Reliability Assessment of Delamination in Chip-to-Chip Bonded MEMS

Packaging
Rajesh Swaminathana, Gowrishankar Subramaniana, Peter Sandborna, Michael Deedsb, Kevin Cochran b,
and Harish Bhaskarana
a-CALCE Electronic Products and Systems Center
University of Maryland
College Park, Maryland 20742
b-Indian Head Naval Surface Warfare Center
Indian Head, Maryland 20640

Abstract - The bond layer is often the weakest link in the reliability of chip packages in the IC
industry. MEMS packages are no exception to this trend. This paper presents a non-destructive
methodology for determining delamination in a chip-to-chip bonded MEMS. Experimental methods
are used to determine the bond strength in samples subjected to environmental testing and the
reliability of the bonds is investigated. A simulation is performed using inputs from scanning acoustic
microscopy and experimental die shear results are correlated with a simulation model to establish
the validity of the non-destructive methodology for determining bond strength.

Index Terms MEMS, reliability, packaging, delamination, chip-to-chip bonding, scanning

acoustic microscopy, die shear.
I. INTRODUCTION
Micro-Electro-Mechanical Systems (MEMS) have found widespread applications in a variety of fields
ranging from simple sensors to complex systems that require the integration of movable structures and
active circuitry. Among the challenges associated with transitioning MEMS from the laboratory to the
marketplace is packaging. Packaging encompasses the transition of an electronic or electromechanical
device from the die to the final product, so that the device is capable of functioning as part of a larger
system. It includes, but is not limited to, the design of interconnections at every level of the system, the
placement of the die on the carrier, removal of excessive heat generated in the case of integrated circuits
(ICs) and issues of reliability associated with the system as a whole [1]. A strong packaging background
exists for microelectronic devices, but research in MEMS packaging is still in its infancy.

MEMS are finding applications in products or systems that require reliable operation over extended
periods of time. The reliability requirements for the final product encompass both the mechanical behaviors
and the electrical characteristics of the overall system. One critical element in many MEMS applications is
chip-to-chip bonding (component bonding) [2], for which long-term operation and storage reliability needs
to be understood [3]. MEMS packages are likely to have a large number of bond layers because of multiple
interfaces inside the package. The bond layers in MEMS devices must often maintain precise chip
alignment in addition to withstanding loading from the macro-environment and loading within the package.
A primary indicator of failure (or impending failure) in a chip-to-chip bonded system is delamination
between the chip and the material used to bond the chips together [4]. In spite of its importance in MEMS
packaging, previous work on bonding in MEMS structures is limited [5]. Very little MEMS-specific work
on the reliability of the chip-to-chip bonds exists, let alone, non-destructive methods for determining the
reliability of chip-to-chip bonded MEMS.
The Use of Acoustic Micro Imaging (AMI) in non-destructive reliability assessment of microelectronic
devices is common. McCluskey et. al. [6] used AMI techniques to assess the delamination in plastic
encapsulated microcircuits. Mahajan and Pecht [7] also describe the use of a Scanning Acoustic
Microscope (SAM) for delamination assessment. AMI techniques have been used by Gunnamani and
Munnamarty to asses the cracks in solder joints in flip-chip assemblies [8]. In MEMS structures bonding
has been studied previously using destructive testing [9]. AMI has been used to assess failures in tunneling
accelerometers [10] but not applied to the assessment of bond reliability. Herenz et. al. proposed the use
of ultrasonic techniques using AMI to check bond quality [11].
The results of a delamination study using non-destructive techniques on chip-to-chip bonding in a
MEMS-based Safety and Arming (S&A) system is described in this paper. A brief description of the
system is provided, the test and metrology approaches are discussed and results from accelerated aging,
thermal cycling, mechanical shock and die shear validation are presented. In Section V a finite element
model is created and results from simulation of die shear validation are compared with the experimental
results in order to understand and verify the delamination measurement methodology.

II. A MEMS-BASED SAFETY AND ARMING SYSTEM

The application discussed in this paper is the Safety and Arming (S&A) device described in [12] and
[13] for use in undersea torpedoes in which two chips containing LIGA MEMS structures must be bonded
together and operate reliably after many years of storage in extreme environmental conditions. This
motivates the development of an experimental approach that can accelerate relevant failure mechanisms
and allow the evaluation of the systems reliability. As previously mentioned, the reliability concern in this
paper is the integrity of the bonds. Non-destructive methods are preferred to destructive testing in most IC
applications. Destructive testing often serves as a verification of the non-destructive testing techniques.
This paper describes a non-destructive approach to measure delamination of the bond layers.
The requirements of the S&A system are to safely and reliably arm and detonate the weapon, but only
after all safety criteria are met and the weapon has reached its intended target. The MEMS S&A system is
shown in Fig. 1.

Fig. 1. MEMS S&A system showing details of various components.

The MEMS components include the S&A chip, deflection delimiter and the initiator chip. The
operation and construction of the S&A system has been discussed in [12] and [13], and a picture of the
system can be found in [14].
The chip-level package includes the S&A, initiator chip, and delimiter. The initiator chip converts
electrical energy to mechanical energy upon demand to start the explosive train. The initiator chip is also
fabricated using MEMS technology. The S&A chip and the initiator chip must be reliably bonded together
while maintaining precise in-plane and out-of-plane alignment. Coefficients of thermal expansion
mismatch between the S&A chip and the initiator chip increases the challenge of maintaining alignment
and bond integrity.
A deflection delimiter is introduced to limit out-of-plane (z-axis) compliance of several structures.
The delimiter ensures that the locks on the barrier are not violated by z-axis displacement between
structures. The deflection delimiter must allow for in-plane movement of all structures, but prevent z-axis
movement of selected structures. In addition, the delimiter must allow for wire bonding and fiber optic
cable routing and mounting
The focus of the reliability analysis in this paper is the bonding of the initiator chip to the S&A chip,
with and without a deflection delimiter.

III. DELAMINATION MEASUREMENT METHODOLOGY

A non-destructive method to measure delamination was developed based on Acoustic Micro-Imaging
(AMI). A Scanning Acoustic Microscope (SAM) was used to perform AMI. The SAM utilizes highfrequency ultrasound to produce accurate high-resolution images of internal construction of objects. One of
the SAMs major advantages is its ability to penetrate optically opaque solids. Images are generated by
mechanically sweeping a sample while emitting ultrasound waves from a transducer. Images indicate the
delamination at the adhesive interface clearly. Although the use of AMI for studying the internal structures
of packaged chips and multi-chip modules is relatively common (see Section I), AMI has not been used to
analyze chip-to-chip bonded structures and thus the development of a new methodology using the SAM
was required for this work.
The structure of the test samples used in the study is shown in Fig. 2.

Ceramic or Silicon
(Initiator Chip)

Ceramic or Silicon
(Initiator Chip)

Silicon

Adhesive or
Solder

Spacer
(Deflection Delimiter)

(S&A Chip)

Silicon
(S&A Chip)

U shaped deflection delimiter

SAM Image

Direction of shear
Fig. 2. Test sample construction: The initiator and S&A chips are both 525 m thick. The deflection
delimiter (when present) is Alloy 42 and 125 m thick. The test samples are 13 mm x 16 mm.

Several combinations of bonding materials and test sample constructions were considered based on
design constraints such as operating and storage temperature, required life, processing and compatibility
issues, performance requirements and material design parameters. The selected materials are shown in
Table I.
Scanning Acoustic Microscopy utilizes high-frequency ultrasound to detect internal discontinuities in
materials and components. The SAM emits acoustic waves in a reflection mode (pulse-echo) at a given
frequency, typically ranging from 15 to 180 MHz. For a thin package such as the chip-to-chip bonded
system treated in this paper, a 110 MHz transducer has been used. The advantage of SAM lies in the ability
of the ultrasound to penetrate optically opaque solids. The distance between the echoes correlates to the
depth they travel in the device being analyzed. A transducer that alternatively acts as a receiver and
transmitter is used. The ultrasound wave uses an inert fluid (de-ionized water) as a coupling medium.
The information from the scan is obtained in the form of a Peak amplitude image and a Phase
inversion image. To obtain both the images, a front surface follower and a data gate need to be used. A
front surface follower compensates for any variation in the water path, which is the distance from
transducer to test specimen. The first signal to cross the Front Surface Follower (FSF) is considered the

front surface of the test specimen. In the analysis performed herein; this is the top of the chip. The data gate
is slaved to the crossing point (used as a reference). A data gate records the amplitude of the highest
peak within the signal being captured. The peak amplitude image will have the true peak amplitude data for
all data points collected during the scan, the amplitude being proportional to the amount of sound reflected.
A normal gray scale palette (white = 100% amplitude, black = 0% amplitude) is used. The data gate also
checks for signal phase and creates a phase inversion C-Scan. An algorithm is used to detect the phase
inversion for which the signal amplitude is required to be well below saturation. The algorithm assumes a
good indication if the first threshold crossing in the data gate is negative, i.e., a peak in the downward
direction and a delaminated indication if the first threshold crossing in the data gate is positive, i.e., a peak
in the upward direction.

Design
No.
1
2
3
4
5
6
7

TABLE I
BOND MATERIALS USED IN THE DELAMINATION STUDY
Bond
Bottom
Top
Spacer
Material
Substrate Substrate
(deflection
delimiter)
Thermoplastic
Silicon
Ceramic
Alloy 42
Paste
Thermoplastic
Silicon
Ceramic
Film
Epoxy Film
Silicon
Ceramic
Thermoplastic
Silicon
Silicon
Film
Epoxy Film
Silicon
Ceramic
Alloy 42
Thermoplastic
Silicon
Ceramic
Alloy 42
Film
Indium Solder
Silicon
Ceramic
Alloy 42

Bond
Thickness
(mm)
0.051
0.178
0.178
0.178
0.051
0.051
0.005

Before delamination can be measured using the SAM, the interfaces between the silicon and/or
ceramic and the bonding material must be accurately located. The first step in obtaining an accurate image
is to find the first reflected echo and fix the front surface follower onto the front surface signal. The data
gate is moved along the signal and fixed at the interface that will be scanned. To increase the amplitude of
the signal from the interface, the transducer is focused onto the interface. The focusing must be done
carefully, to insure that the focused signal does not overlap a static noise signal, which can produce
erroneous results.

If the interface is difficult to find, a multi-layered scan can be used to traverse through the depth of the
sample and capture the peak amplitude image for a large number of layers, each representing 5 s of water
path length. The data gate can be fixed at the required interface after careful analysis of each layer.
The size of the data gate had to be fixed at a required height and length in order to eliminate the noise
in the image. Using trial and error, a height of 75% the peak amplitude signal and a length equal to 10 s
of water path length was finally accepted to be the best data gate size. The validation of this gate height
was accomplished by scanning a large number of samples and correlating the peak amplitude image with
the phase inverted image.
Since a high-frequency transducer was used (115 MHz) the penetration capability is low and there is a
potential problem in obtaining the image of the ceramic interface from the silicon side. Hence the samples
were inverted and rescanned in order to verify the delamination on the ceramic side.
In the cases where the package has a spacer (deflection delimiter), the interfaces that were scanned are
the silicon interface with the adhesive, and the ceramic interface with the adhesive. There are potentially
two other interfaces in the package; they are the adhesive-spacer (from the silicon side), and the spaceradhesive interface (from the ceramic side). Scanning from the silicon side of the sample allows the siliconadhesive interface to be isolated, but the adhesive-spacer interface below it cannot be isolated since the
thickness of the adhesive is smaller than the minimum resolution of the SAM (50 m). Similarly, scanning
from the ceramic side of the sample allows the ceramic-adhesive interface to be isolated, but the adhesivespacer interface below it cannot be isolated.
Packages bonded with indium solder have a different signal compared to that of adhesive bonded
samples (the solders higher density causes the peaks in the reflected signals to be inverted relative to the
adhesively bonded cases). The inversion occurs because the density of the solder is greater than silicon and
ceramic. The size of the data gate was changed and validated in order to ensure that the information was
accurate.
In samples with spacers delamination was quantified by breaking the bonding layer into discreet
regions (Fig. 3).

Regions for quantifying

delamination

Fig. 3. The test sample deflection delimiter was divided into three regions to quantify delamination.
To verify the accuracy of the SAM-based methodology, grooved samples were fabricated using bulk
micro-machining techniques. The width of the grooves ranged from 0.05 mm to 0.25 mm and the depth
ranged from 0.007 mm to 0.035 mm respectively.
Two pieces of grooved samples were placed one on top of the other. This set was inspected using the
SAM, by placing both orientations of the sample (either side up) under the microscope. The peak amplitude
image of one set is shown in Fig. 4. While the grooves on the topside can be clearly seen, the grooves
below as appear as lighter streaks in the SAM. The same observation was made from both sides of the
samples. Using this calibration, the delamination measurements were determined to have an uncertainty of
5%.

Grooves on top

Grooves at the
bottom

Fig. 4. Peak amplitude images of two grooved samples placed one on top the other

IV. EXPERIMENTAL RESULTS

The bond layers are expected to be a crucial reliability concern in a chip-to-chip bonded system,
especially when subjected to harsh environments [3]. In this section, we discuss various environmental tests
employed to assess the reliability of these bond layers. The delamination of the bonded layers before and
after environmental testing is quantified using the AMI-based methodology described in the previous
section. The test plan shown in Fig. 5 was used in this study.

A. Accelerated Aging
In accelerated aging, the samples are maintained at higher temperature and humidity. Samples were
subjected to accelerated aging at 85 C and 85% relative humidity for 100 days. Some samples were
removed periodically for short periods of time to monitor bond performance. The delamination of the
samples was measured before and after accelerated testing. Fig. 6 shows a sample before and after 50 days
of accelerated aging. Table II tabulates example results from accelerated aging tests.

Packaging test structure

assemblies

No
Conditioning

Thermal
Cycling

Accelerated
Aging

Critical Failure

Mechanical
Shock
Critical Failure

Die Shear

Fig. 5. Chip-to-chip bonding reliability test plan

Failure
Analysis

Delaminated

Laminated

Delaminated

Fig. 6. Design 1, Sample 12; Initiator chip side (Thermoplastic film, with deflection delimiter). The
picture on the left is before accelerated aging; the picture on the right is after 50 days of aging.
B. Thermal Cycling
In the thermal cycling test, samples were subjected to a series of hot-and-cold temperature cycles. The
samples underwent 28 cycles of 54 C (8 hours) to 71 C (16 hours), no humidity, with 5 degree/minute
ramps (Fig. 7). Table III shows example results of delamination in samples subjected to thermal cycling

50C/ Minute ramps

16 hours

ture

710C

Time

-540C
8 hours

Fig. 7. Thermal cycling profile.

10

TABLE II
EXAMPLE ACCELERATED AGING RESULTS. DELAMINATION IS MEASURED IN PERCENTAGE OF BONDING AREA.
DELAMINATION UNCERTAINTY, 5%.
Design No. Sample Side exposed
Initial
25 day
50 day
100 day
No.
to SAM
Delamination Delamination Delamination Delamination
1
12
Initiator
43
43
65
100
12
S&A
23
31
17
20
13
Initiator
54
63
69
Not Applicable
13
S&A
77
76
100
100
2
3
Initiator
16
12
13
14
3
S&A
17
17
17
17
8
Initiator
26
27
28
39
8
S&A
30
29
30
29
3
1
Initiator
24
27
29
49
1
S&A
26
30
28
38
13
Initiator
39
35
34
48
13
S&A
31
35
32
35
4
5
Initiator
22
23
25
30
5
S&A
26
27
25
27
7
Initiator
18
20
18
20
7
S&A
21
19
18
19
5
8
Initiator
25
28
24
52
8
S&A
11
34
28
49
5
Initiator
45
81
26
37
5
S&A
14
29
15
19
6
5
Initiator
11
13
13
15
5
S&A
20
12
18
19
13
Initiator
22
20
19
20
13
S&A
17
14
9
14
7
1
Initiator
9
5
4
8
1
S&A
0
0
0
1
8
Initiator
46
48
41
47
8
S&A
10
7
10
11

C. Mechanical Shock
The purpose of mechanical shock for the S&A application considered in this paper is to simulate rapid
acceleration such as the torpedo being launched from a submarine. Mechanical shock could also be used to
simulate forces caused by rough handling, improper transportation, or abusive operation. All samples were
subjected to the mechanical shock profile shown in Fig. 8 after they completed their environmental
conditioning.

11

TABLE III
EXAMPLE THERMAL CYCLING RESULTS. DELAMINATION IS MEASURED IN PERCENTAGE OF BONDING AREA.
DELAMINATION UNCERTAINTY 5%.
Design Sample
Side
Initial
Delamination
No.
No. Exposed to Delamination
after
SAM
Thermal
Cycling
1
1
Initiator
77
69
1
S&A
64
73
8
Initiator
37
37
8
S&A
21
22
2
6
Initiator
32
33
6
S&A
16
16
14
Initiator
36
44
14
S&A
20
21
3
11
Initiator
16
17
11
S&A
17
19
14
Initiator
55
66
14
S&A
33
35
4
1
Initiator
15
28
1
S&A
16
40
13
Initiator
26
27
13
S&A
24
29
5
2
Initiator
81
87
2
S&A
25
39
3
Initiator
19
44
3
S&A
10
23
6
4
Initiator
6
8
4
S&A
4
5
8
Initiator
79
66
8
S&A
73
62
7
2
Initiator
6
18
2
S&A
1
0
12
Initiator
49
14
12
S&A
0
0

All test samples (using all bonding material combinations given in Table I) showed no additional
delamination when inspected under SAM after the mechanical shock.

12

300
250
200
150
100
50
0
0

10

15

20

25

30

35

40

45

50

55

60

65

70

-50
Time (ms)

Fig. 8. Mechanical shock profile.

D. Die Shear Testing
Once the environmental tests were completed, in order to verify the results it was necessary to
determine the strength of the bonds for varying degrees of delamination by assessing the effect of
delamination on bond strength through die-shear testing. In die-shear testing, a pure shear force is applied
to a bond until the bond yields. This force can either be tensile or compressive, theoretically the results
should be identical as long as care is taken in the loading. The chip packages in this testing were loaded in
a compressive manner.
The die-shear testing was performed using a fixture consisting of an aluminum frame that supports an
aluminum plate with a milled well. The well depth is such that when the chip package is placed in the well,
only the back of the bottom chip makes contact with the rear well wall. A Delrin pusher rides in slots in
the fixture frame sides making contact with the front of the top chip only, so that a shear load can be
applied by the pusher front surface. Delrin was used to minimize sliding friction between the pusher and
the slots.
The pusher also serves as a holder for a button type compression load cell that is used to monitor the
force applied to the chip package. Using nominal ultimate shear stress values from the bond material data

13

sheets, it was determined that the failure range for all of the packages in this test would range from 130 to
450 lbs. For this reason, a load cell with a 0 to 500 lb. range was chosen. The pusher and load cell are
pushed with a screw that is threaded into the front plate of the fixture frame (jackscrew configuration).
Friction between the end of the screw and the load cell is minimized with a thrust bearing. All chip
packages were sheared with the shear force acting parallel to the long (16 mm) axis of the chip as indicated
in Fig. 2.
The input voltage to the load cell was provided from a DC power supply and monitored with a
multimeter. A recording system consisting of a second multimeter was to record a voltage value for the
maximum shear stress. The output signal from the load cell was also monitored and recorded using a Data
Acquisition (DAQ) board and LabVIEW software for several reasons. First, since the loading on the
chips is a user-dependent operation (turning a screw), it was felt that a more constant loading profile could
be achieved if the loading could be monitored in real time. A constant loading profile is important because
several of the bond materials used have a low modulus of elasticity and creep that could affect the
maximum die-shear value seen. Second, it was desired to record the loading curve in case further analysis
was deemed necessary. A sample loading curve is shown in Fig. 9.
The chip-to-chip bonded samples were placed on the die shear fixture and the force was applied on the
samples until a drop in load cell voltage was obtained denoting a shear in the sample. The load cell output
tabulated for the all the samples is shown in Table IV. Using the known output range of the load cell (0 to
17.95 mV for a 10 V input), the corresponding shear force values are calculated. Ultimate shear stress
values are also calculated using the as-designed bond areas and the actual bond areas using, = F/A where
is the shear stress, F is the applied force and A the corresponding area (as-designed or actual bond area)
Actual bond areas were found by reducing the designed bond area by the delamination percentage at
the failure interface determined from the SAM analysis discussed in Section III. For packages that failed at
more than one interface, the delamination percentages at both interfaces were averaged in order to estimate
the actual bond area.

14

0.009
0.008
Max Shear = 6.5 mV

Load Cell Output (mV)

0.007
0.006
0.005
0.004
0.003
0.002
0.001
0
-0.001

10

12

14

16

18

20

22

Time (sec.)

Fig. 9. Sample die-shear testing data (multimeter reading at maximum shear = 6.5 mV)

Most of the adhesive packages containing an Alloy 42 spacer sheared cleanly into two pieces at failure.
The spacer remained attached to either the silicon or ceramic chip with all adhesive intact. Most of the
separated interfaces showed no signs of adhesive or only very light residue. A few packages remained
loosely intact with the spacer deforming at failure and remaining partially adhered to both the top and
bottom chip. The adhesive remained completely adhered to the spacer (on both sides) in all cases.
Most of the non-spacer packages did not completely fall apart at failure. The two chips would remain
partially attached by the adhesive, which remained adhered to both chips in portions of the original bond
area. A few of the packages cleanly sheared apart as described above. Packages that completely separated
before die-shear testing were reported as having zero die-shear strength.
Most of the packages where silicon breakage was reported at shear had an incorrect double U-shape
bond layer, i.e., no adhesive was present in the center of the package on the axis perpendicular to the shear
force direction. This could cause the chips to bow in the center during force application and eventually
break (instead of the adhesive bond).
Almost all indium solder packages failed in a unique way. At the maximum shear load, the indium
bond layer would release, but the package would not separate. It was then possible to push the chips back
and forth relative to one another (at a lower shear force level), without the package falling apart entirely.

15

The samples were also visually inspected to check the correlation with SAM images. In most of the
samples it was observed that there is a close correlation between the two, thus further validating the use of
SAM in devices of such sizes and performances. Fig. 10 shows two images, one corresponding to the SAM
image obtained before die-shear testing and one showing an optical image after the die-shear was
performed. Similar correlations were found with the other designs.

Fig. 10. Delamination on the ceramic side in a design without the spacer (Design No. 6, Sample No.13,
Thermoplastic film With Deflection delimiter). The figure on the left shows the SAM image and the one on
the right shows the optical image after die shearing.

The results of die shear were then tabulated to include information that allows correlation with the
amount of delamination after the environmental conditioning for each of the samples at the die-shear failure
interface, the amount of force needed to cause the shear during the die shearing experiment, and the
corresponding strength of the bond. To ensure validity of comparison, graphs of shear strength versus
percentage delamination are plotted for components that undergo the same conditioning and fail at the same
interface as this gives the strength of a particular material specific interface. Hence for example in Design 1
(thermoplastic Paste with deflection delimiter, described in Table I) the components that underwent thermal
cycling and failed at the silicon interface are plotted as a single series. The samples that underwent shock
and no shock and failed at the silicon interface are plotted on the same graph in a different series.1 A table
of sample results and the corresponding shear strength plot for Design 1 samples that underwent thermal
cycling and Mechanical shock is presented in Table IV and Fig. 11 respectively. The bond strength of
samples that went through accelerated aging was zero or close to zero and hence not plotted on the chart.

The samples that failed after accelerated aging are not plotted. In all other designs it is not appropriate to plot samples that
underwent thermal cycling and samples that underwent accelerated aging in the same series because the material properties of the
samples from these conditionings could be different. Also it is not appropriate to plot a sample that failed at the ceramic interface with
a sample that failed by the silicon interface on the same series.

16

Since mechanical shock had practically no effect on delamination the samples that underwent mechanical
shock were plotted on the same series as the samples that did not undergo mechanical shock.
TABLE IV
DIE-SHEAR RESULTS FOR DESIGN 1- THERMOPLASTIC PASTE WITH SPACER
Package
1-1
1-3
1-9
1-14
1-8
1-10
1-15
1-7
1-11
1-5
1-6
1-13
1-2
1-4
1-12
Expected
Mean
S. D.

Enviro.
Test
TC
TC
TC
TC
TC
S
S
S
NS
NS
NS
A
A
A
A

Failure Initial Delam. %Final Delam. %Actual Bond Area

2
in
Interface @ Failure Int. @ Failure Int.
Si
64
73
0.041
Si
30
30
0.107
Cer
52
52
0.073
Si
39
39
0.093
Si
21
22
0.119
Si
39
39
0.093
Cer
66
66
0.052
Si
25
25
0.115
Si
77
77
0.035
Si
48
48
0.080
Si
24
24
0.116
Si
77
100
0.000
Si
56
100
0.000
Si
20
100
0.000
Cer
43
100
0.000
0.153
60
0.062
29
0.045

Shear Load
mV
lbs
2.3
64.1
3.7
103.1
4.5
125.3
0.1
2.8
5.5
153.2
2
55.7
2.4
66.9
2.8
78.0
2.5
69.6
2.3
64.1
3.4
94.7
0
0.0
0
0.0
0
0.0
0
0.0
306
58
48

Ult. Shear Strength (psi)

Design Area Actual Area
418.7
1550.9
673.6
962.3
819.3
1706.8
18.2
29.8
1001.3
1283.8
364.1
596.9
436.9
1285.1
509.8
679.7
455.1
1978.9
418.7
805.3
619.0
814.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2000
2000
382
780
312
657

TC Samples that underwent Thermal Cycling; S - Samples that underwent Shocking; NS Samples that underwent No Shocking; A
Samples that underwent Accelerated Aging; Delam. Delamination; Failure Int. Failure Interface; Design Area - Area of the
adhesive as intended by the manufacture; Actual Area Un-delaminated area of the adhesive = (1-Delamination%)(Design Area);
Ultimate shear strength (for design area) = Force applied/Design area; Ultimate shear strength (for actual area) = Force
applied/Actual Area; Expected Shear strength as expected by the manufacturer; Mean Mean shear strength as predicted by the
experiments; S.D. Standard Deviation of the experimentally obtained shear strength values.
Package notation is as follows: for example 1-1 denotes sample No.1 of design 1 and 1-9 denotes sample no.9 of design 1.

17

2500.0

2000.0

Shock /No-Shock

1500.0

1000.0

Thermal Cycling

500.0

0.0
0

10

20

30

40

50

60

70

80

90

Fig. 11. Shear stress versus delamination of Design 1(Thermoplastic paste with deflection delimiter,
described in Table I and II) for the silicon side.
All Design 1 samples (with Alloy 42 spacer and thermoplastic paste as the adhesive) show 100%
delamination after accelerated aging. Even the Sample No. 4 of Design 1 (thermoplastic paste with
deflection delimiter) that started with 20% delamination (the least delaminated initially) was completely
delaminated at the end of aging. Hence none of the aged samples have been shown in the plots. Of the
remaining samples, the ones that separated at the silicon interface after thermal cycling are shown by the
black line and the ones that separated after shock at the silicon interface are shown by the gray line. The
plots show a mean interfacial stress value of 1200 psi for the silicon-adhesive interface after thermal
cycling and about 1250 psi for the silicon-adhesive interface for the unconditioned samples. But for the
latter we have neglected the extremely low value of 2.8 lbs force obtained for a 39% delaminated sample
(Sample No. 14 of Design 1, thermoplastic paste with deflection delimiter). This is valid because as is
evident in the plot (Fig. 11), this particular sample has one of its edges broken which could be responsible
for the sample to shear at such low forces. However Sample No. 11 of Design 1 (thermoplastic paste with
deflection delimiter) with 77% delamination has an extremely high bond integrity compared to other, which

18

is probably due to better wetting characteristics, better bond attachment, etc. The estimated push off
strength of the adhesive is 1000 psi [15]. Hence we know that the bond integrity is not significantly
affected by thermal cycling but is undermined by accelerated aging.

V. DIE SHEAR SIMULATION

The die shearing experiments summarized in the previous section resulted in measured relations
between various applied loads and the degree of delamination. This is a very useful set of data as it
indicates at what load levels these samples would fail, or in other words provides an indication of the bond
strength. In order to verify that the non-destructively measured delamination fractions used in the load
calculations presented in Section IV are correct, a finite element model (FEM) was developed to determine
the stress levels across the different layers and perform a comparative study with the experimental results
obtained. The development of a three-dimensional model was necessitated owing to the geometry of the
system. For this analysis the eight nodded iso-parametric hexahedral element has been used. A sample
three-dimensional meshed model developed is shown in Fig. 12.
The conditions of the die-shearing experiments were simulated using FEM. The mesh seed has been
appropriately chosen to get a good aspect ratio. Ideally a visco-plastic analysis should have been
performed. But considering the large number of nodes and the associated degrees of freedom for such a
complex geometry an elastic analysis was performed instead. Also the present effort is being run without
any information about the inelastic properties of the adhesive (Staystik 371) material. So this model
serves as only a guideline in estimating the range of the bonding strength and in validating the trends
obtained in the experimental results.2

Valid absolute values for the shear strength will not be obtained through this analysis.

19

Fig. 12. Meshed elements of the silicon-adhesive-Alloy42-adhesive structure (the topmost ceramic surface
is not shown in this figure).

Since the bottom of the silicon chip rests within a well, the z displacements are restrained (S&A
bottom chip in Fig. 12). The sideways motion of the sample was restrained. The side of the bottom chip
touching the rear wall is prevented from moving in the direction of the applied force. The force, specified
from the die shearing experiments, was applied to the samples on the initiator chip. The corresponding
stresses developed on the adhesive at the failure interface were studied.
Fig. 13 shows the stress distribution in a 30% delaminated model. The maximum stress in this case as
predicted by FEM is 6.17 Mpa
Fig. 14 shows the comparison of a FEM stress profile and the SAM image for one sample in Design 1.
Similar stress analysis was performed on the other samples of Design 1 (Thermoplastic paste with
deflection delimiter, described in Table II) that sheared off on the application of force. The samples that
were completely delaminated before undergoing any amount of die shearing were not analyzed. The
maximum stresses obtained in the remaining samples are plotted in Fig. 15. There is a scatter in the FEM
generated results as well due to the varying nature of the geometries in the delamination, but there is a

20

distinct correlation in the trends obtained between the experimentally determined and the finite element
modeled values as can be seen from Fig. 15.

Fig. 13. Force applied = 100 lbs Maximum Stress = 6.17 MPa.

Fig. 14. Comparison of the SAM Image (left) and the FEM model (right) of Sample No. 3 in Design 1 that
had a 30% unsymmetric delamination (the FEM model shows stress distribution on the failure adhesive
surface).

21

2500
FEM

Shear Strength (psi)

2000

model

1500

1000

500

Experimental

20

40
60
% delamination

80

100

Fig.15. Comparison of experimental and FEM predicted shear strength for samples in Design 1
(Thermoplastic paste with deflection delimeter described in Table I).
A similar analysis was carried out for Design 2 (Ceramic chip on silicon chip bonded without
delimiter using Thermoplastic Film, refer to Table I). Once again, all samples subjected to thermal cycling
were modeled and analyzed using Finite Element Method. The modeling in this case is more complicated
due to the highly irregular shape of the delaminated adhesive. A comparison of the SAM image and the
FEM stress distribution for Sample No. 3 of Design 2 is shown in Fig. 16. A three dimensional model is
used to model the geometry and the constraints are applied as described for the earlier model (Design 1
models). The results of the analysis are compared to the experimental results and the chart of shear strength
against percentage delamination is shown in Fig. 17.

22

Fig.17 Comparison of the SAM Image (left) and the FEM model (right) of Sample No. 1 in Design 2
(thermoplastic film without deflection delimiter) that had a 23% unsymmetric delamination

2500

FEM
Shear Strength (psi)

2000

1500

1000

500

Experimental
0
19

21

23

24

% delamination

Fig.17. Comparison of experimental and FEM predicted results for Design 2 (Thermoplastic film without
spacer, refer to table II for description).

23

VI. DISCUSSION
A methodology for non-destructive testing of delamination has been established. The accelerated
testing and thermal cycling results show some influence on delamination, however the influence of
mechanical shock on delamination has been shown to be negligible. The die shearing experiments resulted
in measured relations between various applied loads and the degree of sample delamination indicating the
bond strength.
The suitability of indium solder for use in high reliability systems has been established from these
experiments.
Finite Element modeling carried out as described in Section V established the validity of the
delamination measurements carried out using the scanning acoustic microscope. Further, the close
correlation between the experimental and Finite Element Analysis predicted shear strengths indicates the
possibility of the use of finite element analysis techniques to predict bond strength.
However computing limitations have required that we use an elastic analysis when ideally a viscoplastic analysis should be done. Future work should concentrate on a visco-plastic model to predict bond
strength. For this the visco-plastic properties of the materials must be investigated, since such data is
currently unavailable. A study of the change in material properties after accelerated aging and thermal
cycling is necessary for accurate prediction of shear strength values using Finite Element Analysis. Further,
the interfacial properties of various adhesives with substrates (such as Adhesive with Silicon or Adhesive
with Ceramic) must be studied in order for the results of simulation to be comparable to the experimental
results.

ACKNOWLEDGEMENTS
The authors wish to acknowledge the support of the Naval Surface Warfare Center at Indian Head
Maryland and the membership of the CALCE Electronic Products and Systems Center at the University of
Maryland.

24

REFERENCES
[1] J. W. Dally, Packaging of Electronic Systems: A Mechanical Engineering Approach, New York:
McGrawHill, 1990.
[2] T.-R. Hsu, Packaging design of microsystems and meso-scale devices, IEEE Transactions on
Advanced Packaging, Vol. 23, No.4, pp. 596-601, November 2000.
[3] S. E. Vargo, A. A. Green, J. Mueller, D. P. Barne, and R. H. Reinecke, Characterization of KovarPyrex anodically bonded samples a new packaging approach for MEMS devices, MEMS Reliability for
Critical Applications, Russel A. Lawton, Editor, Proceedings of SPIE, Vol. 4180 pp. 76-83, 2000.
[4] L. Lin, MEMS post-packaging by localized heating and bonding, IEEE Transactions on Advanced
Packaging, Vol. 23, No.4, pp. 608-616 , November 2000.
[5] A. Malshe, W. Brown, W. Eaton, and W. Miller, Challenges in the packaging of MEMS,
International Journal of Microcircuits and Electronic Packaging, Vol. 22 No. 3, pp. 233-241, Third
Quarter 1999.
[6] P. McCluskey, F. Lilie, O. Beysser, and A. Galto, Low temperature delamination of plastic
encapsulated microcircuits, Microelectronics Reliability, Vol. 38, pp. 1829-1834, August 1998.
[7] R. Mahajan and M. Pecht, Reliability assessment of a plastic encapsulated RF switching device,
Microelectronics Reliability, Vol. 38, pp 1607-1610, December 1997.
[8] R. Gannamani and R. Munamarty, A crack-qualification procedure for SMDs, SMT, Vol. 10, pp. 5254, April 1996.
[9] G. Klink, B. Hillerich, Wafer bonding with an adhesive coating, SPIE Conference on Micromachined
Devices and Components IV, Vol. 3514, pp. 50-61, September 1998.
[10] J. D. Olivias, S. Bolin and E. J. Chern, NDE of microelectromechanical systems, EEE Links Vol. 3
No. 1 WWW Site: http://nepp.nasa.gov/imd/eee_links/vol_03/no_01/eee3-1f.htm

25

[11] A. Herenz, K-J. Wolter, E. Meusel, B. Hirschfeld, H. Kraytz and R. Ronniger, C-SAM evaluation of
large bonded areas in the field of micromechanics and microelectronics, International Acoustic Micro
Imaging Symposium, September 1999
[12] L. Fan, H. Last, R. Wood, B. Dudley, C. Khan, C. K. Malek, and Z. Ling, SLIGA based underwater
safety and arming system, Microsystem Technologies vol. 4, pp 168-171, September 1998.
[13] M. Deeds, K. Cochran, P. Sandborn, and R. Swaminathan, Packaging of a MEMS based safety and
arming device, Proc. 2000 Inter Society Conference on Thermal Phenomena (ITHERM), pp 107-112, May
2000.
[14] Cover Photographs, IEEE Transactions on Components and Packaging Technologies, Vol. 22, No.2,
June 1999.
[15] Product data sheet for Thermoplastic Film (TP 7090), Rev B@ 03/15/99, Device push-off strength, AI
Technology, 70 Washington Road, Princeton Jct. NJ 08550
.

26