You are on page 1of 12

Journal of ELECTRONIC MATERIALS, Vol. 37, No.

4, 2008

Special Issue Paper

DOI: 10.1007/s11664-008-0393-8
 2008 TMS

Study of Fracture Mechanics in Testing Interfacial Fracture


of Solder Joints
W.H. BANG,1,2 M.-W. MOON,1 C.-U. KIM,2,4 S.H. KANG,1 J.P. JUNG,3
and K.H. OH1
1.School of Materials Science and Engineering, Seoul National University, Seoul 151-742,
Korea. 2.Department of Materials Science and Engineering, University of Texas at Arlington,
Arlington, TX, USA. 3.Department of Materials Science and Engineering, University of Seoul,
Seoul 130-743, Korea. 4.e-mail: choongun@uta.edu

This paper is concerned with the mechanics of interfacial fracture that are
active in two common testing configurations of solder joint reliability. Utilizing eutectic Pb-Sn/Cu as a reference system and assuming the presence of a
predefined crack size in the intermetallic compound (IMC) layer, stress
intensity factors (KI and KII) at the crack are numerically calculated for the
two given configurations. The analysis of the tensile test configuration reveals
that the fracture occurs by the crack-opening mode (KI mode), as anticipated,
but that it is greatly assisted by the viscoplasticity of the solder. With nonuniform viscoplastic deformation across the joint, KI is found to increase much
more rapidly than it would without the solder, decreasing the critical crack
size to the micron scale. The same mechanism is also responsible for the
development of a KII comparable to KI at the crack tip, that is, |KI/KII|  1. It
is also found that the predominant fracture mode in the bump shear configuration is crack opening, not crack shearing. This is an unexpected result, but
numerical analyses as well as experimental observations provide consistent
indications that fracture occurs by crack opening. During shear testing, bump
rotation due to nonzero rotational moment in the test configuration is found to
be responsible for the change in the fracture mode because the rotation makes
KI become dominant over KII. With rotational moment being affected by the
geometry of the bump, it is further found that the fracture behavior may vary
with bump size or shape.
Key words: Solder joint reliability, intermetallic layer, tensile test, bump
shear test, stress intensity factors (KI and KII)

INTRODUCTION
Sn-based solder alloys are widely used in electronic packaging applications as they offer several
advantageous properties, including low processing
temperatures and excellent wetting of common Cu
metallization components. With the continuing
thrust toward smaller and higher-performance
packaging, however, the alloys are becoming a
source of critical reliability concerns. In advanced

(Received April 13, 2007; accepted January 16, 2008;


published online February 13, 2008)

packaging, solder joints are required to endure


increasingly punishing mechanical load conditions,
and thus they are increasingly prone to physical
failures. Of these physical reliability failures, one of
the most troubling is fracture at the interface between solder and metallization. The interface is
known to be the weakest mechanical link in all
solder joints because the presence of brittle Sn-Cu
intermetallic compounds (IMC, Cu6Sn5 or Cu3Sn)
weakens the overall resistance against fracture and
also creates a point of stress singularity at which
cracks can initiate.16 In order to enhance the
overall reliability of the solder joint, it is therefore

417

418

Bang, Moon, Kim, Kang, Jung, and Oh

critical to prevent or retard the interface fracture,


and this has been the main focus of recent studies in
this field. Progress in this area demands a more indepth investigation of interface fracture mechanics.
There are two common methods of characterizing
the mechanics of interface fracture: the joint tensile
test and the bump shear test.5,710 The tensile test
requires the induction of interfacial fracture by
exerting a load in a direction normal to the interfacial layer, while the bump shear test requires the
application of a force in a direction parallel to the
interfacial layer. The simplicity of these methods,
combined with their ability to provide quantitative
information and fractography, has made them the
most popular approaches for the investigation of
interface fracture in solder joints. However, such
studies produce data with clear limitations. Real
solder joints are subjected to complex load conditions and, as a result, their fracture may proceed by
significantly different mechanisms than the ones
found in the tensile or bump shear test. Further,
because of the time-dependent mechanical properties of the solder, the fracture mechanism under
differing load conditions of a single test method
(bump shear or tensile joint) may be sufficiently
different to lead to the need for nonsimplistic
methods of analysis. In order to extrapolate information from the common test configurations to real
solder joints, it is imperative to investigate more
carefully how fracture occurs in each of these two
methods. This quest requires the study of fracture
mechanics in the IMC layer to determine the fracture criteria, including the critical crack size and
the fracture mode under a given load condition.11,12
This paper presents the results of our investigation on the interface fracture mechanics of the bump
shear and joint tensile test methods. Specifically,
this paper investigates the fracture mode by conducting numerical analysis of the stress intensity
factors (K) at the crack tip in the IMC layer for each
method. The cases simulated here are fractures at
the brittle IMC layer created by a reaction between
the Sn-Pb eutectic solder and Cu substrate under
geometrical constraints and load conditions relevant to the tensile test and a bump shear test. Our
analysis yields several interesting and somewhat
surprising results of practical importance. We find
that the fracture behavior for the two methods, that
is, the crack-opening (KI) mode for the tensile and
the crack-shearing mode (KII) for the bump shear
testing, are not as simple as each method intends to

produce. In the case of the tensile test condition, it is


found that the viscoplastic deformation and Poisson
contraction of the solder matrix have a significant
influence on the stress field around the crack tip.
With viscoplastic deformation of the solder, KI increases much more rapidly than without solder,
making the IMC more prone to cracking. Furthermore, Poisson contraction causes the crack in the
IMC to develop a KII value with a magnitude comparable to that of KI. Even more unexpected is the
result that the crack propagation in the bump shear
test condition occurs by the KI mode not by the KII
mode. It is found that the KI mode prevails because
of bump rotation during crack propagation. This
result is surprising but our experimental investigation produces supporting evidence. The rotation
of solder bumps in the shear direction is clearly
seen, resulting in the fracture site being lifted up
from the substrate and the development of crack
opening (KI) during shearing. These results strongly
suggest that any characterization of the interface
fracture of solder joint, especially characterization
conducted with the two common methods, needs to
include careful consideration of the solder deformation and rotation with variations in geometrical
and load conditions.
BACKGROUND AND NUMERICAL AND
EXPERIMENTAL PROCEDURES
Background
While there exist a few experimental results for
solder joint fracture in tensile tests, the study of SnPb/Cu joints is most comprehensive and thus provides the data needed for our modeling work. The
data used in our modeling is taken from Sn-Pb/Cu
tensile testing conducted by Lee et al., Prakash
et al., and Quan.79 Table I shows the summary of
the tensile test data along with the experimental
conditions used in these studies. While there are
differences in the experimental details, such as the
tension rate and specimen configuration, the tests
show reasonable agreement in the fracture strength
value of 80 MPa. In spite of this agreement,
Prakash et al. and Quan report somewhat different
observations on the crack propagation path. Quan
notes that the fracture proceeds through the IMC
layer (Cu6Sn5) and therefore attributes the data to
the fracture strength of the IMC layer. On the other
hand, Prakash et al. suggest, based on the fractography showing the mixture of IMC and solder

Table I. Tensile Testing Results of Cu/Sn-Pb/Cu Joints (As-Soldered Samples)

Lee et al.7
Prakash et al.8
Quan9

Fracture
Strength

Joint
Thickness

Cross Section
of Joint

Tensioning
Rate

IMC
Thickness

77 MPa
86 MPa
83 MPa

750 lm
200  300 lm
1.0 mm

3 mm (diameter)
10 mm 3 mm
7.6 mm 2.8 mm

0.6 mm/min
0.5 mm/min (0.66 MPa/s)
0.05 mm/min (0.47 MPa/s)

1 lm
2.5 lm
1  3 lm

Study of Fracture Mechanics in Testing Interfacial Fracture of Solder Joints

matrix, that the measured fracture strength may be


related to the decohesion strength between the solder and the IMC layer. However, since such fractography can result from the cracks repeatedly
switching between the solder and the IMC layer,
and in such cases the brittle IMC phase would limit
the overall fracture strength of the joints, it seems
reasonable to assume that the measured fracture
strength is attributable to the fracture strength of
the IMC layer. The tensile load (stress)displacement (strain) curves presented by Prakash et al.
and Quan also provide supporting evidence for this
assessment. In both cases, an abrupt drop in load
after its maximum is visible. This is not possible
without rapid crack growth through a brittle phase.
While numerous studies using the bump shear
method have correlated the test results to IMC
fracture, the resulting data requires consideration
of its complexity. The difficulty of analyzing the
fracture mechanics of this method stems mainly
from the fact that the test load is imparted to the
interface area through a bump that is softer than
the IMC phase at the interface. As is well pointed
out by Huang et al. and Kim et al. in their finite
element method (FEM) analysis,13,14 the viscoplasticity of the bump in such a case may play a
critical role in the fracture at the interface. Perhaps more importantly, shearing load, imparting
nonzero rotational momentum to the bump, may
result in physical motion of the bump that changes
the mechanical constraints at the crack tip. This
possibility has been proven to exist in our experimental and numerical studies. It is therefore
necessary for fracture analysis to include consideration of the viscoplasticity as well as the bump
movement.
FEM Modeling
Our numerical analysis of the fracture mechanics
begins with the assumption that the fracture initiates and proceeds predominantly through the brittle IMC layer in both configurations, which is a
reasonable assumption as previously mentioned.
Consequently, the modeling requires calculation of
the stress intensity factors, K, at the crack tip in the
IMC layer under the two different test configurations while the fracture strength of the IMC layer is
set to 80 MPa. For this numerical analysis, we use a
commercial FEM program, ABAQUS 6.3 with the
finite element mesh structure shown in Fig. 1a and
b. As is shown in these figures, the IMC layer between the eutectic Sn-Pb solder and Cu is assumed
to be 1-lm-thick Cu6Sn5.2,9 Table II lists the
mechanical properties of the materials used in our
FEM analysis. It is further assumed that the test is
conducted at room temperature, 300 K, in order to
be consistent with the existing data and common
testing practice. This temperature is a high homologous temperature for eutectic Sn-Pb solder,
0.65Tm, and thus the viscoplastic characteristics
of eutectic Sn-Pb are incorporated into the finite

419

Fig. 1. Finite element models and crack mesh for estimating


stress intensity factors, K, at the crack tip in the IMC layer during
joint tensile and bump shear tests: (a) FEM meshes for joint
tensile analyses; (b) FEM meshes for bump shear analyses with
calculations performed for different bump sizes, R; and (c) the
crack tip mesh defined in the IMC layer in the dashed-circle area
in (a) and (b).

Table II. Mechanical Properties Used in the FEM


Analysis15

Copper
Cu6Sn5
Eutectic Sn-Pb
FR4 substrate
Steel shear bar

Elastic Modulus

Poissons Ratio, m

117 GPa
85 GPa
30 GPa
22 GPa
220 GPa

0.35
0.31
0.40
0.28
0.3

element calculation. The viscoplastic constitutive


equation used is:


69 kJ=mol
3:2
pl
2
e_ 2:68  10  r  exp 
RT


64 kJ=mol
6:2
6
1:38  10  r  exp 
;
RT
(1)

420

Bang, Moon, Kim, Kang, Jung, and Oh

where e_ pl is the plastic strain rate, r (MPa) is the


stress, R is the gas constant, and T is the temperature.16 Figure 1c shows the mesh structure for a
predefined crack in the IMC layer for the calculation
of the stress intensity factors in the joint tensile and
bump shear tests. The K factors are evaluated at
five finite element contours surrounding the crack
front from one crack face to the opposite crack
face.17 The detailed description of Fig. 1ac and our
FEM analysis procedure follow.
Tensile Analysis
The two-dimensional finite elements in Fig. 1a
are a simplification of the butt-joined tensile specimen used in Quans experiment.9 A quarter section
is modeled using symmetric boundary conditions
along the transverse and longitudinal directions.
Considering the fact that the specimen width is far
greater than the joint thickness, a plane-strain
condition is employed.9,18,19 From the tensile-load
development curve and data presented by Quan, the
applied tensile rate was determined to be 0.3 MPa/s
and this value is used for our main calculation.
Further, since the outer edge of the joint is the most
likely place for crack initiation, being subjected to
the highest interfacial stress singularity,6,20 the
crack tip mesh in the IMC layer is placed at the end
(indicated by a dashed circle in Fig. 1a). In this way,
the calculation of the K factors with variation in the
size of predefined crack is performed.
The finite element model in Fig. 1b is a cross
section of a eutectic Sn-Pb bump reflowed on a
copper pad. The crack mesh in the IMC layer is
defined at the bump corner (indicated as a dashed
circle in Fig. 1b), as this is the expected place for
crack initiation. The development of the K factors
with shear displacement is calculated for a given
size of predefined crack. In our calculation, the
interfacial area is fixed and we include the shape
and size of the bump as variables, as they may vary
in real testing situations and also have a significant
influence on shear stress development at the crack
tip. Table III details the shape and size of the bump
considered in our simulation, expressed in terms of

Table III. Geometrical Parameters in Fig. 1b at


Different Bump Sizes

25
30
35
40
45

deg
deg
deg
deg
deg

Instead of analyzing published data, we conducted the bump shear test for a selected case in
order to acquire the information needed for our
analysis. This test is especially necessary to identify
any bump motion during the test. For this test, we
prepared the eutectic Sn-Pb bump shear samples
following standard procedures. The solder balls
used in our experiment had a diameter of 760 lm.
The balls are attached to the 650-lm-diameter Cu
pad by reflow treatment at 215C for 1 min. They
were then subjected to the shear test while the load
and displacement are recorded. For selected cases,
position markers, in the form of small indentations,
were made on top of the bump before shear testing
for the purpose of tracing the bump movement.
These markers, consisting of three consecutive
indentations along the centerline of the bump, were
made by creating a small impression using a microindenter. The marker position was then compared
before and after shear testing using scanning electron microscopy (SEM, refer to Fig. 8).
RESULTS

Bump Shear Analysis

the geometrical parameters U, DX, and R. The definition of these parameters can be found in Fig. 1b.
The key variation in these parameters is the radius
R. The others are adjusted based on the wetting
behavior of the bump for the given size of the
interfacial area.
Bump Shear Test

DX

35 lm
50 lm
75 lm
100 lm
135 lm

360 lm
375 lm
400 lm*
425 lm
460 lm

*R  400 lm is a case when a 760-lm-diameter solder ball ideally wets on a 650-lm-diameter copper pad (the solder bump
configuration used in our bump shear experiment)

Fracture Analysis in Joint Tensile


Our numerical results reveal that the stress
intensity factors in the butt-joint tensile test show
radically different behaviors from those found in the
ordinary situation in which homogeneous material
is tested. Immediately noticeable is the fact that, in
addition to the nominal tensile stress, a substantial
amount of shear stress (in the direction of crack
propagation) develops around the crack tip. It is in
fact found that the level of KI and KII developed at
any crack size and load stress is almost comparable,
KII/KI  1. This behavior, i.e., almost equal development of KI and KII, is shown in Fig. 2 in which the
stress intensity factors (KI as positive and KII as
negative) are shown as a function of incremental
tensile load for various sizes of predefined cracks.
The loading rate used in this calculation was
0.3 MPa/s. It can be seen that, for any given crack
size and comparable loading rate, the magnitude of
KI and KII increases with the applied stress.
The reason for the KII development shown in
Fig. 2 is found to be nonuniform Poisson contraction
in the transverse direction (normal to the axial load
and parallel to the crack propagation direction). The
elastic modulus and yield strength of the solder
matrix are far lower than those of Cu and IMC, and
therefore it is subjected to a large Poisson deformation upon axial loading. However, full-scale solder
deformation can occur only in the area where there

Study of Fracture Mechanics in Testing Interfacial Fracture of Solder Joints

Fig. 2. The developments of KI and KII at the IMC crack tip for different crack lengths Da.

Fig. 3. Transverse directional displacement contour in a solder joint


at a tensile load level of 80 MPa (a quarter section model with ydirection boundary conditions along the bottom and x-direction
boundary conditions along the left edge, refer to Fig. 1a).

is little constraint imposed by the stiffer Cu and


IMC. Therefore, the contraction of the solder is
greater in the area distant from the interface, while
it is practically limited to the level of the Poisson
contraction of Cu (and IMC) near the interface. This
nonuniform Poisson deformation field creates a large
shear stress in the transverse direction, resulting
in a high value of KII at the crack tip in the IMC.
Figure 3 shows the displacement contour in the
transverse direction of the joint under 80 MPa
tensile load at the macroscopic scale. An uneven
contraction deformation field in the joint is observed.
Note that a significant transverse contraction occurs
in the solder in the area distant from the IMC while
it becomes smaller closer to the interface. Figure 4
shows the influence of the nonuniform Poisson contraction in the solder on the deformation field at the
crack tip area. This figure compares the degree of
axial opening displacement (DDT) and transverse
shearing displacement (DDS) calculated in the crack
tip area as a function of axial load. The displacements presented here are taken by selecting a nodal
plane that is 0.01 lm behind the tip of a 0.5-lm-long
crack and subsequently calculating the relative displacement of the nodes in the axial and transverse

421

Fig. 4. Crack opening degree (DDT) and shearing degree (DDS) at


the IMC crack tip during tensile test: DDT = tensile-direction coordinate of the upper crack-plane node () minus the tensile-direction
coordinate of the lower crack-plane node (s); DDS = shear-direction
coordinate of the upper-crack plane node () minus the transversedirection coordinate of the lower crack plane node (s). The upper
crack-plane node () and the lower crack-plane node (s), at which
data are derived, are 0.01 lm distant from the crack tip.

directions upon loading. Since the measurements


are taken in close proximity to the tip, the absolute
magnitude of the displacement in both directions
appears small, but it should be noted that the axial
and shear stresses developed by such a displacement
are in reality extremely large.
Even though the stress field around the crack tip
in the IMC includes a considerable shear stress
component, it is not likely that the fracture mode is
significantly affected by it. KI is known to dominate
fracture even when KI and KII are comparable.
Furthermore, since our calculations find that KI is
actually higher than KII at the crack tip, it is easy to
conclude that IMC fracture occurs by the crackopening mode. Nevertheless, the mechanism leading to shear stress development at the crack, that is,
the restriction of solder deformation near the IMC,
causes the IMC fracture behavior to be substantially different from that expected from tensile
testing of a single-phase material in two ways. The
first, perhaps with the most significant practical
importance, is the fact that the IMC phase becomes
far more brittle than without the solder. A simple
analysis of the critical crack size for IMC fracture
may be helpful in elaborating our conclusion.
According to the available data,3,4,19 the critical
KI (KIC) for the fracture ofpthe
Cu6Sn5 IMC ranges
from 1.4 MPa to 2.4 MPa m. The data in Fig. 2
then suggest that, at the assumed fracture strength
of 80 MPa, KI would reach these critical values with
just a 0.5lm to 0.8 lm crack. This means that a
small crack of submicron dimensions would be
sufficient to trigger fracture of the IMC layer in the
solder joint. In contrast, in the case in which
the IMC layer is not structurally associated with the

422

Bang, Moon, Kim, Kang, Jung, and Oh

solder, the critical crack size is found to be much


larger. If it is assumed that the IMC is purely elastic
and tested as a homogeneous sample, the mode I
stress intensity factor
developed at the crack tip
p
should be given by r pc. Under an identical fracp
ture condition, that is KIC = 1.4 MPa to 2.4 MPa m
and rf = 80 MPa, the critical crack size is estimated
to be 98 lm to 290 lm, which is roughly two orders
of magnitude longer than that for the IMC with the
solder. Although simplified, our analysis suggests
that the presence of the solder in the joint causes
the crack in the IMC to propagate much more easily.
It is our finding that the reduced critical crack size
with solder stems from exaggerated crack opening
caused by nonuniform solder deformation. Similar
to the case of Poisson contraction, the axial deformation of the solder is restricted near the IMC but is
not restricted further away. This nonuniform
deformation in the axial direction induces a bending
deflection at the IMC, making the crack open more
than in the homogeneous case, as illustrated in
Fig. 5a and b. These figures compare the deformation field at the crack in the case of homogeneous

IMC (Fig. 5a) and the case of IMC with solder


(Fig. 5b). Note that the crack opens considerably
more with the solder present, causing KI to increase
much more quickly upon loading. Therefore, it is not
so surprising that the critical crack size for the IMC
is far smaller in solder joints.
The second impact of the restricted solder deformation near the IMC is found to be the relative
insensitivity of fracture conditions to the strain
rate. Since solder is a material with considerable
viscoplasticity, it is not unreasonable to expect that
fracture, even if it occurs at the IMC, would be
influenced by the strain rate. However, as shown in
Table I, experimentally measured fracture strength
is remarkably insensitive to the tensioning rate
used in these studies. As can be seen in Fig. 6, in
which calculated K factors as a function of loading
rate are presented, our numerical calculations appear to reproduce such insensitivity. The results
shown here assume the presence of a 0.5-lm-long
crack at the IMC and the application of a 80 MPa
load stress. Note that the K factors remain nearly
constant in the loading rates between 0.03 MPa/s
and 3 MPa/s, while they start to show a noticeable
decrease as the rate exceeds 3 MPa/s. Close
inspection of the results, including the calculated
deformation field in the joint, suggests that the
loading rate insensitivity of the K factors up to
3 MPa/s stems from the quick saturation of the
viscoplastic deformation of th solder near the IMC
layer. As has been previously discussed, the K factors are strong functions of the crack-opening
magnitude and are greatly affected by the deformation field in the solder. At the low and intermediate loading rate, the viscoplastic deformation
contributes greatly to the total deformation field
responsible for crack opening. However, after its
quick development, viscoplastic deformation stops
(saturates) because of the constraints from he stiffer
Cu and IMC. Since the total solder deformation near
the IMC is ultimately limited by the deformation of

Fig. 5. Crack tip deformation at tensile stress of 80 MPa for two


different structures: (a) homogeneous elastic material of Cu6Sn5; (b)
Cu6Sn5 intermetallic layer formed between Sn-Pb and Cu.

p
Fig. 6. K factors (MPa m) at 80 MPa for tensioning rates ranging
between 0.03 MPa/s and 16 MPa/s.

Study of Fracture Mechanics in Testing Interfacial Fracture of Solder Joints

423

the Cu and IMC, the total amount of plastic deformation, and thus the degree of crack opening,
remains unchanged when the applied strain rate is
low enough. This results in near-constant K factors
regardless of the loading rate, between 0.03 MPa/s
and 3 MPa/s. The K factors start to decrease when
the strain rate is fast enough that viscoplastic
deformation of the solder does not saturate, resulting
in a lesser degree of crack opening. It should be noted
that the strain rate insensitivity, seen in both the
experimental work in Table I and the numerical K
factor analysis, is possible because the critical crack
size is small. If the critical size were much longer
than that determined in our study, significant viscoplastic deformation should occur before the crack
reaches the critical size, making K factors vary more
sensitively with the strain rate and joint geometry.
Fracture Analysis on Bump Shear Test
Our investigation of fracture in bump shear tests
includes both experimental characterization and
numerical analyses of the fracture mechanics, and
these studies also yield several new findings of
practical importance. The most striking finding
made in our investigation is probably the result that
the fracture in bump shear tests does not proceed by
the pure shear mode but occurs mainly by the crackopening mode. This mode I fracture in the bump
shear test may be an unexpected result because the
test induces fracture by exerting shear force to the
bump. However, our investigation reveals that
nonzero rotational moment in the test configuration
makes the solder body rotate in the shear direction
during testing, and thus results in significant
change in KI than KII. This result, combined with
the viscoplasticity of solder, causes the fracture
analysis of bump shear testing to be much more
complicated than desired, yet it produces several
interesting results. Among these findings, the
occurrence of solder bump rotation and its impact on
the fracture process are detailed in this paper.
Figure 7 displays an example of the load
displacement results obtained in our experiments.
This particular test is conducted using 760-lmdiameter solder ball reflowed on a 650-lm-diameter
Cu pad with a shear probe displacement rate of
200 lm/s. While similar tests are conducted with
variation of test conditions, including the bump size
and displacement rate, they all show essentially the
same trend, that is, a rapid increase of the load to
saturation (30 lm in the considered in Fig. 7) followed by a gradual decrease of the load after some
duration at saturation. The loaddisplacement
relationship shown in Fig. 7 is also consistent with
the results obtained in many other studies in which
the saturation load is attributed to the point of
cracking. As the cracking appears to occur by shear
loading, the saturation (maximum) load is typically
reported relative to the shear strength or adhesion
strength of the solder joint. However, our SEM
observations reveal that fracture in bump shear

Fig. 7. Shear loaddisplacement in a bump shear test.

tests does not occur by a simple shear process but


involves a rotation of the solder body. Evidence for
the bump rotation is shown in Fig. 8, which presents a series of SEM micrographs of bumps
(760 lm diameter, 200 lm/s shear rate) after being
sheared to four different distances: 50 lm, 150 lm,
300 lm, and 500 lm. The occurrence of bump rotation can be seen from the movements of the indentation markers on each bump. Their position shows
a gradual increase in the rotation of the bump in the
shearing direction. In the case in which the shear
distance is small, the rotational motion of bump is
not clear because it is too small to resolve under
SEM. However, the rotation is clearly visible at the
larger shear distances of 300 lm and 500 lm.
The rotational motion of the bump seen in our
experiment may appear to be an anomalous result;
however, it is found that the rotation is in fact
inevitable because the test configuration yields a
nonzero rotational moment. If the test object had a
uniform surface contact area with the shear probe,
e.g., a cubic bump, the rotational moment would not
exist. However, in the case of the bump test, the
contact between the probe and the solder is not
uniform but rather is sharp because the bump has a
round surface. Because the exertion of the shear
force is achieved by a small contact while the wholebody displacement of the bump is prevented by the
soldered interface at the bottom, a rotational moment arises and the crack tip acts as a rotation axis.
Our numerical analysis of the deformation field also
provides supporting evidence for the existence of the
rotational motion of the solder. Figure 9 show the
displacement contour developed in the vertical
direction at 20 lm shear of a R  400 lm size bump
with a 0.5 lm crack located at the bump corner
(indicated as a circle). Figure 9a shows clear evidence of this nonzero rotational moment, as it shows
the development of vertical deformation contours
during shear. In this case, the solder is assumed to

424

Bang, Moon, Kim, Kang, Jung, and Oh


50m shearing

150m shearing

300m shearing

500m shearing

Side view of bump


before shearing

Side view of bump


after shearing

Rotated view of
sheared bump

Fig. 8. SEM observation of the mechanical motion of a solder bump in a bump shear test (the arrow in the first column indicates the shearing
direction of the shear probe). Two important motions of the solder bump can be seen: the movements of the micro-indented mark at bump top
indicate the rotation of the solder bump in the shearing direction; the arrows point to compressed areas of solder bumps due to contact with the
shear probe.

Fig. 9. Upward-displacement contour in bumps at a shear distance


of 20 lm. Notice that the dashed circles indicate the location of the
crack tip, which is lifted up from the substrate by the rotation motion
of the bump in the shearing direction: (a) Perfect elastic bump: a
result from the FEM model incorporating only the elasticity of Sn-Pb
but not incorporating the viscoplasticity; (b) viscoplastic solder bump:
a result from the FEM model that includes the viscoplasticity of
Sn-Pb (Eq. 1).

be purely elastic for simpler analysis and better


visualization. Note that the displacement contour is
consistent with what is expected from deformation
with rotational motion, that is, the development of
vertical tensile strain at the probe contact area,
including near the crack tip, and compressive strain
at the opposite end. Inspection of the deformation
contour with varying shear distance reveals that the
tensile component arises immediately with bump
contact of the probe tip. The advancement of the
shear probe increases the tensile strain as the bump
rotates more in the shear direction.
The viscoplasticity of solder does not change the
general trend of bump rotation and its evolution with
shear distance, but creates a more complex evolution
of the strain field at the crack-tip area due to the
addition of plastic mass flow. The initial displacement response of the bump is the same as in the
elastic case. Upon contacting the probe tip, the bump
starts to rotate and creates a tensile strain field at the
crack-tip area. However, the continuation of the
shear induces an increasing degree of plastic deformation in the area in contact with the tip (as seen in
Fig. 8), which induces mass flow into adjacent areas
and adds a compressive strain component to the total
deformation field. The strain field near the crack tip
is found to be the most significantly affected by such
plastic flow. Our calculation shows that it decreases
the overall rate of tensile strain development at the
bump corner compared to the pure elastic case, and
in some shear distance ranges the tensile strain
shows a negative dependence on the probe displacement. The vertical displacement contour shown in
Fig. 9b demonstrates the influence of the viscoplastic
flow. It can be seen that, compared to the elastic
contour shown in Fig. 9a, the bump corner area
exhibits much smaller tensile deformation.

Study of Fracture Mechanics in Testing Interfacial Fracture of Solder Joints

Fig. 10. K developments at the IMC crack tip during shearing of a


400-lm-radius solder bump (refer to Fig. 1b and Table III).

The most critical impact of the bump rotation is


the fact that it promotes mode I fracture because the
resulting deformation opens up the crack tip in the
vertical direction and thus makes KI greater than
KII. Figure 10 shows one of many examples evidencing KI dominance. This figure shows the stress
intensity factors (KI and KII) as a function of the
shear distance for R  400 lm with a 0.5 lm crack in
the IMC layer. Note that KI is far greater than KII
and, at any shear distance, KI/KII  3. Similar
calculations conducted with varying bump and crack
sizes produce the same result. The dominance of KI
suggests that the fracture mechanics in bump shear
testing is not much different from that in tensile testing, and that the crack propagation would commence
p
when KI exceeds KIC = 1.4 MPa to 2.4 MPa m.
Another critical impact of the bump rotation is
that it makes the interfacial fracture behavior sensitive to the bump size or shape. One of the main
applications of the bump shear test is the quantitative characterization of interfacial adhesion
strength. It is desired that the resulting data should
ideally yield consistent interfacial strength without
sensitivity to bump size as long as the interfacial
area is the same. However, our analysis indicates
that such consistency may not be possible in the
bump test configuration due to intrinsic dependence
of the KI development on bump size. Such sensitivity is shown in Fig. 11, in which KI development
with shear distance is shown for different bump sizes. In this calculation, the interfacial area is kept
constant, copper size  325 lm, while the bump
radius varies from 360 lm to 460 lm (see Fig. 1b
and Table III). Since the interfacial area is assumed
to be the same, different bump size changes the
bump shape, and its influence is included in our
calculation, as indicated in Table III as the parameters U and DX. Also, the presence of a 0.5 lm crack
at the bump corner is assumed in the calculation.
The KI development shown in Fig. 11 presents a

425

Fig. 11. KI values at a 0.5 lm crack tip for different bump sizes (see
Fig. 1b) and Table III for detailed geometry descriptions).

clear indication that the interface fracture is sensitively affected by the variation in bump size. It can
be seen that the KI development with shear distance
is much more rapid
pin
larger bumps (or longer DX).
If KIC = 1.4 MPa m is assumed for the IMC, the
interfacial fracture would take a shear distance of
about 6 lm for a 460-lm-radius bump (DX = 135
lm), while it takes almost 18 lm when the bump
radius is reduced to 375 lm (DX = 50 lm). This result indicates that larger bumps become more brittle and are more prone to interfacial fracture.
Bump rotation is the primary source of the size
effect shown in Fig. 11. The decisive factor affecting
KI at any given shear distance is the degree of crack
opening at the crack tip. As previously discussed,
crack opening occurs by bump rotation, and its
magnitude is proportional to the rotational moment.
Because the rotation axis is located at the crack tip,
the rotational moment at any given time should be a
linear function of the distance between the crack tip
and probe tip. In this sense, the DX defined in
Table III and Fig. 1b is a good measure of the rotational moment, as larger DX means larger crack tipto-probe distance. In a larger bump, DX is larger,
making it subject to a higher rotational moment. As a
result, the rate of KI increase with shear distance
becomes higher in bigger bumps. This explains the
result shown in Fig. 12, in which the KI data shown in
Fig. 11 is normalized with DX and plotted as a function of shear distance. Notice that the normalization
makes the KI of different bumps reasonably close.
While the normalized KI values in Fig. 12 exhibit
reasonable agreement for different bump sizes,
there still exist minor but measurable differences.
The difference is particularly significant for the
smallest bump (R = 360 lm, DX = 35 lm). Interestingly, unlike in other bumps, KI in this case does
not increase monotonically with the shear distance
but shows a sinusoidal dependence on the shear
distance. Such a behavior in KI is a result of two

426

Bang, Moon, Kim, Kang, Jung, and Oh

Fig. 12. Normalization of KIs in Fig. 11 by DX at various bump sizes.


A relation between R and DX can also be seen in Table III .

competing strain components at the crack tip. As


previously described and also shown in Fig. 9b, the
viscoplastic mass flow to the crack-tip area adds a
compressive strain component to the total deformation. This compressive strain competes with the
tensile strain from the bump rotation, working in
the direction of reducing KI. However, such a contribution becomes influential in small bumps in
which the strain from rotational moment is small
enough to be affected by the strain from the viscoplastic mass flow. Furthermore, the contribution is
not uniform but varies with shear distance. Such a
mechanism may be better understood with the
conceptual representation of the strain field displayed in Fig. 13. In this diagram, C+ and C- represent the degree of crack opening by the tensile
field and crack closing by the compressive field,
respectively. A simple approximation based on the
results shown in Fig. 12 finds that the crack opening by bump rotation may be expressed as:
C / DXDu

Fig. 13. Schematic illustration of crack-tip opening by rotation, C+,


and crack-tip closing by compressive deformation of solder bump, C-.
In the C+ diagram, the thick solid line provides the value for a large
solder bump, while the thin solid line illustrates that for a small bump.
This is also the case for the C- diagram. C+ + C-, which provides the
net crack-tip opening in degree, is displayed by the dashed (for large
bump size) and dotted lines (for small bump size).

(2)

where, Du is the shear distance. On the other hand,


the dependence of C- on the shear distance may be
approximated to be proportional to the contact arc
length between the bump and a shear probe because
it is related to the volume of the solder that is
plastically displaced during shearing:




1 R  Du
;
(3)
C /  R cos
R
where R is the bump radius, and the negative sign
represents crack closing. In the diagram, C- is displaced by Du in the shear direction because viscoplastic flow at small shear distances is localized to
the bump/probe contact area and cannot affect the
strain field at the crack tip of the bump corner (in
addition, Du will be larger for larger bumps since
the viscoplastic mass flow distance will be longer).
Then, C+ + C- determines the net rate of crack-tip

opening with shear distance. As shown in the diagram, for large bumps, C+ is predominant in determining the value of C+ + C-, making it increase
monotonically with shear distance. For small bumps
C+ is small enough to be affected by C-, and therefore the total opening of the crack (C+ + C-) shows a
sinusoidal variation with shear distance (dotted
line). Note the resemblance between the two cases of
C+ + C- with the K data for the 460 lm and 360 lm
bumps shown in Fig. 11. It is therefore reasonable
to conclude that the viscoplasticity of the solder
causes the bump to resist fracture, especially when
the bump is small.
DISCUSSION
It is our belief that the analysis presented in this
investigation provides several insights that are
useful in enhanced utilization of the two testing

Study of Fracture Mechanics in Testing Interfacial Fracture of Solder Joints

methodologies for solder joint evaluation. The most


obvious conclusion of this study is that both the
methods test the worst case of fracture resistance of
the solder joint. In the case of tensile testing, the
viscoplasticity of the solder and its restriction near
the IMC layer create the greatest level of stress
concentration at the IMC layer, making even
micron-size cracks fatal. The shear test configuration also puts the IMC layer under the strongest
possible fracture conditions because of the rapidly
developing KI value due to bump rotation. Since
these testing methods are often utilized to identify
problematic areas in joint reliability and also to
quantify the degree of reliability enhancement due
to processing or material design, both of these
methods can be highly successful in providing relevant information. Furthermore, the fact that fracture in the bump shear testing occurs by mode I
fracture may increase its usefulness. Some of the
greatest merits of the bump shear test are its simplicity in sample preparation and the proximity of
the test to real solder configuration, yet its application has conventionally been limited to the evaluation of adhesion strength. Since the bump shear
method shares similarity in fracture mechanics
with tensile test method, the method has a potential
for providing more information than just the adhesion strength, such as fracture toughness of the IMC
layer. Nevertheless, our study indicates that more
theoretical development is necessary to make bump
testing more useful. In particular, the size effect seen
in our study will pose significant challenges. As
pointed out by our results, the critical shear distance
(and therefore the fracture load) is sensitive to the
bump size even if the adhesion area is the same. Since
the mechanism responsible for the size effect is the
rotational moment, variation can also happen with
varying bump shape even if the starting solder ball is
the same. Depending on the contact angle or soldering conditions, the DX value that determines the
rotational moment can vary significantly. Because of
all these variations, bump size and shape will exist
within a sample population, the resulting fracture
data may be difficult to deconvolute. This complication created by the size effect needs further study to
resolve, part of which is ongoing in our laboratories.
The finding that the critical crack for IMC fracture is micron sized under tensile testing configuration may have significant practical importance
and deserve further attention. The very mechanism
that makes the critical size such an extreme scale is
the excessive crack-tip opening due to the nonuniform viscoplasticity of solder at the joint. This result
suggests that the viscoplasticity of solder, unlike
common expectation, may not be helpful in retarding joint fracture but, in fact, it may make joint
more prone to fracture. With the complexity of
stress conditions and also variation in the crack
initiation location in real solder joints, the threat of
IMC failure found in tensile tests may be masked. It
should be noted that, even if a crack initiates at the

427

solder, it becomes fatal once it encounters the IMC


layer. The usual IMC thickness in real solder joints
is in the micron range, i.e., it is in the same size
range as the critical crack. Therefore, once a crack
meets the IMC layer during its growth, it is likely to
be close to or larger than the critical size. In this
sense, engineering solutions for enhancing the
reliability of solder joints may need to focus more on
enhancing IMC properties.
Finally, the analysis shown in this paper is not
limited to the case of Pb-Sn eutectic solder. The
fundamental conclusion is equally valid for other
solder materials such as lead-free solder. All solders achieve adhesion with the substrate through
the reaction between Sn and metallization layers
such as Cu and Ni. While variations in material
properties may exist, such variations do not change
the fundamental fracture mechanics found in this
study. It is therefore reasonable to expect that a
simple extension of the current analysis would
yield valid assessments of the fracture mechanics
in joints made by other alloys as well as the
identification of the critical factors affecting their
reliability.
CONCLUSIONS
The analysis conducted in this study has focused
on fracture mechanics that is active in two common
solder joint reliability test configurations. With an
assumption of a predefined crack in the IMC layer,
the development of stress intensity factors at the
crack tip is numerically calculated with variations
in relevant contributing parameters such as crack
size, the viscoplasticity of the solder, and testing
conditions. The calculations assume the presence of
a crack in the IMC layer. Therefore, the fracture
mechanics calculated here appears to be valid only
when the interface fracture is considered. It is found
that the assumption of the IMC being the fracture
site correctly represents fracture in real testing
situations as well as in real solder joints. This is due
to the fact that both of the configurations produce
generic mechanical constraints that make the IMC
layer the most susceptible to fracture.
In the case of joint tensile testing, it is found that:
(a) the critical crack size for IMC fracture is on the
micron scale; (b) the brittleness of the IMC layer
stems from the generic mechanical constraints in
the test configuration that assist crack-tip opening;
and (c) with the brittleness of the IMC as well as the
restriction of the viscoplasticity of the solder near
the IMC, the fracture is insensitive to loading rate
in the range of practical importance. In the case of
bump shear testing, the IMC layer is again the
weakest link for fracture because: (a) the test load
does not impart pure shear load on the interface but
rather creates more of a crack-opening load because
of bump rotation; (b) the bump rotation occurs with
its axis located at the crack tip; and (c) because the
bump rotation (and thus KI level) is proportional to

428

Bang, Moon, Kim, Kang, Jung, and Oh

the rotational moment, the fracture is subjected to


the bump size effect.
ACKNOWLEDGEMENTS
The work at Seoul National University was
supported by Brain Korea 21, Korea. The work at
the University of Texas at Arlington was supported
by a grant from the Korea Research Foundation
(KRF-2005-000-10442).
REFERENCES
1. A.J. Sunwoo, J.W. Morris, and G.K. Lucey, Metall. Trans. A
23A, 1323 (1992).
2. B.J. Lee, N.M. Hwang, and H.M. Lee, Acta Mater. 45, 1867
(1997).
3. R.J. Fields, R.R. Low, and G.K. Lucey, The Metal Science of
Joining, ed. M.J. Cieslak, M.E. Glicksman, S. Kang, and
J. Perepezko (Ohio: TMS, 1991), p. 165.
4. B. Balakrisnan, C.C. Chum, M. Li, Z. Chen, and T. Cahyadi,
J. Electron. Mater. 32, 166 (2003).
5. L.K. Quan, D. Frear, D. Grivas, and J.W. Morris, J. Electron. Mater. 16, 203 (1987).
6. R.J. Asaro, N.P. ODowd, and C.F. Shih, Mater. Sci. Eng.
162A, 175 (1993).
7. H.-T. Lee and M.-H. Chen, Mater. Sci. Eng. A333, 24 (2002).

8. K.H. Prakash and T. Sritharan, Mater. Sci. Eng. 379, 277


(2004).
9. L.K. Quan (MS Thesis, U.C. Berkeley, CA, 1988).
10. JESD22-B117: BGA Bump Shear, Published by JEDEC
Solid State Technology Association (Virginia: JEDEC, 2000).
11. G.J. Irwin, Appl. Mech. 24, 361 (1957).
12. T.L. Anderson, Fracture Mechanics: Fundamentals and
Applications, 2nd ed. (London: CRC Press, 1994), pp. 2996.
13. J.-W. Kim and S.-B. Jung, Mater. Sci. Eng. A371, 267
(2004).
14. X. Huang, S.-W.R. Lee, C.C. Yan, and S. Hui (Proceedings of
51st Electronic Components & Technology Conference,
ECTC, Florida, 2001), p. 1065.
15. J. L. Marshall, The Mechanics of Solder Alloy Interconnects,
ed. D.R. Frear, S.N. Burchett, H.S. Morgan, and J.H. Lau
(New York: Van Nostraind Reinhold, 1994), p. 42.
16. E.W. Hare and R.G. Stang, J. Electron. Mater. 24, 1473
(1995).
17. Hibbit, Karlson & Sorensen: ABAQUS/Standard Users
Manual, Ver. 6.3 (2003), ch. 7.9.
18. ASTM E399-90: Standard Test Method for Plane Strain
Fracture Toughness of Metallic Materials (ASTM, Philadelphia), (1990), E1290-02.
19. D.R. Frear and P.T. Vianco, Metall. Mat. Trans. 25A, 1509
(1994).
20. D.B. Bogy, J. Appl. Mech. 38, 377 (1971).

You might also like