Development of Multi Chip Modules for Extreme Environments
by
Hyun Joong Lee
A thesis submitted to the Graduate Faculty of
Auburn University
in partial fulfillment of the
requirements for the Degree of
Master of Science
Auburn, Alabama
August 9, 2010
Keywords: Multi Chip Module, Flip Chip, Wire Bond, Thermal Cycle Test
Copyright 2010 by Hyun Joong Lee
Approved by
Bogdan M. Wilamowski, Chair, Professor of Electrical and Computer Engineering
Hulya Kirkici, Associate Professor of Electrical and Computer Engineering
Robert N. Dean, Assistant Professor of Electrical and Computer Engineering
Abstract
Multi chip module technology offers numerous advantages such as reduced signal delays,
higher performance, lower power consumption, smaller space, its simplification. In this paper, the
fabrication process of multi chip module for extreme environments will be presented including 2
IC interconnection techniques, flip chip and wire bond. The chips has been subjected to extreme
thermal cycles, from -180 C to 125 C, for a reliability test. Resistance measurements were also
performed at room temperature after thermal shock cycles. Finally, conclusions are made from the
results.
ii
Acknowledgments
I would like to express my appreciation and sincere thanks to my research advisor, Dr. Bogdan
M. Wilamowski for the academic guidance, inspiration and encouragement through my educational
career. I would also like to thank the members of my committee, Dr. Hulya Kirkici and Dr. Robert
Dean, for their advice on this work. I would like to thank my lab director, Charles Ellis, for his
support.
Finally, I would also like to thank my parents for their support toward my education.
iii
Table of Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Potential Reliability Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3 Two Possible Solutions for MCM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1 Filp Chip Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1.1 Coating of Polyimide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1.2 Pattern the Seed Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.1.3 Plating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.1.4 Coating of Polyimide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.5 Photoresist Layer 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.6 Polyimide Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1.7 Backside Metal Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Wire Bond Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.1 Seed Layer Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.2 Pattern the Seed Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2.3 Plating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2.4 Coating of Polyimide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2.5 Etch Openings in Polyimide . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2.6 Polyimide Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2.7 Photoresist Layer 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
iv
3.2.8 Backside Metal Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4 Selected Optimum Solutions for Multi Chip Module Technology . . . . . . . . . . . . 17
4.1 MCM Process Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.1.1 Seed Layer Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1.2 Ground plane Patterning . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1.3 Plating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1.4 BCB Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.1.5 2nd Layer BCB Application . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.1.6 Descum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.7 Copper Brightening(oxidized during descum) . . . . . . . . . . . . . . . . 22
4.1.8 Seed Layer Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.1.9 Layer 2 Patterning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.1.10 Plating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.1.11 Polyimide Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1.12 Photoresist Patterning for Polyimide Etching . . . . . . . . . . . . . . . . 23
4.1.13 Polyimide Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.14 Seed Layer Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.15 Photoresist Patterning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.16 Plating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.1.17 Polyimide Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.1.18 Photoresist Patterning for Polyimide Etching . . . . . . . . . . . . . . . . 26
4.1.19 Polyimide Etching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.1.20 Backside Metal Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.1.21 Dice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Test Chip Procedures for MCM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.2.1 Patterning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
v
4.2.2 Plating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5 Test & Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.1 Interaction between Die to Substrate Attach . . . . . . . . . . . . . . . . . . . . . 32
5.1.1 Wire Bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.1.2 Flip Chip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2 Interaction between Substrate to Package Attach . . . . . . . . . . . . . . . . . . . 35
5.3 Test Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.4 MCM Package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
vi
List of Figures
3.1 Flip chip structure with underfill . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Flip chip mask1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3 Cu, Ni, Au electroplating setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.4 Flip chip mask2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.5 Die shear strength with backside metal deposition . . . . . . . . . . . . . . . . . . . . 12
3.6 Wire bond mask1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.7 Wire bond mask2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.1 MCM process flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 MCM Layout 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.3 MCM Layout 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4 MCM Layout 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.5 MCM Layout 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.6 MCM Layout 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.7 MCM Layout 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.8 Test chip layout for MCM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.1 Wire bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2 Strain distribution in the whole wire bond die attach type package[14] . . . . . . . . . 33
5.3 Flipchip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.4 Strain distribution in the whole flip chip type package[14] . . . . . . . . . . . . . . . 34
5.5 FEM model of substrate to package[14] . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.6 Wire bond test vehicle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
vii
5.7 Flip chip test vehicle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.8 MCM before sealed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.9 MCM after sealed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.10 PAD-LEAD connection-1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.11 PAD-LEAD connection-2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.12 Temperature cycling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.13 Resistance changes after 50 thermal shock cycles . . . . . . . . . . . . . . . . . . . . 41
viii
List of Tables
3.1 YES Oven Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2 Plating Mixing Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3 Cr Etching Mixing Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.1 Comparison of BCB with Polyimide . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 YES Oven Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
ix
Chapter 1
Introduction
The extreme environmental conditions on the surfaces of the Moon and Mars are excep-
tionally harsh with the worst case from -180 C to 125 C. These extremes need thermally pro-
tective electronic warm boxes to protect the electronics. This leads to increased launch weight,
system complexity, lack of modularity, degradation in system performance and reliability. Now,
if a System-In-Package operates correctly in an earth-like climate in space, it brings about huge
advantages. This paper focuses on a multi chip module process; developments including flip chip
and wire bond IC interconnection techniques for extreme environments.
A single chip package was commonly used in a variety of applications in the past, but MCM
(Multi Chip Module) technology is widely accepted now. MCM is a hybrid manufacturing tech-
nique for multiple IC (Integrated Circuits) chips as a single IC. The basic concept of MCM is to
reduce the space, decrease the interconnection-delays between the paths on the different chips.
Therefore, it has higher performance results from reduced signal delays between chips, improved
signal quality, lower power supply needed because of shorter interconnect lengths, reduced weight,
much higher functionality in a smaller space, and simplification of complexity and reduced number
of external components.
There are three types of MCMs; MCM-C, MCM-D, and MCM-L. It is based on substrate
selection. An MCM-C uses a ceramic substrate. MCM-D technology uses deposited dielectric [1].
There are two types of IC interconnection solutions, flip chip and wire bond. The choice be-
tween flip chip and wire bond is determined by production cost, packaged device performance, and
overall size. Flip chip is called Direct Chip Attach (DCA), because the chip is directly attached to
the substrate. This means the chip is assembled face down onto the board without wire bonding.
1
Even though flip chip is not a new technology, it is becoming popular because it has several ad-
vantages. Since IBM introduced flip chip technology in the 1960s for their mainframe computers,
it has been widely used in various engineering fields [2]. Today it is applied in most electronic
watches and increasingly in cellular phones, pagers, LCD and high speed microprocessors. This
technology gives several advantages including smaller size, higher performance, lower cost, and
flexibility. However, it has also several disadvantages as well; such as difficulty in the testing of
bare dies, limited availability of bumped dies, fine pitch substrates, limited repair possibility, and
the difficult handling of bare chips. Wire bonding is another semiconductor device fabrication
method to make electrical interconnects. The difference with Flip chip is that each bond is con-
nected individually. The biggest advantages of wire bonding are its process flexibility and the sheer
quantity of wire bonders. As a result, it is a mature technology in which the production process
is thoroughly researched. In addition, wire bonding is very flexible compared to flip chip. If the
die size changes, it can be accommodated without additional costs as well as tighter control of
wire length. Wire bonding also has several disadvantages which are its larger size, higher cost and
increase of inductance and capacitance of the connection.
The goal of this thesis is to develop Multi-Chip-Module technology with high reliability in
extreme environments in both high and low temperature. A spacecraft flying in space is faced
with harsh surroundings. Along with temperature, the atmosphere is totally different from the
earth. Therefore, special requirements are necessary for chip interconnections used in packaging
of semiconductor devices. The process is based on thin copper/polyimide substrate technology to
build electronics capable of operating in extreme environments, without the need for warming or
cooling, to maintain an earth-like environment. First, it provides a background of flip chip, wire
bond and a multi chip module with their basic architectures. In Chapter 3, the processes of flip
chip and wire bond are presented which is satisfied in these extreme conditions. In Chapter 4 the
process of MCM is presented which is satisfied in these extreme conditions. Then several tests are
implemented with extremely high temperature and low temperature in chapter 5. It demonstrates
the results as well as a discussion. Finally, it comes to a conclusion about the research.
2
Chapter 2
Potential Reliability Issues
A Thermo-mechanical reliability issue of chip packaging is a major concern when the assem-
blies are exposed to harsh environments like space. There are several main potential problems such
as the coefficient of thermal expansion (CTE), thermal stress and cracks.
The coefficient of thermal expansion (CTE) is a very basic physical property which has con-
siderable importance in mechanical and structural design applications of a material. CTE is de-
fined as relative increase in length per unit temperature rise. The stored energy typically expands
in response to temperature change [3]. This dimensional response is called coefficient of thermal
expansion (CTE). As a component of electronic packages is comprised of dissimilar materials, the
operation reliability issues are often present due to the CTE mismatch between these materials.
If the CTEs of the substrate and the packages are different, they will expand or contract as the
temperature changes. This CTE mismatch causes serious problems. It will stress the joint thus,
the solder joint may be fractured. In addition, mechanical stresses usually happen at every stage of
fabrication with a CTE mismatch. These stresses may cause degradation of device characteristics,
reliability of the assembled components, and the power dissipation of the circuit. Therefore, it is
key to match CTE between different layers.
One severe limitation on the reliability issues of present electronics is the stress caused by
CTE. Typical IC packages are made by a variety of materials which have different CTEs. If one of
them has unique CTE compared with the other materials, it leads to thermal stresses. It is recom-
mended to solve stresses at an early stage of semiconductor manufacturing. As a semiconductor
wafer undergoes fabrication processes, it is stressed due to the coefficient of thermal expansion
mismatches between the silicon substrate and the deposited thin film layers. Therefore, it may
3
cause the severe problems; quality, reliability of the assembled components, and deformation be-
tween the chip and substrate[4]. Such stresses can be increased due to several reasons such as
CTE mismatches, geometrical discontinuities, cyclical and random thermal loadings, and handling
during assemblies operations including wafer preparation, oxidation, diffusion, metallization, die
and wire bonding, encapsulation, and curing [5]. The amount of stress is normally proportional to
the CTE mismatch and the polymer film thickness. The polymer tends to be delaminated from the
substrate with a growing stress [6].
The occurrence of microcracks is also a potential reliability problem because the cracks are
normally time dependent, which means they grow in response to thermal cycling. In most cases,
microcracks which are not quickly detectable can also occur between, or within, individual grains
of brittle materials. The microcracks are usually observed at the edges of the chip and then prop-
agate into the entire chip [7]. Optimization of the fabrication process could have an effect on
minimizing it. Several guide lines have been proposed to avoid high stress such as low CTE with
low CTE mismatch, optimized curing processes, and prevention of moisture absorption [8], [9],
[10]. Package cracking, wire cracking, thin film cracking on the die, and thermal shock cycle tests
are also associated issues that need to be considered [11], [12].
The substrate plays a crucial role in ensuring the electrical, thermal, and mechanical reliability
of electronic assemblies. It is quite important to have an intimate CTE between die to substrate and
substrate to package. Otherwise, it incurs huge stresses on the edge of the corner due to the CTE
mismatch. If the CTE of the test die and package increase with decreasing temperature, substrate
should increase with decreasing temperature and then at last it is condiderable. Thermally induced
strains lead to a severe problem in their reliability under these conditions. Thus, material selection
can be critical. Ceramic substrate was used for wire bond and silicon substrate was used for flip
chip.
4
Chapter 3
Two Possible Solutions for MCM
Two possible interconnect solutions for MCM will be discussed with respect to mechanical
and electrical performance. Because of CTE mismatch, thermal strains, stresses, and cracks will
occur at the both chip to substrate interconnection and substrate to package interconnection. The
initial concept is to match the coefficient of thermal expansion (CTE) of the package. Both Si and
AlN substrates were used for flip chip and wire bond. The temperature range used to study the
stress in flip chip and wire bond from CTE mismatches was from -180 C to 125 C.
Devices operating under high power conditions may be exposed to high voltages, and so
high voltage breakdown may occur between adjacent electrical connections. To prevent this, an
effective dielectric passivation layer must be applied to ensure reliable package operation [13].
Introduced interconnect substrate technology was made of thin film copper and polyimide (HD
Microsystems, PI-2611 with CTE of 3ppm/ C). Polyimides are the most commonly used mate-
rial for flexible, high-density interconnection circuit applications. Polyimide films are amber in
color and highly transparent, being formed by the condensation reactions of aromatic dianhydrides
with aliphatic or aromatic diamines. They belong to the family of high temperature thermosetting
polymers originally developed by the DuPont Company, and have been mass produced since 1955.
This stiff structure exhibits an exceptional combination of high thermal stability, a very low CTE,
high heat resistance, excellent dielectric properties (2.9 of dielectric constant), and relatively low
moisture uptake (0.5%) compared to most other polymeric materials.
3.1 Filp Chip Solution
An underfill encapsulation, shown in figure 3.1, is normally used for the flip chip die assemble
process to minimize the solder joint strains and to improve the reliability. A polymer material is
5
usually used to fill in the space between the die and the substrate. Epoxies are the primary material
used for an underfill because it has great storage stability, long-established safety, wide availability,
and high versatility. However, an underfill with epoxy was not applied for this flip chip process
because epoxy becomes brittle at extremely low temperature, while indium remains malleable at
the same temperature[15]. An indium was used for substrate attachment to package and solder
bump was used for flip chip die to substrate attachment. Silicon substrate was chosen for flip chip
because it is well matched to Si die without underfill.
Figure 3.1: Flip chip structure with underfill
Flip chip was fabricated on 4-inch diameter Si substrates. The Si substrate provides a perfect
CTE match to the Si flip chip die. An oxide insulating layer of thickness 6000 ?A was grown on the
Si substrate in an oxidation furnace. The oxidation was performed in the presence of oxygen and
hydrogen gases at 1050 C.
3.1.1 Coating of Polyimide
The first step is to make a coating of polyimide on the surface. Prior to the application of
polyimide, the silicon wafer was placed in a dehydration bake oven at 120 C for 30 minutes. For
non-polyimide surfaces - a thin layer of adhesion promoter was applied by spinning VM652 (HD
Microsystems) onto the surface at 5000 rpm for 30 seconds to improve the adhesion [16], and to
avoid silicon delamination or peel off. The layer was soft-baked at 120 C for 60 seconds. And
then polyimide PI2611 was spun on the surface at 500 rpm for 5 seconds with an acceleration
speed of 250 rpm, then spun at 3000 rpm for 30 seconds with an acceleration speed of 1000 rpm.
6
It was then soft-baked at 120 C for 5 minutes in the dehydration bake oven. The next step was a
?low stress? curing process. The curing process converts the polyamic acid into a fully aromatic
with high temperatures to completely imidize the film [17]. It can be done in the programmable
YES oven (model 450-PB8-2P-CP). After the wafer was placed in a programmable nitrogen oven
at ambient temperature, the process began by purging the oven with N2. Table 3.1 is the process
used to cure the polyimide layer under low stress using YES oven.
After the curing process, E-beam process was applied to deposit thin layers of metals to form
the electrodes, conducting traces, and connection pads. It was selected with Ion Clean (5 minutes),
Chrome (250 ?A), and Copper (2000 ?A) as the seed layer deposition.
Table 3.1: YES Oven Procedure
STEP PROCESS
Step1 Target Temperature: 90 Ramp rate: 1/Min Curing time:30 min
Step2 Target Temperature: 150 Ramp rate: 1/Min Curing time:30 min
Step3 Target Temperature: 350 Ramp rate: 2/Min Curing time:60 min
Step4 The nitrogen oven is then ramped down to ambient temperature at a rate of 2 C/min
3.1.2 Pattern the Seed Layer
The second step is to make the seed layer pattern. The silicon wafer was placed in the dehy-
dration bake oven at 120 C for 20 minutes ahead of the application of photoresist. HMDS was then
evaporated on the surface of the silicon wafer for 5 minutes to improve the photoresist adhesion. A
layer of thick photoresist, AZ9245 (AZ Electronic Materials), was used for patterning. It was spun
on the surface at a speed of 1700 rpm for 5 seconds with an acceleration speed of 500 rpm and
spun at 2700 rpm for 30 seconds with an acceleration speed of 1000 rpm. It was then soft-baked
7
at 110 C for 90 seconds. After that, the silicon wafer was exposed (Mask Alligner MA6/BA6) to
make a pattern with 1 cycle for 60 seconds. Exposure was performed immediately after soft bake,
as soon as the wafer had cooled to room temperature. It was then developed with AZ400K positive
photoresist developer for 5 minutes which was in the ratio of 1:2.5 with deionized water. Then,
a plasma descum with Matrix (300W, 20sec) was performed to remove any residual photoresist.
After completing this step, photoresist thickness was about 4.5 um. The mask layout is shown in
figure 3.2.
Figure 3.2: Flip chip mask1
3.1.3 Plating
The next step is metallization. The electroplating setup is shown in Figure 3.3. First, copper
(2.5um, 12mA) was plated. Nickel was then plated (0.5um, 10mA, 35-40 C). Finally, Au was
plated (0.5um, 6mA, 35-40 C). The Nickel layer served as a thermal barrier; It was used to im-
prove the adhesion to the resistor material. As a top layer, Au usually protects it from oxidation and
corrosion to the Ni layer [17]. Plating was very slowly done in order to obtain a smooth, even coat-
ing of the plated metal. During the process, it should have no interval, especially between nickel
plating and gold plating. It was observed to be contaminated if interval existed between nickel and
Au plating. As a result, it was not evenly plated and thus, it was peeling off. To avoid unexpected
failure, orostrike C RTU was suggested. After finishing nickel plating, it gave the wafer a short
8
soak in orostrike C RTU solution for 30 seconds and then plated Au immediately using 434 HS
RTU Gold without water cleaning. After plating, positive photoresist was stripped by acetone and
was cleaned by methanol and water. Then the seed layers were etched by Cu etchant, Potassium
Permanganate.
Table 3.2: Plating Mixing Formula
SOLUTION QUANTITY
Copper Microfab SC Makeup 1L
Microfab SC MD 8mL/L
Microfab 10 70/30 2mL/L
Nickel Nickel Sulfamate RTV 1L
Nickel Stress Reducer 8mL/L
Gold RTV HS434 - Straight
Polyimide adhesion Prepoly is done to coat polyimide on polyimide
Adhesion promoter for polyimide on oxide
Table 3.3: Cr Etching Mixing Formula
SOLUTION
80 gm KmnO4 (Potassium Permanganate)
20 gm NaOH (Sodium Hydroxide)
1 L H2O
9
Figure 3.3: Cu, Ni, Au electroplating setup
3.1.4 Coating of Polyimide
The following step is to make the second coating of polyimide on the surface. Prior to the
application of polyimide, the silicon wafer was placed in a dehydration bake oven at 120 C for
30 minutes. Openings were then etched by AOE Nitrogen for 30 seconds. After that, a thin layer
of adhesion promoter was formed by spinning VM652 (HD Microsystems) onto the surface at
3000 rpm for 30 seconds. Next, the layer was soft-baked at 120 C for 60 seconds. And then
polyimide PI2611 was spun on the surface at 500 rpm for 5 seconds with an acceleration speed
of 250 rpm then, spun at 3000 rpm for 30 seconds with an acceleration speed of 1000 rpm. It
was then soft-baked at 120 C for 5 minutes in the dehydration bake oven. Then, it was placed in
a programmable nitrogen YES oven (model 450-PB8-2P-CP) at ambient temperature for a ?low
stress? curing process. Table 3.1 shows the curing process under low stress using YES oven.
3.1.5 Photoresist Layer 1
The silicon wafer was placed in the dehydration bake oven at 120 C for 20 minutes ahead
of the application of photoresist. HMDS was then evaporated on the surface of the silicon wafer
for 5 minutes to improve the photoresist adhesion. Positive photoresist AZ4620 was then spun on
10
the surface at a speed of 1000 rpm for 30 seconds with an acceleration speed of 250 rpm and soft-
baked at 110 C for 90 seconds. After that, it was immediately exposed (Mask Alligner MA6/BA6)
to make a pattern with 3 cycles for 90 seconds as soon as the wafer had cooled down to room
temperature. It was then developed with AZ400K positive photoresist developer for 4 minutes
which was in the ratio of 1:2.5 with deionized water. Next, a plasma descum with Matrix (300W,
20sec) was performed to remove any residual photoresist. After completing this step, photoresist
thickness is about 20 um. The mask layout is shown in figure 3.4.
Figure 3.4: Flip chip mask2
3.1.6 Polyimide Etching
The last step is polyimide etching using A.O.E. The etching process was P-imide2 which
contains ICP-2min, Polyetch1-3min, Polyetch2-4min.
3.1.7 Backside Metal Deposition
It began with E-beam deposition process. It was selected with Titanium (500 ?A) and Nickel
(800 ?A) Gold (1200 ?A) as backside metal deposition [14]. Average shear strength with Ti/Ni/Au
backside metal deposition is shown in figure 3.5. Then, the silicon wafer is diced to separate all
the individual probe arrays.
11
Figure 3.5: Die shear strength with backside metal deposition
3.2 Wire Bond Solution
Ceramic substrates (AlN) were selected for wire bond solution because they provide an inter-
mediate CTE match between the die (si) and the package (Al203). Indium was used for Die and
Substrate Attach. Indium is known that it deforms viscoplastically with over than 300 C operating
temperature range [15]. The attached die were thermosonically wire bonded using an automatic
wire bonder and 25.4 m diameter Au wire. The stage temperature was 125 C. It is also used for
substrate to package electrical interconnection.
An oxide insulating layer of thickness 6000 ?A was grown on the ceramic substrate in an ox-
idation furnace. The oxidation was performed in the presence of oxygen and hydrogen gases at
1050 C . The general concept of ceramic substrates is that they have the same CTE with ceramic
package. That is, they do not face any problems related to any CTE mismatch. So, ceramic sub-
strates are widely used in hybrid applications.
3.2.1 Seed Layer Deposition
Wire bond process began with E-beam deposition process. It was selected with Titanium
(250 ?A) and Copper (2000 ?A) as the seed layer deposition
12
3.2.2 Pattern the Seed Layer
The second step is to make the seed layer pattern. The silicon wafer is placed in the dehy-
dration bake oven at 120 C for 20 minutes before the application of photoresist. HMDS was then
evaporated on the surface of the silicon wafer for 5 minutes to improve the photoresist adhesion. A
layer of thick photoresist, AZ9245 (AZ Electronic Materials), was used for patterning. It was spun
on the surface at a speed of 1700 rpm for 5 seconds with an acceleration speed of 500 rpm and
spun at 3000 rpm for 30 seconds with an acceleration speed of 1000 rpm. It was then soft-baked
at 110 C for 90 seconds. After that, the silicon wafer was exposed (Mask Alligner MA6/BA6) to
make a pattern with 1 cycle for 60 seconds. Exposure was performed immediately after soft bake,
as soon as the wafer had cooled to room temperature. It was then developed with AZ400K positive
photoresist developer for 5 minutes which was in the ratio of 1:2.5 with deionized water. Then, a
plasma descum with Matrix (300W, 20sec) was performed to remove any residual photoresist. Af-
ter completing this step, photoresist thickness should be about 4.5 um. The mask layout is shown
in figure 3.6.
Figure 3.6: Wire bond mask1
13
3.2.3 Plating
The next step is metallization. First, copper (1.25um, 12mA) was plated. Nickel was then
plated (1.0um, 10mA, 35-40 C). Then, Au was plated (1.5um, 6mA, 35-40 C). The Nickel layer
served as a thermal barrier; It was used to improve the adhesion to the resistor material. After
plating, positive photoresist was stripped by acetone and was cleaned by methanol and water.
Then the seed layers were etched by Cu etchant, Potassium Permanganate.
3.2.4 Coating of Polyimide
This step is to create a coating of polyimide on the surface. Prior to the application of poly-
imide, the silicon wafer was placed in a dehydration bake oven at 120 C for 30 minutes. For
non-polyimide surfaces - a thin layer of adhesion promoter was applied by spinning VM652 (HD
Microsystems) onto the surface at 5000 rpm for 30 seconds to improve the adhesion [16], and to
avoid silicon delamination or peel off. Next, the layer was soft-baked at 120 C for 60 seconds.
Then, polyimide PI2611 was spun on the surface at 500 rpm for 5 seconds with an acceleration
speed of 250 rpm, then spun at 3000 rpm for 30 seconds with an acceleration speed of 1000 rpm.
It was then soft-baked at 120 C for 5 minutes in the dehydration bake oven. The next step was a
?low stress? curing process in the programmable YES oven (model 450-PB8-2P-CP). The wafer
was placed in a programmable nitrogen oven at ambient temperature and the process began by
purging the oven with N2. Table 3.1 is the process used to cure the polyimide layer under low
stress using the YES oven.
3.2.5 Etch Openings in Polyimide
And then, it made each opening in polyimide. The silicon wafer was placed in the dehydra-
tion bake oven at 120 C for 20 minutes before the application of photoresist. HMDS was then
evaporated on the surface of the silicon wafer for 5 minutes to improve the photoresist adhesion. A
layer of thick photoresist, AZ4620 (AZ Electronic Materials), was used for patterning. It was spun
on the surface at a speed of 1000 rpm for 30 seconds with an acceleration speed of 250 rpm. It
14
was then soft-baked at 110 for 90 seconds. After that, the silicon wafer was exposed with 3 cycles
for 90 seconds. Exposure was performed immediately after the soft bake, as soon as the wafer had
cooled to room temperature. It was then developed with AZ400K positive photoresist developer
for 5 minutes which was in the ratio of 1:2.5 with deionized water. Then, a plasma descum with
Matrix (300W, 20sec) was performed to remove any residual photoresist. After completing this
step, photoresist thickness was about 20 um. The mask layout is shown in figure 3.7.
Figure 3.7: Wire bond mask2
3.2.6 Polyimide Etching
The last step is polyimide etching using A.O.E. The etching process was P-imide2 which
contains ICP-2min, Polyetch1-3min, Polyetch2-4min.
3.2.7 Photoresist Layer 1
The silicon wafer was placed in the dehydration bake oven at 120 C for 20 minutes ahead
of the application of photoresist. HMDS was then evaporated on the surface of the silicon wafer
for 5 minutes to improve the photoresist adhesion. Positive photoresist AZ4620 was then spun on
the surface at a speed of 1000 rpm for 30 seconds with an acceleration speed of 250 rpm and soft-
baked at 110 C for 90 seconds. After that, it was immediately exposed (Mask Alligner MA6/BA6)
15
to make a pattern with 3 cycles for 90 seconds as soon as the wafer had cooled down to room
temperature. It was then developed with AZ400K positive photoresist developer for 4 minutes
which was in the ratio of 1:2.5 with deionized water. Then, a plasma descum with Matrix (300W,
20sec) was performed to remove any residual photoresist. After completing this step, photoresist
thickness is about 20 um. The mask layout is shown in figure 3.4.
3.2.8 Backside Metal Deposition
It began with E-beam deposition process. It was selected with Titanium (500 ?A) and Nickel
(800 ?A) Gold (1200 ?A) as backside metal deposition. Then, the ceramic wafer is diced to separate
all the individual probe arrays.
3.2.9 Summary
Due to reasons which will be discussed in a later part of this paper, it was found that flip chip
solution has larger inelastic stress / strain than wire bond solution since flip chip has the largest
CTE mismatch between Si substrate and Al203 package compared to wirebond, AlN substrate and
Al203 package. Moreover, wire bond solution is more suitable for multi chip module in harsh
environments.
16
Chapter 4
Selected Optimum Solutions for Multi Chip Module Technology
As more complex IC (Integrated Circuit) designs are utilized, IC needs to be packaged more
densely than before. Now MCMs are being widely used in several applications for military, space,
medical, and commercial products. This chapter addresses an overview of MCM design for harsh
conditions implemented in systems in package (SiP) technologies. The basic idea of the MCM is
to combine multiple chips as a single chip.
It uses two different dielectric materials, BCB and Polyimide. An advanced electronic resin,
BCB which is drived from B-staged bisbenzocyclobutene chemistry [18], has been developed for
use as dielectrics in thin film microelectronics applications. BCB has significant advantages over
the polyimide such as lower dielectric constant, a lower dissipation factor, reduced water absorp-
tion, a shorter cure time and a lower cure temperature [18], [19], [20]. However, BCB has signifi-
cantly higher CTE than polyimide. That is, it is important to use a passivation film at the uppermost
layer in MCM with thick polyimide. Moreover, BCB was used as dielectrics at lower layers and
polyimide was used as dielectrics at higher layers. Table 4.1 shows a comparison of BCB with
Polyimide.
Table 4.1: Comparison of BCB with Polyimide
Properties BCB Polyimide
Dielectric constant(1kHz) 2.65 2.9
CTE(ppm/ C) 42 3
Tensile strengh(MPa) 87 9 350
Modulus(GPa) 2.9 0:2 8.5
Dissipation factor 0.0008 0.002
17
4.1 MCM Process Procedures
This is the MCM solution using BCB and Polyimide. Figure 4.1 shows basic MCM process
flows. Either BCB or Polyimide can be used as a dielectric layer.
MCM were fabricated on 4 X 4 inch AIN substrates. An oxide insulating layer of thickness
1600 ?Awas grown on the ceramic substrate in an oxidation furnace. The oxidation performed in the
presence of oxygen and hydrogen gases at 1050 C.
Figure 4.1: MCM process flow
18
4.1.1 Seed Layer Deposition
The multi Chip Module process began with E-beam deposition process with 1 minute Ion
Clean. It was selected with Titanium (250 ?A) and Copper (2000 ?A) as the seed layer deposition.
4.1.2 Ground plane Patterning
The second step is to make the Ground pattern. The ceramic wafer was placed in the dehydra-
tion bake oven at 120 C for 20 minutes ahead of the application of photoresist. HMDS was then
evaporated on the surface of the ceramic wafer for 5 minutes to improve the photoresist adhesion.
A layer of thick photoresist, AZ9245 (AZ Electronic Materials), was used for patterning. It was
spun on the surface at a speed of 1700 rpm for 5 seconds with an acceleration speed of 500 rpm and
spun at 2700 rpm for 30 seconds with an acceleration speed of 1000 rpm. It was then soft-baked
at 110 C for 90 seconds. After that, the ceramic wafer was exposed (Mask Alligner MA6/BA6) to
make a pattern with 1 cycle for 60 seconds. Exposure was performed immediately after soft bake,
as soon as the wafer had cooled to room temperature. It was then developed with AZ400K positive
photoresist developer for 5 minutes which was in the ratio of 1:2.5 with deionized water. Then,
a plasma descum with Matrix (300W, 20sec) was performed to remove any residual photoresist.
After completing this step, photoresist thickness is about 4.5 um. The mask layout is shown in
figure 4.1.
4.1.3 Plating
The next step is metallization. First, copper (2um, 26mA) was plated. Nickel was then plated
(0.2um, 23mA, 35-40 C). The Nickel layer served as a thermal barrier. It was used to improve the
adhesion to the resistor material. After plating, positive photoresist was stripped by acetone and
was cleaned by methanol and water. Then the seed layers were etched by Cu etchant, Potassium
Permanganate.
19
4.1.4 BCB Application
Prior to the application of BCB, the ceramic wafer was placed in a dehydration bake YES
oven at 150 C. For BCB application - a thin layer of adhesion promoter was formed by spinning
AP3000 onto the surface at 3000 rpm for 30 seconds to improve the adhesion to the BCB layer.
Poor adhesion of ceramic materials can show up as delamination or peel off during the rest of
the processing steps. The layer was soft-baked at 120 C for 60 seconds. After that, BCB was
dispensed on the surface at 500 rpm for 5 seconds with an acceleration speed of 100 rpm then,
spun at 500 rpm for 5 seconds. It was then spun at 5000 rpm for 10 seconds with an acceleration
speed of 450 rpm then, spun at 5000 rpm for 30 seconds. Next, it was soft-baked at 60 C for 90
seconds. Then, the ceramic wafer was exposed (Mask Alligner MA6/BA6) to make a pattern with
1 cycle for 30 seconds. Then, pre-develop baked at 60 C for 5 minutes. Without the pre-develop
process, the development time will increase. Next, it was then developed with DS3000 at 35 C
for 2.2minutes and then, it was develop-rinse step with DS3000 at the room temperature for 2.2
minutes. Next, it was done Nitrogen-dry. When it was water cleaned, it got cracks everywhere on
areas with metal underneath. After blow-drying the wafer, post-develop bake was done at 100 C
for 1 minute. When inspecting wafers after bake, a lot of particles were exposed all over the
surface which means cleaning was very important on them using BCB and they needed N2 blow
every step. The next step was a ?low stress? curing process in the programmable YES oven (model
450-PB8-2P-CP). The wafer was placed in a programmable nitrogen oven at ambient temperature
and the process began by purging the oven with N2. Table 4.1 is the process used to cure the
polyimide layer under low stress using YES oven.
4.1.5 2nd Layer BCB Application
Prior to the application of BCB, the ceramic wafer was placed in a dehydration bake YES
oven at 150 C. And then, the layer was soft-baked at 120 C for 60 seconds. Next, BCB was
dispensed on the surface at 500 rpm for 5 seconds with an acceleration speed of 100 rpm and spun
at 500 rpm for 5 seconds. It was then spun at 5000 rpm for 10 seconds with an acceleration speed
20
Table 4.2: YES Oven Procedure
STEP PROCESS
Step1 Target Temperature: 150 Ramp rate: 1/Min Curing time:15 min
Step2 Target Temperature: 150 Soak rate: 1/Min Curing time:15 min
Step3 Target Temperature: 210 Ramp rate: 2/Min Curing time:60 min
Step4 Target Temperature: 210 Soak rate: 2/Min Curing time:60 min
Step5 Cool down to ambient temperature at a rate of 2degree/min
of 450 rpm and then spun at 5000 rpm for 30 seconds. After this, it was then soft-baked at 60 C for
90 seconds. Then, the ceramic wafer was exposed (Mask Alligner MA6/BA6) to make a pattern
with 1 cycle for 40 seconds. The next step was pre-develop bake at 60 C for 5 minutes. It was then
developed with DS3000 at 35 C for 2.2minutes and then it was develop-rinse step with DS3000 at
the room temperature for 2.2 minutes. After blow-drying the wafer, post-develop bake was done
at 100 C for 1 minute. The next step was a ?low stress? curing process in the programmable YES
oven (model 450-PB8-2P-CP). The wafer was placed in a programmable nitrogen oven at ambient
temperature and the process began by purging the oven with N2. Table 4.2 is the process used to
cure the polyimide layer under low stress using YES oven.
4.1.6 Descum
To remove a thin film of polymer residue left behind in the develop process, it was descummed
by plasma matrix. An etch gas was 80:20 02/CF4(30 seconds) [18].
21
4.1.7 Copper Brightening(oxidized during descum)
An O2/CF4 plasma will cause corrosion of copper. After the descum, a 30 second dip in 10
percent acetic acid was immediately necessary to prevent corrosion and discoloration of the copper
surface [18].
4.1.8 Seed Layer Deposition
For the next layer deposition, E-beam deposition was processed with 3 minutes Ion Clean. It
was selected with Chrome (250 ?A) and Copper (2000 ?A) as the seed layer deposition.
4.1.9 Layer 2 Patterning
The second step is to make the second plane pattern. The ceramic wafer was placed in the
dehydration bake oven at 120 C for 30 minutes ahead of the application of photoresist. HMDS
was then evaporated on the surface of the ceramic wafer for 10 minutes to improve the photoresist
adhesion. A layer of thick photoresist, AZ9245 (AZ Electronic Materials), was used for patterning.
It was spun on the surface at a speed of 1700 rpm for 5 seconds with an acceleration speed of 500
rpm and spun at 2700 rpm for 30 seconds with an acceleration speed of 1000 rpm. It was then
soft-baked at 110 C for 90 seconds. After that, the ceramic wafer was exposed (Mask Alligner
MA6/BA6) to make a pattern with 1 cycle for 60 seconds. Exposure was performed immediately
after soft bake, as soon as the wafer had cooled to room temperature. It was then developed with
AZ400K positive photoresist developer for 5 minutes which was in the ratio of 1:2.5 with deionized
water. Then, a plasma descum with Matrix (300W, 20sec) was performed to remove any residual
photoresist. After completing this step, photoresist thickness should be about 4.5 um. The mask
layout is shown in figure 4.4.
4.1.10 Plating
The next step is metallization. First, copper (2um, 26mA) was plated. Nickel was then plated
(0.2um, 23mA, 35-40 C). The Nickel layer served as a thermal barrier. It was used to improve the
22
adhesion to the resistor material. After plating, positive photoresist was stripped by acetone and
was cleaned by methanol and water. Then the seed layers were etched by Cu etchant, Potassium
Permanganate.
4.1.11 Polyimide Application
This step is to make a coating of polyimide on the surface. Prior to the application of poly-
imide, pre-poly etch was accomplished. Then, the ceramic wafer was placed in a dehydration bake
oven at 120 C for 30 minutes. For non-polyimide surfaces - a thin layer of adhesion promoter
VM652 (HD Microsystems) was formed to improve the adhesion to the polyimide layer. It was
spun on the surface at a speed of 500 rpm for 5 seconds with an acceleration speed of 500 rpm
and spun at 4000 rpm for 30 seconds with an acceleration speed of 1000 rpm. Poor adhesion of
silicon materials can show up as silicon delamination or peel off during the rest of the processing
steps. The layer was soft-baked at 120 C for 60 seconds. Next, the polyimide PI2611 was spun
on the surface at 500 rpm for 5 seconds with an acceleration speed of 500 rpm then, spun at 3000
rpm for 30 seconds with an acceleration speed of 1000 rpm. Next, it was soft-baked at 120 C
for 5 minutes in the dehydration bake oven. The next step was a ?low stress? curing process in
the programmable YES oven (model 450-PB8-2P-CP). The wafer was placed in a programmable
nitrogen oven at ambient temperature and the process began by purging the oven with N2. Table
3.1 is the process used to cure the polyimide layer under low stress using YES oven.
4.1.12 Photoresist Patterning for Polyimide Etching
The next step is to make the patterning for polyimide etching. The ceramic wafer was placed
in the dehydration bake oven at 120 C for 60 minutes ahead of the application of photoresist.
HMDS was then evaporated on the surface of the ceramic wafer for 10 minutes to improve the
photoresist adhesion. A layer of thick photoresist, AZ 4620(AZ Electronic Materials), was used
for patterning. It was spun on the surface at a speed of 1000 rpm for 30 seconds with an acceleration
speed of 250 rpm and it was then soft-baked at 110 C for 90 seconds. Rest period (43.5, 73.5, 1hr)
23
was applied. To make a pattern, it was exposed with 3 cycles 25 seconds, 30 seconds in between.
Exposure was performed immediately after soft bake, as soon as the wafer had cooled to room
temperature. It was then developed with AZ400K positive photoresist developer for 3.5 minutes
which was in the ratio of 1:2.5 with deionized water. Next, a plasma descum with Matrix (300W,
20sec) was performed to remove any residual photoresist. After completing this step, photoresist
thickness should be about 15 um. The mask layout is shown in figure 4.5.
4.1.13 Polyimide Etching
After mounting on 5-in wafer with blue wax(105, 30 seconds), it was etched by (A.O.E).
Conversely, it released substrate from the backing (105 to remove, clean with Amyl Acetate).
4.1.14 Seed Layer Deposition
For the next layer deposition, E-beam deposition was processed with 3 minutes Ion Clean. It
was selected with Chrome (250 ?A) and Copper (2000 ?A) as the seed layer deposition.
4.1.15 Photoresist Patterning
The ceramic wafer was placed in the dehydration bake oven at 120 C for 30 minutes ahead
of the application of photoresist. HMDS was then evaporated on the surface of the ceramic wafer
for 10 minutes to improve the photoresist adhesion. A layer of thick photoresist, AZ9245 (AZ
Electronic Materials), was used for patterning. It was spun on the surface at a speed of 1700
rpm for 5 seconds with an acceleration speed of 500 rpm and spun at 2700 rpm for 30 seconds
with an acceleration speed of 1000 rpm. It was then soft-baked at 110 C for 90 seconds. After
that, the ceramic wafer was exposed (Mask Alligner MA6/BA6) to make a pattern with 1 cycle
for 60 seconds. Exposure was performed immediately after soft bake, as soon as the wafer had
cooled to room temperature. It was then developed with AZ400K positive photoresist developer
for 5 minutes which was in the ratio of 1:2.5 with deionized water. Then, a plasma descum with
24
Matrix (300W, 20sec) was performed to remove any residual photoresist. After completing this
step, photoresist thickness should be about 4.5 um. The mask layout is shown in figure 4.6.
4.1.16 Plating
The next step is metallization. First, copper (1.25um, 26mA) was plated. Nickel was then
plated (1.0um, 23mA, 35-40 C). And then Au was plated (1.5um, 20mA, 35-40 C). The Nickel
layer served as a thermal barrier. It was used to improve the adhesion to the resistor material.
And then, a plasma descum with Matrix (300W, 20sec) was performed to remove any residual
photoresist. It was very helpful to prevent any current leakages in testing MCMs. After plating,
positive photoresist was stripped by acetone and was cleaned by methanol and water. Then the
seed layers were etched by Cu etchant(49-1), Potassium Permanganate.
4.1.17 Polyimide Application
This step is to make a coating of polyimide on the surface. The ceramic wafer was placed
in a dehydration bake oven at 120 C for 30 minutes. For non-polyimide surfaces - a thin layer
of adhesion promoter VM652 (HD Microsystems) was formed to improve the adhesion to the
polyimide layer. It was spun on the surface at a speed of 500 rpm for 5 seconds with an acceleration
speed of 500 rpm and spun at 4000 rpm for 30 seconds with an acceleration speed of 1000 rpm.
Poor adhesion of silicon materials can show up as silicon delamination or peel off during the rest of
the processing steps. The layer was soft-baked at 120 C for 60 seconds. Next, polyimide PI2611
was spun on the surface at 500 rpm for 5 seconds with an acceleration speed of 500 rpm and then
spun at 3000 rpm for 30 seconds with an acceleration speed of 1000 rpm. It was then soft-baked at
120 C for 5 minutes in the dehydration bake oven. The next step was a ?low stress? curing process
in the programmable YES oven (model 450-PB8-2P-CP). The wafer was placed in a programmable
nitrogen oven at ambient temperature and the process began by purging the oven with N2. Table
3.2 is the process used to cure the polyimide layer under low stress using YES oven.
25
4.1.18 Photoresist Patterning for Polyimide Etching
Next, it is to make the patterning for polyimide etching. The ceramic wafer was placed in the
dehydration bake oven at 120 C for 60 minutes ahead of the application of photoresist. HMDS
was then evaporated on the surface of the ceramic wafer for 10 minutes to improve the photoresist
adhesion. A layer of thick photoresist, AZ 4620(AZ Electronic Materials), was used for patterning.
It was spun on the surface at a speed of 1000 rpm for 30 seconds with an acceleration speed of 250
rpm and it was then soft-baked at 110 C for 90 seconds. Rest period (43.5, 73.5, 1hr) was applied.
To make a pattern, it was exposed with 3 cycles 25 seconds, 30 seconds waiting time in between.
Exposure was performed immediately after soft bake, as soon as the wafer had cooled to room
temperature. It was then developed with AZ400K positive photoresist developer for 3.5 minutes
which was in the ratio of 1:2.5 with deionized water. Then, a plasma descum with Matrix (300W,
20sec) was performed to remove any residual photoresist. After completing this step, photoresist
thickness should be about 15 um. The mask layout is shown in figure 4.6.
4.1.19 Polyimide Etching
After mounting on 5-in wafer with blue wax(105, 30 seconds), it was etched by (A.O.E).
Conversely, it released substrate from the backing (105 to remove, clean with Amyl Acetate).
4.1.20 Backside Metal Deposition
It was selected with Chrome (500 ?A), Nickel (800 ?A), and Gold (1200 ?A) as backside metal
deposition with 1 minute Ion Clean.
4.1.21 Dice
The silicon wafer is diced to separate all the individual probe arrays.
26
Figure 4.2: MCM Layout 1
Figure 4.3: MCM Layout 2
27
Figure 4.4: MCM Layout 3
Figure 4.5: MCM Layout 4
28
Figure 4.6: MCM Layout 5
Figure 4.7: MCM Layout 6
29
4.2 Test Chip Procedures for MCM
The test chips are fabricated on 4-inch diameter silicon substrates. Based on the analysis
presented in the previous chapter, wire bond solution is adopted to interconnect the test chip to
MCM.
4.2.1 Patterning
The first step is to make the seed layer pattern. The silicon wafer was placed in the dehydration
bake oven at 120 C for 20 minutes ahead of the application of photoresist. HMDS was then
evaporated on the surface of the silicon wafer for 5 minutes to improve the photoresist adhesion. A
layer of thick photoresist, AZ9245 (AZ Electronic Materials), was used for patterning. It was spun
on the surface at a speed of 1700 rpm for 5 seconds with an acceleration speed of 500 rpm and
spun at 2700 rpm for 30 seconds with an acceleration speed of 1000 rpm. It was then soft-baked
at 110 C for 90 seconds. After that, the silicon wafer was exposed (Mask Alligner MA6/BA6) to
make a pattern with 1 cycle for 60 seconds. Exposure was performed immediately after soft bake,
as soon as the wafer had cooled to room temperature. It was then developed with AZ400K positive
photoresist developer for 5 minutes which was in the ratio of 1:2.5 with deionized water. Next,
a plasma descum with Matrix (300W, 20sec) was performed to remove any residual photoresist.
After completing this step, photoresist thickness should be about 4.5 um.
4.2.2 Plating
The next step is metallization. First, copper (2.5um, 12mA) was plated. Nickel was then
plated (0.5um, 10mA, 35-40 C). And then Au was plated (0.5um, 6mA, 35-40 C). Positive pho-
toresist was stripped by acetone and was cleaned by methanol and water. Then the seed layers
were etched by Cu etchant, Potassium Permanganate. Next, the silicon wafer is diced to separate
all the individual probe arrays.
30
Figure 4.8: Test chip layout for MCM
31
Chapter 5
Test & Results
Flip chip and Wire bond test vehicles were packaged by Dr. Johnson?s group.
Various packaging structures were tested in order to evaluate the effectiveness of their design and
to lay the groundwork for future models. The first step was to expose the temperature and stress
distributions of the various models to extreme temperature cycling. This was simulated using com-
mercially available software. Both of the two types of packaging structures, die to substrate and
substrate to package, were designed and developed. An indium was used for substrate attachment
to package of wire bond and solder was used for flip chip. Both indium substrate attachments were
used.
5.1 Interaction between Die to Substrate Attach
They were exposed to temperature cycles(100), -180 C and +125 C, by dipping them into
liquid nitrogen, then they were allowed to cool down to room temperature. Then, a half-diagonal
finite element model has been constructed, which covers all the critical joints between die to sub-
stract attach and substrate to package attach.
5.1.1 Wire Bond
As it was discussed, each bond of face-up chips is connected individually with a wire. It was
used Indium(0.05mm thick) which is used with high temperature alloys to attach a 5mm x 5mm
die onto an AlN substrate(48mm x 13mm) and to attach an an AIN substrate(48mm x 13mm) onto
an Al203 package(53mm x 14mm) as shown in Figure 5.1. The commercial FEM software package
ANSYS 5.2 was utilized to perform a detailed analysis of temperature and stress/strain distribution
in the package. According to simulation (Figure 5.2), the edge-corner of the substrate attachment
32
has the maximum stress/strain instead of that of the die attachment because it has smaller CTE
mismatch, AIN to Al203, compared with that of die to substrate, SiGe to AIN CTE mismatch. In
the result, substrate to package interconnection is the critical issue in reliability.
Figure 5.1: Wire bond
Figure 5.2: Strain distribution in the whole wire bond die attach type package[14]
5.1.2 Flip Chip
As it was discussed, it uses the direct electrical connection of face-down chips in opposition to
wire bond. It was used solder bump(In50Pb) to attach a die onto an Si substrate as shown in Figure
5.3. It has the largest CTE mismatch between the substrate(Si) and the package(Al203). Therefore
the maximum stress/strain was found to be accumulated at substrate and package interconnection
(Figure 5.4).
33
Figure 5.3: Flipchip
Figure 5.4: Strain distribution in the whole flip chip type package[14]
34
5.2 Interaction between Substrate to Package Attach
The CTE mismatch between die and substrate is negligible compared to that of between sub-
strate(Si / AIN) and package(Alumina). The strain is mainly concentrated at the outer upper side of
the indium substrate attach (Figure 5.5). Moreover, the substrate attachment is the critical concern
in reliability.
Figure 5.5: FEM model of substrate to package[14]
5.3 Test Vehicles
Two test vehicles were fabricated for test. A single array chip of wire bond and flip chip are
shown in Figure 5.6 and Figure 5.7. The 5mm x 5mm PB6 wire bond test die and 5mm x 5mm
FA-10 flip chip test die were purchased from Delphi Electronics. The die was backside metal de-
posited with Ti/Ni/Au. Two hermetically sealed packages were subjected to temperature thermal
shock cycles(100), -180 C and +125 C, by repeatedly dipping them into liquid nitrogen and then
they were allowed to cool down to room temperature [20]. It was not observed solder joint failures
and degradation of the dielectric layers. However, flip chip was not as good as the wire bond,
because flip chip has the largest CTE mismatch between Si substrate and Al203 package compared
35
to the wire bond, AlN substrate and Al203 package. The maximum deformation of flip chip and
wire bond were 0.004992 and 0.073313. The corresponding maximum stress values are 0.066784
and 0.098391 at the edge of the corner between substrate and package.
(a) Wire bond single array (b) Wire bond testing chip
Figure 5.6: Wire bond test vehicle
(a) Flip chip single array (b) Flip chip testing chip
Figure 5.7: Flip chip test vehicle
5.4 MCM Package
Two MCM packages shown in Figure 5.8 and 5.9 were hermetically sealed and each sealed
package has been cycled between -180 C and 125 C for 20 complete cycles. The inter-connections
between wire bond pad and lead are shown in figure 5.10 and 5.11. Reliability screening tests were
performed for MCM. It was observed that no electrical or solder joint failures occured.
36
Figure 5.8: MCM before sealed
Figure 5.9: MCM after sealed
37
Figure 5.10: PAD-LEAD connection-1
38
Figure 5.11: PAD-LEAD connection-2
39
Figure 5.12 shows temperature cycles. The resistances of MCM were initially 140.5W and
183.1W at room temperature. It was observed that the resistances were slightly increased after 1
cycling to 142.2W and 185W at room temperature shown in Figure 5.13. Then, they were 144.8W,
189W at room temperature after 5 thermal shock cycles, and 144.2W, 188.7W at room temperature
after 10 thermal shock cycles. Next, they were 141.8W, 187.2W at room temperature after 20
thermal shock cycles. The possible reason in decreasing the resistance after 20 cycles was the
difference of the room temperature on that day. Resistance was observed 145.0W, 188.6W after 30
cycles and 144.4W, 188.1W after 50 cycles.
There was an initial increase in resistance. However, this result seems to be very reasonable
since the resistance remained almost constant after 10 thermal shock cycles. Even though the
resistance was changed slightly by temperature, the processes have been demonstrated and initial
reliability screening results are verified.
Figure 5.12: Temperature cycling
40
Figure 5.13: Resistance changes after 50 thermal shock cycles
41
Chapter 6
Conclusion
This paper presents a design fabrication technology of Multi Chip Module which is suitable
for wide range temperature. In this work, multi chip module chips using wire bond have been
successfully fabricated. Thermal shock cycle tests have been performed for flip chip and wire
bond before adapting MCM. Then, wire bond and flip chip solution were compared. It was shown
that Flip chip and Wire bond are working exactly as expected even though wire bond has less stress
campared with flip chip. Minimizing the stress is especially important when mulitiple copper/low-
k interconnection layers are present. These results have indicated that wire bond is more suitable
for the multi chip module.
The surface of flip chip and wire bond which is electroplated on the gold layer tends to disin-
tegrate because of repeated use of the plating solution. Identifying the reason for this occurrence
and the improvement in the plating process will make the chips more durable by preserving the
required electrical characteristics of the electrode.
Multi chip module packages using wire bond solution were fabricated. Then, thermal relia-
bility testing of MCM was performed. The resistance of MCM was getting larger, but it was stable
after 10 thermal shock cycle tests. Additional long-term stability test will be subjected in the future
including failure analysis.
42
Bibliography
[1] Blum N. A., Charles H. K. Jr., Francomacaro A. S., Multichip Module Substrates, Johns
Hopkins APL Tech. Dig. 20(1), 1999, Page(s): 62-69
[2] E. M. Davis, W. E. Harding, R. S. Schwartz and J. J. Coring, Solid Logic Technology: Ver-
satile High Performance Microelectronics, IBM Journal of Research & Development, Vol. 8,
1964, Page(s): 102
[3] Rajen Chanchani, Peter M. Hall: Temperature Dependence of Thermal Expansion of Ce-
ramics and Metals for Electronic Packages, IEEE Transactions on Components, Hybrids and
Manufacturing Technology, vol. 13, no. 4, Page(s): 743-750, 1990.
[4] Zhuqing Zhang and C. P. Wong, Recent advances in flip-chip underfill materials, process, and
reliability, IEEE Transactions on Advanced Packaging, vol. 27, no. 3, 2004, Page(s): 515-524
[5] Dale J.R., Oldfield R.C., Mechanical Stresses Likely to be Encountered in the Manufacture
and Use of Plastically Encapsulated Devices, Microelectronics and reliability, Vol. 16, 1977,
Page(s): 255-258
[6] Md. Sayed Kaysar Rahim: Die Stress Characterization and Interface Delamination Study in
Flip Chip on Laminate Assemblies
[7] Thomas R.E., Stress-Induced Deformation of Aluminum Metallization in Plastic Encapsu-
lated Semiconductor Devices Caused by Thermal Shock, Proceedings of the 35th Electronic
Components conference, IEEE, 1985, Page(s): 37-45
[8] Schroen W.H, Spencer J.L., Bryan J.A., Cleveland R.D., Metzgar T.D., Edwards D.R., Reli-
ability Tests and Stress in Plastic Integrated Circuits, Proceedings of the 19th Annual Relia-
bility Physics Symposium, IEEE, Page(s): 81-87
[9] Kitano M., Nishimura A., Kawai S., Analysis of Packaging Cracking During Reflow Solder-
ing Process, Proceeding of the 26th International Reliability Physics Symposium, Page(s):
90-95
[10] JPL Publication 99-1, MEMS Reliability Assurance Guidelines for Space Applications
[11] Lau H., Thermal Stress and Strain in Microelectronics Packaging, Van Nostrand Reinhold,
1993
[12] Edwards D.R., Heinen K.G., Martinez J.E. and Groothuis S., Shear Stress Evaluation of
Plastic Packages, Proceedings of the 37th Electronic Components Conference, IEEE, 1987,
Page(s): 84-95
43
[13] Malik, N.H., Al-Arainy, A.A., and Qureshi, M.I., ElectricalInsulation in Power System,
Page(s): 164-167, 1998 Marcel Dekker Inc, New York
[14] Senthil Sivaswamy, Rui Wu, Charles Ellis, Michael Palmer, R. Wayne Johnson, Patrick Mc-
Cluskey2 and Kevin Petrarca: System-in-Package for Extreme Environments, Electronic
Components and Technology Conference, 2008. ECTC 2008. 58th Volume , Issue , 27-30
May 2008, Page(s): 2044 - 2050
[15] Wu Rui and McCluskey, F. Patrick, Constitutive relations of indium solder joint in cold tem-
perature electronic packaging based on anand model, Proceedings of Itherm2008 (accepted),
May, 2008, Florida, USA
[16] Soo-Jin Park, Eun-Jung Lee and Soo-Han Kwon: Influence of Surface Treatment of Poly-
imide Flim on Adhesion Enhancement between Polyimide and Metal Films, Korean Chem.
Soc. 2007, Vol. 28, No. 2
[17] Seung Wook Yoon, Vaidyanathan Kripesh, Li Hong Yu, Li Chao Yong, Su Yong Ji Jeffrey
and Mahadevan K. Iyer: Flipchip Bump Integrity with Copper/Ultra Low-k Dielectrics for
Fine Pitch Flipchip Packaging, Microelectronics and Reliability, Volume 46, Issue 1, January
2006, Page(s): 155-163
[18] Cyclotene 4000 Series Advanced Electronic Resins (Photo BCB), The Dow Chemical Com-
pany
[19] PI-2600 Series Brochure - Low Stress Polyimides, HD MicroSystems
[20] June Nguyen, Oksun Dydasco, Harutoshi Saigusa, Chang-Hwang Hua, Skyworks Solutions,
Inc., Sunnyvale Operations, 1230 Bordeaux Drive, Sunnyvale, CA 94089, A comparison of
BCB with Polyimide process in manufacturing HBT devices
44