Thermal fatigue life evaluation of CSP joints by mechanical shear fatigue testing
By Yoshihiko Kanda*1, Kunihiro Zama*1, Yoshiharu Kariya*2, Hironori Oota*3, Syunichi Kikuchi*4, Hideki Yamabe*5,
Graduate school of Shibaura Institute of Technology, *2Materials Science and Engineering Department, Shibaura
Institute of Technology, *3NEC Corporation *4FUJITSU Advanced Technologies Limited, *5NIHON SUPERIOR CO.,
LTD., *6HDP User Group International Inc.
The interrelation between a thermal cycle test and a mechanical shear fatigue test has been studied for CSP joints
from the view point of fatigue life and the microstructural damage of solder joints. The fatigue lives in both methods
are almost equivalent even though loading method is different. From the viewpoint of microstructure, the fact is
attributed to that the transgranular failure is predominant mode and a microstructural coarsening which is induced by
thermal loading and stress equivalent for both the thermal cycle test and the mechanical shear fatigue test.
Since differences in the coefficient of thermal expansion of components under a heat cycle condition imposition
thermal strain on solder joints, the fatigue reliability evaluation is important for the joints in a semiconductor
package. Fatigue reliability evaluation of semiconductor packages is mainly performed through thermal cycle
testing on the package assembled on substrate. The thermal cycle test is very reliable; however, the long test time of
several months makes it rather costly. Therefore, the mechanical fatigue test is expected, in which the test
temperature, strain and loading profile can be independently controlled, and which also has a short test time
compared to the thermal cycle test . However, whether the fatigue life and failure mode for the solder joints are
identical or not for mechanical and thermal testing is unclear, and so clarifying this is an urgent issue. This research
investigates the correlation between both test methods from observations of the fatigue behaviors of CSP joints
obtained from the thermal cycle test and mechanical fatigue test.
2.1 Thermal cycle testing
The test specimen for the thermal cycle test was a CSP (Chip Sized Package) consisting of a 0.4mm pitch size, 144
pin 6 mm x 6 mm x 0.5 mm Si chip joined to a 20 mm x 30 mm x 0.8 mm thick FR-4 substrate by 230 µm dia.
solder balls. Figure 1 shows the schematic illustration of the thermal cycle test CSP. Also, to measure the electric
resistance of the solder joint, a daisy chain circuit was formed in the package. The solder ball consisted of
Sn-3mass%Ag-0.5mass%Cu. The two types of thermal cycle test conditions were as follows: maximum
temperature 398 K, minimum temperature 233 K with a 25-min holding time trapezoidal wave (condition A) and
maximum temperature 373 K, minimum temperature 273 K with a 25-min holding time trapezoidal wave (condition
B). Fatigue life is defined as the cycle number at which the resistance has risen by 10% of its initial value.
2.2 Mechanical fatigue test
The mechanical shear fatigue test specimen, as shown in Fig. 2, was the CSP substrate implemented in the thermal
cycle test provided with a central slit along the short direction and φ3.5 mm and φ2.5 mm holes in the substrate.
The conditions of the mechanical shear fatigue test were set based on the equivalent inelastic strain range of the
solder bump in the thermal cycle test. First, the relationship between the displacement range for the mechanical
shear fatigue test (actuator displacement) and the equivalent inelastic strain range of the corner bumps were
calculated by FEM. Secondly, the inelastic strain range developed in the corner bump by the thermal cycle test was
calculated for the target condition, and the displacement range of the mechanical shear test corresponding to the
equivalent inelastic strain range was found from the above relation. This was then set as the condition for the
mechanical shear fatigue test. Further, the test temperature of the mechanical shear fatigue test was 398 K, which is
the maximum temperature of the thermal cycle test normally carried out as an accelerated test.
The mechanical shear fatigue test used a micro-test fatigue machine . This machine used a piezo stage with a
displacement enlargement structure having a maximum stroke of ±250 µm and a maximum load-cell load of ±40 N.
Displacement was measured using a capacitance displacement sensor (displacement resolution: 0.01 µm) attached to
the test specimen fixture tip and this measured value was fed back to the actuator for control. The heating device—a
ceramic heater attached to the underneath of the tool—heated the test sample fixtures and a thermocouple affixed to
the test sample terminals measured the temperature, which was controlled to ±2 K during testing. The definition of
fatigue life is the same as for the thermal cycle test, that is, the number of cycle at which the impedance has risen by
10% of its initial value.
Interposer Cu post
Solder resist Solder bump
FR-4 substrate Cu pad
Fig. 1 Schematic illustration of CSP specimen for thermal cycle test.
Fig.2 Schematic illustration of CSP specimen for mechanical shear fatigue test.
Fig. 3 FEA models of specimen for (a) thermal cyclic test and (b) mechanical fatigue test.
2.3 Finite Element Analysis
In the FEA, the 1/4 symmetrical model as shown in Fig. 3 is adopted by considering its symmetric boundaries. For
simplicity, solder bumps are assumed to be cylindrical-shaped, with the exception of corner bumps. Elasto-creep
analysis was performed with 3D solid type element (hexagonal 8nodes). The material constants for each component
are summarized in Table 1, and the creep property of a solder bump is given by the Galofaro’s law in Eq. (1).
ε ss = 5.75 × 10 6 [sinh( 0.03σ )] 7.8 exp( −70000 / RT )
where, ε ss is steady-state creep rate(/s) ， σ is stress(MPa) ， R is universal gas constant(J/(mol ･ K))and T is
Table 1 Material constants of each component
Resin (T g=358K) Solder
Solder Si Cu FR-4
T <T g T >T g resisit
76087-0.109T (K) 173 129.3 14 7 31.619-0.0305T (K) 2.4
Poisson's ratio 0.3 0.35 0.34 0.3 0.15 0.29
CTE (ppm/K) 29 2.6 16.6 12.5 25 60
2.4 Microstructural observation
The microstructures were observed under an optical microscope. The specimens were mechanically polished using
SiC polishing paper and diamond paste, and then polished using a colloidal silica suspension liquid to remove the
layers that were damaged in the mechanical polishing process.
3.1 Fatigue life
Figure 4 shows the relationship of the equivalent inelastic strain range and the displacement range calculated by
FEM for mechanical strain fatigue test. From FEM analysis, the equivalent inelastic strain range developed in the
corner bump was around 0.035 for condition A and 0.018 for condition B in the thermal cycle test. From Fig. 4, the
corresponding displacement range of the mechanical strain fatigue test was 19 µm for condition C and 15 µm for
condition D, respectively. The mechanical strain fatigue test was carried out under these two conditions. The
relationship between the fatigue life and equivalent inelastic strain range of the solder joint obtaiend from both test
methods are shown in Fig. 5. The fatigue life obtained in the thermal cycle test had a cumulative fracture
probability of 63.2% from Weibull statistics of sample number 9. The fatigue life of the solder joint obtained from
the mechanical shear fatigue test and the fatigue life obtained from the thermal cycle test show slight variations, but
both lives obtained from the two different methods are almost equivalent.
Equivalent inelastic strain range, Δεin
10 12 14 16 18 20
Displacement range, Δd / μm
Fig. 4 Relationship between equivalent inelastic strain range and displacement range for mechanical shear fatigue
Equivalent inelastic strain range, Δεin
Thermal cycle (ΔT=100K)
Mechanical fatigue (Δd=15μm)
Mechanical fatigue (Δd=19μm)
Chip Size Package
101 102 103 104
Number of cycles to failure, Nf
Fig. 5 Relationships between equivalent inelastic strain range and fatigue lives of thermal cyclic test and mechanical
shear fatigue test.
3.2 Microstructural damage
When comparing thermal cycle life and mechanical fatigue life, it is necessary to investigate the failure mode as
well as the cycle number. Figure 6 shows the initial microstructure of the solder joint and the microstructure
underwent the thermal cycle test, condition B and the mechanical shear fatigue test, condition D (upper photos:
Low magnification polarization image, lower photos: high magnification dark field. White parts represent
intermetallic components of Cu6Sn5 and Ag3Sn).
Firstly, the initial microstructure of the solder joint consists of about 2 crystal grains, as seen in the low
magnification polarization image. After testing in condition B of the thermal cycle test, the microstructure had
about 4 crystal grains, and microstructural change, such as re-crystallization, was not observed. For condition B,
the fatigue cracks were initiated at the solder bump corner and propagated within the crystal grain.
Similarly to the thermal cycle test, the microstructure in condition D of the mechanical shear fatigue test had about
3 crystal grains, and microstructural change, such as re-crystallization, was also not observed. Moreover, also
similarly to the thermal cycle test, the fatigue cracks were initiated in the solder bump corner and then propagated in
grain interior. Thus, the fatigue crack mode for both cases in the thermal cycle test and mechanical shear fatigue
test is thought to be transgranular fracture.
In addition to crack propagation, the important coarsening of intermetallic compounds was investigated from the
viewpoint of the fatigue load environment. From the dark field image, in the initial microstructure, the
intermetallic compounds formed a fine eutectic region. On the other hand, after the thermal cycle test, the
intermetallic components were clearly enlarged, and the eutectic regions is coarsened. In the mechanical shear
fatigue test enlargement of the intermetallic compounds was also confirmed. This enlargement of the intermetallic
components, from the point that they develop due to the heat and stress, suggests that the fatigue load of the thermal
cycle test and the mechanical shear fatigue test are equivalent. The above results indicate that the fatigue life
obtained from the thermal cycle test and the mechanical shear fatigue test are almost the same, the fatigue crack
modes are both equally transgranular failure and, as suggested from the microstructural coarsening, the fatigue
load of both tests are thought to be equal.
10 0μm 1 00μm 100μm
20μm 20μm 20μm
Fig. 6 Optical micrographs showing initial and fatigue tested microstructure.
(a: Initial microstructure b: Thermal cyclic test at temperature range 233K~373K c: Mechanical shear fatigue
test at Δd=15μm)
The results of the thermal cycle test (thermal range 233K - 398K and 273K - 373K) and the mechanical shear
fatigue test (which gave the equivalent inelastic strain range under the same test conditions as the thermal cycle test)
both gave an almost equal fatigue life for the solder bump. This is considered to be due, from a microstructural
viewpoint, to the fatigue load for both testing methods being equivalent and the fatigue failure mode of the solder
bump also being equal. Therefore, the fatigue reliability for the thermal cycle environment can be predicted from the
mechanical shear fatigue test.
 Yoshiharu kariya, Takuya Hosoi, Takashi Kimura, Shinichi Terashima and Masamoto Tanaka, Mater. Trans., Vol.
45, No. 3 (2004) pp. 689 to 694
 Yoshihiko Kanda, Yoshiharu Kariya and Yusuke Mochizuki, Mater. Trans., Vol. 49, No. 7 (2008) pp. 1524 to
This research was supported by HDP User Group International Inc.