An Acceleration Model for Lead-Free (SAC) Solder Joint Reliability
under Thermal Cycling

Vasu Vasudevan* and Xuejun Fan**
* Intel Corporation, 5200 Elam Young Pkwy, Hillsboro, OR 97124
vasu.s.vasudevan@Intel.com
** Department of Mechanical Engineering
Lamar University, PO Box 10028, Beaumont, TX 77710
xuejun.fan@lamar.edu


Abstract
The electronics industry has successfully transitioned from
Sn/Pb to Pb free (LF) solder for computing and consumer
electronics applications. However, there is no industry-wide
standardized LF solder joint reliability model (neither
empirical nor FEA-based) available for solder fatigue
reliability assessment. A LF solder fatigue model has been
proposed in this paper based on a 3-parameter modified
Coffin-Manson approach. The proposed model showed best
fit to the experimental data (17 pairs of temperature cycle test
data) from different sources for multiple package types and
sizes including various test conditions. The model fit to the
experimental data was excellent and the error was less than
6%. This analysis showed that the LF acceleration factor (AF)
model is not significantly different from the Sn/Pb model and
proposed model provides best fit to experimental results.
1. Introduction
While the electronics industry has successfully completed
the transition to Pb-free (LF) technology for computing and
consumer electronics applications, there is still no
standardized LF solder joint reliability model (neither
empirical nor FEA-based) available for solder fatigue.
Numerous studies have been published regarding the thermal
cycling reliability of SnAgCu solder joints under accelerated
test conditions [e.g., 1-5]. The acceleration factor (AF)
models for LF are of particular interests to the electronics
industry for the simple reason that the solder joint reliability
in field power cycling conditions can be predicted in a simple
yet accurate manner. Pan et al [6] first attempted to obtain an
acceleration factor model using several thermal cycle profiles,
in which the temperature range, maximum temperature and
the cycle time or frequency were systematically varied.
However, it has been shown in several studies [7,8] that this
model prediction is not well correlated with experimental
results. Studies by Salmela [8], and Zhang and Clech [7]
evaluated several different types of AF models. Those AF
models also include models based on the compact strain
energy analysis and the finite element analysis. It showed that
the FEA based models have limited success in predicting the
AF factors, especially when test conditions and package types
vary dramatically.
In this paper, a LF acceleration factor model is proposed
based on the Norris-Landzberg equation. A total of 17 pairs
of experimental data suitable for acceleration factor
computations, taken from Intel and industry published test
data, were used to calibrate and support the model. These data
come from various types of packages, such as FCBGA,
CBGA, CSP, QFP and die-side capacitor packages. The test
data also include extended cycle time effect and maximum
temperature as well as the temperature range effect.
2. Review: Lifetime Prediction Models for Solder Joint
Fatigue
Solder joint fatigue is considered as low-cycle failure.
Therefore almost all lifetime prediction models originate from
the Coffin-Manson’s equation
CN
n
p
=∆ )( ε
(1)

where N is the number of cycles to failure, ∆ε
p
is the plastic
strain range per cycle, n is an empirical material constant, and
C is a proportionality factor. A fatigue failure always begins
at a local discontinuity such as the area of stress concentration
near solder/pad interface. Plastic strain accumulates each
cycle resulting in failure. The acceleration factor (AF) can be
defined based on equation (1) as follows
n
p
p
N
N
AF
−
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
∆
∆
==
2
1
2
1
ε
ε
(2)

where N
1
and N
2
are fatigue life under two different cyclic
loading conditions, which correspond to the plastic strain
range ∆ε
p
1
and ∆ε
p
2
respectively. Equation (2) has been
widely used since the field life (e.g. N
1
) can be related to a
test life (e.g. N
2
) with an empirical material constant n if the
plastic strain range at each loading condition can be
calculated. Norris and Landzberg [9] assumed that the plastic
strain range is proportional to the temperature excursion
range, and further introduced two more factors to account for
the effects of temperature-cycling frequency f and the
maximum temperature Tmax of the solder material. By doing
this, the AF equation obtains the form:
⎟
⎟
⎟
⎠
⎞
⎜
⎜
⎜
⎝
⎛
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
∆
∆
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
==
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
−
−−
testfield
a
TTk
E
nm
e
T
T
f
f
N
N
AF
max,max,
11
test
field
test
field
test
field
(3)
where “ field ” and “ test ” denote the field and the test
condition, respectively. The Norris-Landzberg model can also
be used to compare two different field or test conditions. E
a
is
the activation energy, and k is Boltzmann’s constant. For
SnPb eutectic solder,
m=1/3, n=1.9, and E
a
/k = 1414 (4)
978-1-4244-2231-9/08/$25.00 ©2008 IEEE 139 2008 Electronic Components and Technology Conference