 journal articles, is an
Institute Fellow, a Fellow of
ASM International, and the
Year 2000 recipient of the
John S. Rinehart award for pioneering work in the field of
dynamic fracture.