G. Z.

Wang
e-mail: guozhong.wang@daimlerchyrsler.com Applying Anand Model to
Z. N. Cheng
e-mail: zhaonian.cheng@daimlerChrysler.com
Represent the Viscoplastic
Daimler Chrysler SIM Technology,
865 Changning Road,
Deformation Behavior of Solder
Shanghai 200050, China
Alloys
K. Becker A unified viscoplastic constitutive law, the Anand model, was applied to represent the
University of Applied Science Bingen, inelastic deformation behavior for solders used in electronic packaging. The material
Bingen 55411, Germany parameters of the constitutive relations for 62Sn36Pb2Ag, 60Sn40Pb, 96.5Sn3.5Ag, and
e-mail: becker@fh-bingen.de 97.5Pb2.5Sn solders were determined from separated constitutive relations and experi-
mental results. The achieved unified Anand model for solders were tested for constant
strain rate testing, steady-state plastic flow and stress/strain responses under cyclic load-
J. Wilde ing. It is concluded that the Anand model can be applied for representing the inelastic
IMTEK, University of Freiburg, deformation behavior of solders at high homologous temperature and can be recom-
Freiburg im Breisgan D-79085, Germany mended for finite element simulation of the stress/strain responses of solder joints in
e-mail: wilde@imtek.uni-freiburg.de service. 关DOI: 10.1115/1.1371781兴

Introduction and Kitano 关9兴兲 proposed a unified model for 60Sn40Pb solder
which accounts for the measured stress-dependence of the activa-
In general, temperature fluctuations experienced by IC pack-
tion energy and for the Bauschinger effect exhibited by the solder.
ages and assemblies in service cause progressive damage in solder Qian et al. 关10兴 employed the back stress to describe the transient
joints; eventually, this damage accumulation beyond certain limits stage of a stress/strain curve in a unified constitutive model for
leads to the electrical failure. One of the major goals of thermo- tin-lead solder. Some works 共Skipor et al. 关11兴; Ma et al. 关12兴兲
mechanical analysis in the electronics industry is to be able to employed the Bodner-Partom constitutive relations which use a
simulate the stress/strain responses of the solder joint and then state variable to represent the internal inelastic structure for the
predict its reliability in service. In order to gain accurate simula- solder with eutectic composition. However, some parameters of
tion and reliable prediction, realistic constitutive relations for sol- these unified models are empirical and dependent on temperature
der alloys are warranted. and strain rate, resulting in complex calculation of the stress strain
The high homologous temperatures, e.g., 0.65 T m 共in K兲 for the responses and some scattering predictions from the experiments.
eutectic tin-lead共SnPb兲 at room temperature, experienced by the Usually, a specially viscoplastic constitutive law must be de-
solder joint and the thermally activated strains imposed on it due fined as a user-defined subroutine code to represent the nonlinear
to the thermal expansion mismatch between the materials gives rate-dependent stress-strain relations in some finite element pro-
rise to a complex deformation behavior. This deformation behav- grams 共Busso et al. 关8兴, Qian et al. 关10兴兲. Such a work is often
ior is associated with the irreversible, temperature and rate 共or complex, expert dependent and largely time consumptive. Some
time兲 dependent inelastic characteristics, producing strain- user material subroutines with unified constitutive models are al-
hardening, dynamic recovery, and in many instances dynamic re- ready available in current commercial finite element codes, e.g.,
crystallization. This deformation behavior is known to be visco- the UMAT in ABAQUS 共Weber et al. 关13兴兲. A unified constitu-
plastic with, generally, revelations such as creep and stress tive model, which is referred to as the Anand model, is offered by
relaxation phenomena of materials working in high homologous the ANSYS code. In order to apply this Anand model to simulat-
temperature regime. ing the thermomechanical responses of solder joints in electronic
There has already been a great deal of effort applied to reason- packaging, the material parameters of the constitutive relations
able experimental data and constitutive models for this material. must be determined in prior.
The previous researchers 共Darveaux and Banerji 关1兴; Weinbel The objective of this paper is to obtain the material parameters
et al. 关2兴, Kashyap and Murty 关3兴兲 presented extensive experimen- of the Anand model for solders from experimental results and the
tal data on tin-lead based solders. There are also some constitutive separated elasto-plasto-creep constitutive relations. The material
relations for solder alloys, ranging from elasto-plastic model 共e.g., parameters with viscoplastic constitutive relations for solders are
Ramberg-Osgood relation兲 using empirical stress-strain curves used to simulate the steady-state creep behavior, constant strain
共Lau and Rice 关4兴兲 to a purely phenomenological model where the rate behavior, and stress/strain responses under thermal cyclic
time-dependent and time-independent deforms are artificially loading for comparison and verification. Some discussions on rec-
separated 共Sarihan 关5兴, Pao et al. 关6兴; Knecht and Fox 关7兴兲. Some ommendation of using unified Anand model in the finite element
creep models from literature 共power law creep, Harper Dorn simulation of electronic packaging reliability were also presented.
creep, hyperbolic sine creep, etc.兲 have been applied to the time
dependent creep data. However, from the viewpoint of continuum Model Formulation
mechanics, the time-dependent and time-independent inelastic
strains are assumed to arise from similar mechanisms due to dis- 1 The Anand Model. A simple set of constitutive equations
location motion. So, a unified framework for viscoplastic behavior for large, isotropic, viscoplastic deformations but small elastic de-
of solder materials is highly desired. However, there exist insuf- formations is the single-scalar internal variable model proposed
ficient works until now. Some researches 共Busso et al. 关8兴, Busso by Anand and Brown 共Anand 关14兴, Brown et al. 关15兴兲. There are
two basic features in this Anand model. First, this model needs no
Contributed by the Electronic and Photonic Packaging Division for publication in
explicit yield condition and no loading/unloading criterion. The
the JOURNAL OF ELECTRONIC PACKAGING. Manuscript received by the EPPD Oc- plastic strain is assumed to take place at all nonzero stress values,
tober 20, 1998. Associate Editor: Ye-Hein Pao. although at low stresses the rate of plastic flow may be immea-

Journal of Electronic Packaging Copyright © 2001 by ASME SEPTEMBER 2001, Vol. 123 Õ 247
surably small. Second, this model employs a single scalar as an where the function g( ␴ ,s,T) is associated with dynamic process,
internal variable to represent the isotropic resistance to plastic that is, strain hardening and dynamic recovery. A simple form of
flow offered by the internal state of the material. This internal evolution equation of Eq. 共4兲 was given by Anand 关14兴 as follows:

再冏 冏 冉 冊冎
variable is denoted by s, which has the dimensions of stress, and is
called to be deformation resistance. There are some reasonable s a s
considerations for the simplifications that only a single scalar is ṡ⫽ h 0 1⫺ •sign 1⫺ •␧˙ P ; a⬎1 (5)
s* s*
used to characterize the internal structural characteristics of a ma-
terial 共Brown et al. 关15兴兲. The set of Anand constitutive equations with

冋 冉 冊册
presented here accounts for the physical phenomena of strain-rate
and temperature sensitivity, strain rate history effects, strain- ␧˙ P Q n
s * ⫽ŝ exp (6)
hardening and the restoration process of dynamic recovery. A RT
The internal variable s represents an averaged isotropic resis-
tance to macroscopic plastic flow offered by the underlying iso- where h 0 is the hardening/softening constant, a is the strain rate
tropic strengthening mechanisms such as dislocation density, solid sensitivity of hardening/softening. The quantity s * represents a
solution strengthening, subgrain, and grain size effects, etc. The saturation value of s associated with a set of given temperature
deformation resistance s is consequently proportional to the and strain rate as shown in Eq. 共6兲. ŝ is a coefficient, and n is the
equivalent stress ␴. That is strain rate sensitivity for the saturation value of deformation re-
sistance, respectively.
␴ ⫽c•s; c⬍1 (1) From the above viscoplastic Anand model, there are nine ma-
terial parameters: A,Q, ␰ ,m,h 0 ,ŝ,n,a, and s 0 , the last one is the
where c is a material parameter and constant in the constant strain initial value of the deformation resistance, which is needed to
rate test, which is defined as determine the evolution of deformation resistance in Eq. 共5兲.

c⫽
1
␰
sinh⫺1 冋冉 ␧˙ P Q/RT
A
e 冊册
m
(2)
2 The Separated Constitutive Model. In the separated
constitutive descriptions for solder alloys, the total inelastic strain
␥ is the sum of time-independent plastic strain ␥ P , the steady-
where ␧˙ P is the inelastic strain rate, A is the pre-exponential fac- state creep ␥ S , and the transient 共or primary兲 creep strain ␥ T .
tor, Q is the activation energy, m is the strain rate sensitivity, ␰ is That is
the multiplier of stress, R is the gas constant, and T is the absolute
temperature, respectively. ␥ ⫽ ␥ P⫹ ␥ S⫹ ␥ T (7)
The following functional form of the flow equation was se- with
lected to exactly accommodate the strain rate dependence on the
stress at constant structure:
␥ P ⫽C 2 冉冊␶ mp

冉 冊冋 冉 冊册
(8)
Q ␴ 1/m G
␧˙ P ⫽A exp ⫺ sinh ␰ (3)
RT s ␥ S ⫽ ␥˙ S •t (9)
Note that the internal state variable enters into the flow equation ␥ T ⫽ ␥ t 关 1⫺exp共 ⫺B ␥˙ S t 兲兴 (10)
only as a ratio with the equivalent stress. The temperature depen-
dence in Eq. 共3兲 is incorporated via a classical Arrhenius term. where ␶ is the applied shear stress, G is the shear modulus, C 2 is
While the stress and state dependence is a simple modification of the plastic strain coefficient, m p is the stress sensitivity of plastic
the hyperbolic sine form first proposed by Garofalo 关16兴 to model strain, ␥˙ S is the steady-state strain rate. B is the transient creep
steady-state plastic flow 共secondary creep兲. coefficient, ␥ t is the transient creep strain, and t is the time,
The evolution equation for the internal variable s is assumed to respectively.
be of the form as The steady-state creep rate can be described by a single expres-
sion, a hyperbolic sine form from Darveaux and Baneji 关1兴. That
ṡ⫽g 共 ␴ ,s,T 兲 ␧˙ P (4) is

Table 1 Material parameters of separated model for solders „Darveaux and Banerji †1‡…

Steady State Transient

Elastic Creep共2兲 Creep Plastic共2兲
G 0 共at 0°SC兲 G1 C1 Q
Solder (104 MPa) (102 MPA) 共K/s/MPa兲 ␣ nS 共ev兲 B ␥t C2 mp
60Sn40Pb
Shear 1.310 0.559 28.720 1300 3.3 0.548 440 0.020 2.3(1013) 5.6
Tensile共1兲 3.447 1.517 16.534 751 3.3 0.548 762 0.012 6.1(1011) 5.6
62Sn36Pb2Ag
Shear 1.310 0.559 14.344 1300 3.3 0.548 137 0.037 6.6(107 ) 4.5
Tensile共1兲 3.447 1.517 8.282 751 3.3 0.548 237 0.021 5.6(106 ) 3.5
96.5Sn3.5Ag
Shear 1.931 0.689 7.09(10⫺2 ) 1300 5.5 0.40 147 0.086 2.0(1011) 4.4
Tensile共1兲 5.240 1.931 4.09(10⫺2 ) 751 5.5 0.40 255 0.050 1.0(1010) 4.4
97.5Pb2.5Sn
Shear 0.896 0.103 7.31(109 ) 1200 7.0 1.20 137 0.115 2.8(1010) 4.0
Tensile共1兲 2.413 0.283 4.22(10 9 ) 693 7.0 1.20 237 0.066 1.8(109 ) 4.0

共1兲
Inelastic tensile constants derived from shear constants using relations ␴ ⫽ ␶ 冑3 and ␧⫽ ␥ / 冑3 due to the assuming of von Mises yield criteria.
共2兲
Use shear modulus when calculating tensile strain rate in Eq. 共11兲 and plastic strain in Eq. 共8兲.

248 Õ Vol. 123, SEPTEMBER 2001 Transactions of the ASME

 ␥˙ S ⫽C 1
G
T 冋 冉 冊册 冉 冊
sinh ␣
␶
G
ns
exp ⫺
Q
RT
(11)
共c兲 Determination of ␰ and ŝ. By using the values obtained in
step 共b兲, the parameter ␰ was slected such that the constant c in
Eq. 共2兲 is less than unity, and ŝ was then determined from the
where C 1 is a coefficient, ␣ is the multiplier of stress, and n s is the combined term ŝ/ ␰ .
stress sensitivity of steady-state creep rate, respectively. A hyper- 共d兲 Nonlinear fitting of a,h 0 and s 0 . The stress of Anand
bolic sine dependence of the strain rate on the stress results from model can be expressed as
an assumption of forward and backward thermal activation of va-
cancies. However, some investigators 共Arrowood et al. 关17兴兲 as- ␴ ⫽ ␴ * ⫺ 关共 ␴ * ⫺cs 0 兲 共 1⫺a 兲 ⫹ 共 a⫺1 兲 兵 共 ch 0 兲共 ␴ * 兲 ⫺a 其 ␧ p 兴 1/共 1⫺a 兲
sociated power law breakdown with a transition from dislocation (15)
climb dominated steady-state plastic flow to dislocation glide The material parameters a,h 0 , and s were fitted from the ␴ ⬃␧ P
dominated one. curves 共especially the transient state兲 with various temperatures
To obtain the total inelastic strain, the temperature dependence and strain rates. In the fitting, the saturation stresses given in step
of the shear modulus G must be incorporated 共a兲 were also used.
G⫽G 0 ⫺G 1 •T 共 °C兲 (12) The nine material parameters for 62Sn36Pb2Ag, 60Sn40Pb,
96.5Sn3.5Ag, and 97.5Pb2.5Sn solder materials, determined from
where G 0 is the shear modulus 0°C, and G 1 gives the temperature the above procedure, are listed in Table 2. For a whole deforma-
dependence. Table 1 lists the material parameters of the separated tion response for solder due to a load, the elastic behavior should
elasto-plasto-creep model for 60Sn40Pb, 62Sn36Pb, also be included. The elastic properties, e.g., the temperature de-
96.5Sn3.5Ag, and 97.5Pb2.5Sn solders. pendent elastic moduli, for these solders are the same as listed in
Table 1. The Poisson’s ratios can also be derived from the rela-
tionship between shear modulus and tensile modulus in Table 1.
Parameters Determination of Anand Model
1 Steady-State Plastic Flow and Saturation Stress. When Verification and Application
steady-state plastic flow occurs, the stress reaches the saturation 1 Model Prediction. The achieved material parameters of
value. For viscoplastic deformation behavior, this steady-state unified Anand model for solders were tested for constant strain
plastic flow happens when the plastic flow has become fully de-
veloped and its rate equals the applied strain rate at the given
temperature and strain rate. The steady-state plastic flow of Anand
model can be derived as

␧˙ P ⫽␧˙ ⫽A exp ⫺ 冉 冊冋 Q
RT
sinh␰
␴*
s* 冊 1/m
(13)

where ␧˙ is the applied strain rate in constant strain rate test. From
Eq. 共13兲, the saturation stress ␴ * is obtained as follows:

␴ * ⫽cs * ⫽ 冉
ŝ ␧˙ p Q/Rt
␰ A
e 冊 n
sinh⫺1 冋冉 ␧˙ p Q/RT
A
e 冊册
m
(14)

From the above, the material parameters: Q/R,A,ŝ/ ␰ ,m and n in

Eq. 共14兲 can be determined by applying a nonlinear fitting method
using ␧˙ p ⬃ ␴ * pair data from steady-state creep results.
2 Determination Procedures and Results. The material
parameters of the elastic, plastic, and creep deformation behavior
shown in Table 1 can be transformed to the viscoplastic Anand
model. The values of the nine material parameters
A,Q, ␰ ,m,h 0 ,a,ŝ, and n for any given solder alloy were deter-
mined by following the procedures outlined below:
共a兲 Determination of the saturation stresses ␴ * under strain
rates ␧˙ and temperatures T from steady-state creep relation.
共b兲 Nonlinear fitting of Q/R,A,ŝ/ ␰ ,m and n in Eq. 共14兲.

Table 2 Material parameters of viscoplastic Anand model for

solders

Solders
Material
Parameters 60Sn40Pb 62Sn36Pb2Ag 96.5Sn3.5Ag 97.5Pb2.5Sn
A 共s⫺1兲 7
1.49(10 ) 2.30(10 ) 7
2.23(10 ) 4
3.25(1012)
Q/R 共°K兲 10830 11262 8900 15583
␰ 11 11 6 7
m 0.303 0.303 0.182 0.143
ŝ 共MPa兲 80.42 80.79 73.81 72.73
n 0.0231 0.0212 0.018 0.00437
h 0 共MPa兲 2640.75 4121.31 3321.15 1787.02
a 1.34 1.38 1.82 3.73 Fig. 1 Constant strain rate behavior of 60Sn40Pb solder „Ã:
s 0 共MPa兲 56.33 42.32 39.09 15.09 derived from elasto-plasto-creep model; —: prediction from
Anand model…

Journal of Electronic Packaging SEPTEMBER 2001, Vol. 123 Õ 249

 Fig. 3 Plastic strain hardening data for 60Sn40Pb solder „ ␧˙
Ä1.0Ã10À3 sÀ1, TÄÀ25°C…

data can then be calculated. Figure 3 shows the relationship be-

tween d ␴ /d␧ p and ␴ data of 60Sn40Pb solder at temperature
⫺25°C and strain rate 1.0⫻10⫺3 s⫺1 . The corresponding data
based on the separated model were also calculated and plotted in
this figure. It is seen that although the Anand model underesti-
mates the strain hardening at low stresses, both models produce
the coincident results.
As independent verification is provided by steady-state creep
data. The steady-state creep rates of 60Sn40Pb and 96.5Sn3.5Ag
solders from experiments of Darveaux and Banerji 关1兴 and the
model predictions are plotted versus shear stresses in Fig. 4共a兲 and
Fig. 5, respectively. Figure 4共b兲 also shows another independent
verification of creep data for 60Sn40Pb solder around room tem-
perature reported in the literature 共Grivals et al. 关18兴, Solomon
关19兴, Busso and Kitano 关9兴兲. Both Fig. 4 and 5 indicated that the
experimental data and the prediction results are coincident well.
Fig. 2 Constant strain rate behavior of 62Sn36Pb2Ag solder Also, the model predictions capture the characteristic of the power
„Ã: derived from elasto-plasto-creep model; —: prediction from law breakdown for steady-state creep behavior.
Anand model…
2 Steady-State Creep Approximation. One can find the
flow equation 共Eq. 共3兲兲 of Anand model follows the same hyper-
rate behavior and steady-state plastic flow. The temperature of the bolic sine form of the secondary creep law 共Eq. 共11兲兲 with modi-
constant strain rate behavior ranges from ⫺55°C–⫹125°C and fication of internal state variable as denominator in the stress term,
the strain rate ranges from 1.0⫻10⫺2 s⫺1 – 1.0⫻10⫺5 s⫺1. As the together with providing an evolution equation for the state vari-
representative results, Fig. 1 and Fig. 2 illustrate the comparisons able. In electronic packaging applications, because the steady-
between the stress-strain data derived from separated model and state creep deformation processes dominated the deformation ki-
the prediction results from Anand model for 60Sn40Pb and netics of solder alloys due to their low melting point 共Frost and
62Sn36Pb2Ag solders, respectively. A good agreement between Ashby 关11兴兲, the material model for inelastic response of solders
the model predictions and the data from separated model is is often treated as steady-state creep 共to name a few, Holden and
achieved for the above conditions, especially in the regimes where Wallach 关20兴, Wang et al. 关21兴, Hong and Burrell 关22兴兲. There-
the steady-state plastic flow apparently occurs. Also, the tempera- fore, it is interesting to use the Anand model to represent the
ture and strain rate dependence of the onset of plastic flow, and steady-state plastic flow of solders.
the general features of the strain hardening behavior of the stress- As mentioned above, the steady-state plastic flow occurs when
strain curves can be clearly seen from these figures. One can also the deformation resistance s, equals the saturation value s * 共see
see there are some small differences in the transients 共the Eq. 共13兲兲. Then, the evolution equation can be skipped and there is
nonsteady-state segments in the stress-strain response兲. Neverthe- no hardening/softening effect, the hardening/softening constant is
less, the Anand model represents the deformation behavior quite given a value of zero, i.e., h 0 ⫽0 and its strain rate sensitivity a
well. ⫽1. The strain rate sensitivity for saturation s * is also set to be
Under isothermal and constant strain rate conditions, for plastic zero, i.e., n⫽0, resulting parameters s 0 and ŝ possess the same
flow not developed fully, i.e., s⬍s * , the strain hardening data values as s * . Thereafter, the Anand model is simplified to the
d ␴ /d␧ p , that is the slope of the stress versus inelastic strain hyperbolic sine creep form. Some researchers 共Wiese et al. 关23兴,
curve, can be represented as Fusaro and Darveaux 关24兴兲 have already used this rate dependent
␴
d␧ p 冉
⫽ch 0 1⫺
s
s*
a 冊 (16)
model to represent the hyperbolic sine steady-state creep of
solders.
Moreover, the Anand model can also be simplified to be the
The strain hardening data reflects the rate sensitivity and the evo- power law creep form. The secondary creep of solders associated
lution of the internal variable. Based on Eq. 共16兲, the hardening with Norton’s equation is

250 Õ Vol. 123, SEPTEMBER 2001 Transactions of the ASME

 Table 3 Material parameters with Anand model of 60Sn40Pb
solder for power law steady-state creep approximation

A Q/R ␰ m ŝ n h0 a s0

2.09(108 ) 5686 1 0.019 100 0 0 1 100

to be: B * ⫽0.205 1/MPans, ⌬H⫽0.49 ev, and n * ⫽5.25. Because

the stresses induced in solder joints for electronic packaging ap-
plications are usually less than 100 MPa, with an additional con-
sideration to let the condition ␰ /s * ⫽1/100, then the flow equation
of the Anand model has the same power law representation. The
obtained material parameters are listed in Table 3.
3 A Simple Application Example. The viscoplastic Anand
model has been previously used to finite element simulate the
stress/strain responses of solder joints of some electronic pack-
ages, e.g., a chip scale assembly 共Wilde et al. 关26兴兲. Here, only a
simple specimen was selected to perform the simulations. The
specimen is composed of two elastic beams, the Al2O3 ceramic
and FR4 substrate, clamping with solder sandwiched at both ends,
which is similar to the Ford joint specimen 共Pao et al. 关27兴兲, as
shown in Fig. 6. Moreover, the unified constitutive framework
was also simplified to power law steady-state creep 共hereafter,
referred to as Anand approximation model兲. The 60Sn40Pb solder
was selected to form the solder joints. A power law creep model
was also utilized in the simulation to clarify the differences of
stress/strain responses in the solder joint under thermal cycling.
The corresponding material constants for both models were pre-
sented in the above Section.
Due to the symmetrical geometry, only half of the specimen
was meshed with 2D plane strain elements. In the simulation with
the Anand approximation model, the material parameters were
defined in the ANSYS and a viscous element type 共visco106兲 was
selected for the solder. Table 4 lists the elastic properties of the
involved materials. The temperature of thermal cycling ranged
from ⫺55°C–⫹125°C with ramp rate of 36°C/min and dwell time
of 10 min at the two temperature extremes. The reference tem-
perature was treated as 125°C.
Fig. 4 Steady-state creep behavior of 60Sn40Pb solder Figure 7 shows the simulation results of inelastic shear stain
distributions in solder joints at the start of the ⫺55°C dwell of the
third thermal cycle based on both models. The results show that

␥˙ S ⫽B * exp冋 册⫺⌬H n
kT
␶ * (17)
the strain distributions based on both models are coincident well.
The inner lower corner near the solder/FR4 interface demonstrates
the strain localization and may be the potential failure site. Tradi-
where B * is a material constants, ⌬H is the activation energy, k is tionally, the prediction on thermal fatigue life of solder joint is
the Boltzmann’s constant, and n * is the stress component. The
creep properties of eutectic SnPb are determined 共Pao et al. 关25兴兲

Fig. 6 Schematic diagram of a simple specimen

Table 4 Elastic mechanical properties of materials

Young’s modulus Poisson’s Coefficient of thermal

Materials 共MPa兲 ratio expansion 共ppm/°C兲
60Sn40Pb 3.447(104 )⫺151.7T共°C兲 0.316 25
Al2O3 2.76(105 ) 0.3 6.7
FR4 1.6(104 ) 0.3 16
Fig. 5 Steady-state creep behavior of 96.5Sn3.5Ag solder

Journal of Electronic Packaging SEPTEMBER 2001, Vol. 123 Õ 251

 Fig. 7 Distributions of inelastic shear strains in solder joints at start À55°C
dwell of the third thermal cycle

based on the estimation of failure indicator such as the cyclic

inelastic strain range through Coffin-Manson empirical equation
and its modified forms. The cyclic strain range can be obtained
from the stress-strain hysteresis loops. The hysteresis loops of a
element 330 close to the inner lower corner with strain localiza-
tion based on the two models are shown in Fig. 8. It is seen that
the hysteresis loops follow the same cyclic pattern, and the ther-
mal cycle lifetimes can then be assumed approximately the same
due to almost the same inelastic strain ranges.
It is found that the CPU time 共3892 s兲 of finite element simu-
lations for 4 thermal cycles with Anand model is greatly less than
that 共20169 s兲 with creep model for the same precision in the
simple application. Finite element analysis of highly nonlinear
problems is generally very CPU time-consumption as the structure
stiffness matrix must be updated during each time step. The creep
equations are integrated with an explicit Euler forward algorithm
in ANSYS software, which is only efficient for problems having
small amounts of contained creep strains 共0.25%兲. Since an ex-
plicit integration procedure is used, a stability is placed on the
time step size according to the creep ratio, a measure of the in-
crement of creep strain. It is recommended to use a time step size
such that the creep ratio is less than 0.1. Therefore, the simulation
for problems dominated by creep, is very time-consumption. The
integration of the flow equation for the Anand model is, however,
run with an Euler backward algorithm. A consistent stress update
procedure which is equivalent to the backward Euler scheme is
used to enforce the consistency condition and the evolution equa-
tion at the end of the time step. Hence, the CPU time of simula-
tion can be reduced considerably with no strict stability limit.

Discussions
Due to the above results, the unified Anand model can be ap-
plied for representing the inelastic deformation of solders. This
model can also be conveniently used and should be recommended
in finite element simulation of stress/strain responses of solder
joints in service. The reasons are as follows:
Fig. 8 Stress-strain hysteresis loops of the selected element 共a兲 A unified framework for solder behavior. Currently, the
in solder joints under thermal cycling constitutive models of solder alloys, used in electronic packaging,

252 Õ Vol. 123, SEPTEMBER 2001 Transactions of the ASME

are mostly separated as rate independent plasticity and steady- the inelastic deformation of solders. This model is useful in finite
state creep. Classical plastic flow theories and power law creep element simulation of stress/strain responses of solder joints in
are still widely used. The Anand model unifies both rate- service.
dependent creep 共transient creep and steady-state creep兲 and rate-
independent plastic occurring concurrently at the same time in the References
material. The internal state variable, i.e., the deformation resis- 关1兴 Darveaux, R., and Banerji, K., 1992, ‘‘Constitutive Relations for Tin-Based
tance, can represent an averaged isotropic resistance to macro- Solder Joints,’’ IEEE Trans. CHMT, 15, No. 6, pp. 1013–1024.
scopic plastic flow. Moreover, it is usually difficult to separate the 关2兴 Weinbel, R. C., Tien, J. K., Pollak, P. A., and Kang, S. K., 1987, ‘‘Creep
plastic strain from creep strain to calibrate the parameters of a Fatigue Interaction in Eutectic Lead-Tin Solder Alloys,’’ J. Mater. Sci. Lett., 6,
pp. 3091–3096.
separated model based on the mechanical experimental tests. The 关3兴 Kashyap, B. P., and Murty, G. S., 1981, ‘‘Experimental Constitutive Relations
parameters in a unified model can be determined in a direct for the High Temperature Deformation of a Pb-Sn Eutectic Alloy,’’ Mater. Sci.
method combining both rate dependent and rate independent plas- Eng., 50, pp. 205–213.
tic strains into a viscoplastic strain term. 关4兴 Lau, J., and Rice, J. R., 1990, ‘‘Thermal Stress/Strain Analyses of Ceramic
Quad Flat Pack Packages and Interconnections,’’ Proc. Elect. Components and
共b兲 A creep approximation with hyperbolic form. It is shown in Technology 40th Conf., Las Vegas, Nevada, Vol. 1, pp. 824–834.
this paper that the Anand model can be simplified to represent the 关5兴 Sarihan, V., 1993, ‘‘Temperature Dependent Viscoplastic Simulation of Con-
secondary creep of solders. The creep approximation with hyper- trolled Collapse Solder Joint Under Thermal Cycling,’’ ASME J. Electron.
bolic form captures the characteristic of the regime of power law Packag., 115, pp. 16–21.
关6兴 Pao, Y. H., Badgley, S., Jih, E., Govila, R., and Browning, J., 1993, ‘‘Consti-
breakdown, which may be reached for the corners and interfaces tutive Behavior and Low Cycle Thermal Fatigue of 97Sn3Cu Solder Joints,’’
of solder joints under thermal shock and accelerated test condi- ASME J. Electron. Packag., 115, No. 6, pp. 147–152.
tions. The major deficiency of power law creep models is the lack 关7兴 Knecht, S., and Fox, L. R., 1990, ‘‘Constitutive Relation and Creep-Fatigue
of predictive power to correctly describe the stress/strain hyster- Life Model for Tin-lead Solder,’’ IEEE Trans. CHMT, 13,, No. 2, pp. 424–
433.
esis loop during thermal cycle loading. The inaccuracy of the 关8兴 Busso, E. P., Kitano, M., and Kumazawa, T., 1994, ‘‘Modeling Complex
hysteresis loop calculation substantially magnifies the error of fa- Inelastic Deformation Processes in IC Packages’ Solder Joints,’’ ASME J.
tigue life prediction of solder joints. Electron. Packag., 116, No. 3, pp. 6–15.
共c兲 Available in commercial finite element code. There already 关9兴 Busso, E. P., and Kitano, M., 1992, ‘‘A Visco-Plastic Constitutive Model for
60/40 Tin-Lead Solder Used in IC Package Joints,’’ ASME J. Eng. Mater.
existed some unified models for solders in electronic packaging, Technol., 114, pp. 331–337.
but most of them should provide the time integration scheme and 关10兴 Qian, Z., and Liu, S., 1997, ‘‘A Unified Viscoplastic Constitutive Model for
be implemented in the commercial finite element code as user- Tin-Lead Solder Joints,’’ Advances in Electronic Packaging, ASME EEP-Vol.
specified subroutines. Such a work is often complex, expertly de- 19-2, pp. 1599–1604.
关11兴 Skipor, A. F., et al., 1996, ‘‘On the Constitutive Response of 63/37 Sn/Pb
pendent, and time-consuming. The rate dependent plasticity in Eutectic Solder,’’ ASME J. Eng. Mater. Technol., 118, pp. 1–11.
ANSYS code, i.e., the Anand model, provides a useful unified 关12兴 Ma, X., Wang, L., Fang, H. Y., and Qian, Y. Y., 1998, ‘‘Bodner-Partom
model. Constitutive Relations for Eutectic Tin-Lead Solder,’’ Chinese J. Applied Me-
共d兲 Iteration stability and lower CPU time-consumption. As chanics, 15, No. 2, pp. 87–90 共in Chinese兲.
关13兴 Weber, G. G., Lush, A. M., Zavaliangos, A., and Anand, L., 1991, ‘‘An Ob-
mentioned above, the finite element simulation with the Anand jective Time-Integration Procedure for Isotropic Rate-Independent and Rate-
model is more efficient than the creep model. Dependent Elastic-Plastic Constitutive Equations,’’ Int. J. Plast., 6, pp. 701–
744.
It should be noted that only a single scalar used to characterize 关14兴 Anand, L., 1985, ‘‘Constitutive Equations for Hot Working of Metals,’’ J.
the internal structural characteristics of a material is of course a Plasticity, 1, pp. 213–231.
gross simplification. The Anand model was originally formulated 关15兴 Brown, S. B., Kim, K. H., and Anand, L., 1989, ‘‘An Internal Variable Con-
stitutive Model for Hot Working of Metals,’’ Int. J. Plast., 5, pp. 95–130.
to be used more for hot working process rather than for thermal 关16兴 Garofalo, F., 1963, ‘‘An Empirical Relation Defining the Stress Dependence of
cyclic loading. Two features for modeling thermal cyclic loading Minimum Creep Rate in Metals,’’ Trans. AIME, 227, 351–362.
are not considered in the Anand model: 共1兲 kinematic hardening 关17兴 Arrowood, R., Mukherjee, A., and Jones, W. B., 1991, ‘‘Hot Deformation of
and 共2兲 a thermal 共static兲 recovery term in the evolution equation. Two-Phase Mixtures,’’ Solder Mechanics: A State of the Art Assessment.
Frear, D. R., Jones, W. B., and Kinsman, K. R., eds., TMS, Warrendale, pp.
Some experiments of the hysteresis loops under isothermal strain 107–153.
fatigue condition 共Busso and Kitano 关9兴兲 revealed that the solder 关18兴 Grivas, D., Murty, K. I., and Morris, J. W., 1979, ‘‘Deformation of Pb/Sn
exhibits the Bauschinger effect. However, modeling the internal Eutectic Alloys at Relatively High Strain Rates,’’ Acta Metall. 27, pp. 731–
mechanisms, which give rise to the macroscopic Bauschinger ef- 737.
关19兴 Solomon, H. D., 1986, ‘‘Creep, Strain-Rate Sensitivity and Low Cycle Fatigue
fect, would require further state variables. In addition, the impor- of 60/40 Solder,’’ Brazing and Soldering, 11, pp. 68–75.
tance of the Bauschinger effect is somewhat reduced at high ho- 关20兴 Holden, S. J., and Wallace, E. R., 1999, ‘‘Modeling the Effects of Microstruc-
mologous temperatures and relatively low strain rate regimes ture in BGA Joints,’’ Int. J. Microcircuits Electron. Packag., Second Quarter,
important to solder alloys. The static recovery of solder alloys is 22, No. 2, pp. 80–85.
关21兴 Wang, J., Qian, Z., Zou, D., and Liu, S., 1998, ‘‘Creep Behavior of a Flip-Chip
still not presently sufficiently studied. Some further studies on the Package by Both FEM Modeling and Real Time Moire Interferometry,’’
kinematic hardening and the static recovery of solder alloys are ASME J. Electron. Packag., 120, pp. 179–185.
needed in the future. 关22兴 Hong, B. Z., and Burrell, L. G., 1997, ‘‘Model Thermally Induced Viscoplastic
Deformation and Low Cycle Fatigue of CBGA Solder Joints in a Surface
Mount Package, IEEE Trans. CPMT-Part A, 20, No. 3, pp. 280–285.
Conclusions 关23兴 Wiese, S., Feustel, F., Rzapka, S., and Meusel, E., 1998, ‘‘⬘Experimental Char-
In this paper, a unified viscoplastic constitutive model, the acterization of Material Properties of 63Sn36Pb Flip Chip Solder Joints, Proc.
Mat. Res. Soc. Symp., Vol. 515, pp. 233–238.
Anand model, which uses a single scalar internal variable to de- 关24兴 Fusaro, J. M., and Darveaux, R., 1997, ‘‘Reliability of Copper Base-plate High
scribe the deformation resistance state, was applied to represent Current Power Modules,’’, Int. J. Microcircuits Electron. Packag., Second
the inelastic deformation behavior for solders in electronic pack- Quarter, 20, No. 2, pp. 81–88.
aging. The material parameters of Anand model for 60Sn40Pb, 关25兴 Pao, Y. H., Jih, E., Badgley, S., and Browning, J., 1992, ‘‘Thermal Cyclic
Behavior of 97Sn-3Cu Solder Joints,’’ ASME 92-WA/EEP-21, pp. 1–6.
62Sn36Pb2Ag, 96.5Sn3.5Ag, and 97.5Pb2.5Sn solders were de- 关26兴 Wilde, J., Cheng, Z. N., and Wang. G. Z., 1999, ‘‘Influences of Packaging
termined from experimental results and separated constitutive re- Materials on the Solder Joints Reliability of Chip Scale Assemblies,’’ Proc.
lations. The model verifications show that this rate dependent 1999 Inter. Symp. on Advan. Packag. Mater., 共Braselton, Georgia兲, pp. 141–
model can excellently reproduce the base and experimental data. 149.
关27兴 Pao, Y. H., Badgley, S., Govila, R., Baumgartner, L., Allor, R., and Cooper,
Also, a simple application example of simulation on stress/strain R., 1992, ‘‘Measurements of Mechanical Behavior of High Lead Lead-Tin
responses of solder joints under thermal cycling was presented. It Solder Joints Subjected to Thermal Cycling,’’ ASME J. Electron. Packag.,
is concluded that the Anand model can be applied for representing 114, pp. 135–145.

Journal of Electronic Packaging SEPTEMBER 2001, Vol. 123 Õ 253