Microstructural Evolution of Al-Cu
Intermetallic Phases in Wire-Bonding
Research Thesis
Submitted In Partial Fulfillment of the Requirements for the
Degree of Master of Science in Materials Engineering
Maria Drozdov
Submitted to the Senate of
The Technion – Israel Institute of Technology
Tishrey 5768
Haifa
September 2007
ACKNOWLEDGMENTS
The research thesis was done under the supervision of Prof. Wayne D. Kaplan in the
department of Materials Engineering at the Technion-Israel Institute of Technology.
I would like to express my gratitude to my supervisor, Prof. Wayne Kaplan for his
guidance, endless patience, encouragement and availability also beyond conventional
working hours.
I thank Prof. D.G. Brandon, Prof. R. Chaim, Dr. Y. Kaufmann, Dr. A. Berner, Dr. A.
Lahav and M. Kalina for fruitful discussions, help and guidance.
I thank my colleagues and friends from the metal-ceramic interface group: H. Sadan,
L. Miller, A. Karpel, M. Baram, A. Shlayer, Dr. T. Paramasivam, G. Gluzer, S. Zamir,
R. Segev, M. Libertal, M. Katz and M. Levin for their friendship and support.
Many thanks to my friends: Ruth, Guy, Yury and others, for supporting me
throughout this period.
A special thank to my mother Mina, for her encouragement and love.
The generous financial assistance from “Kulicke and Soffa” is gratefully
acknowledged. I also thank Dr. Z. Atzmon and G. Gur for their help.
II
TABLE OF CONTENTS
Abstract ...................................................................................................... 1
List of Symbols and Acronyms ..................................................................... 3
CHAPTER 1 :
Introduction ....................................................................... 4
CHAPTER 2 :
Literature Survey ............................................................... 6
2.1.
The Wire-Bonding Process .................................................................................... 6
2.2.
Copper Wires in the Wire-Bonding Process ........................................................... 8
2.3.
The Al-Cu Phase Diagram .....................................................................................10
2.4.
Some Properties of Intermetallic Al-Cu Phases ....................................................11
2.5.
Phase Formation in Bulk Al-Cu Welded Joins ........................................................13
2.6.
Phase Formation in Al-Cu Thin Films ....................................................................15
2.7.
Influence of Oxygen on Al-Cu Intermetallics .........................................................19
2.8.
Phase Formation in Al-Cu Wire Bonds ..................................................................21
2.9.
Influence of Al-Cu Wire-Bonding Process Parameters on Bond Reliability ............23
2.9.1.Factors Affecting Al Cracks, Si Cratering and Bond Strength ..............................23
2.9.2.Factors Affecting the Copper Ball Hardness .......................................................24
2.10. Kirkendall Effect ..................................................................................................24
2.11. Characterization of Al-Cu Wire-Bonds ..................................................................25
2.12. Investigation of Al-Cu Intermetallics Using Focused Ion Beam .............................27
2.13. Research Goals....................................................................................................29
CHAPTER 3 :
Experimental Procedures & Characterization
Techniques
30
3.1.
Specimen Preparation .........................................................................................30
3.1.1.Free Air Balls ......................................................................................................30
3.1.2.Al-Cu Reference Specimen .................................................................................30
3.1.3.Al-Cu Wire-Bonds ................................................................................................30
3.2.
Characterization ..................................................................................................33
3.2.1.Initial Characterization ........................................................................................33
3.2.2.Morphological Characterization ..........................................................................34
3.2.3.Compositional Analysis .......................................................................................35
CHAPTER 4 :
4.1.
Results ............................................................................. 37
Free Air Balls .......................................................................................................37
4.1.1.Morphology .........................................................................................................37
4.1.2.Heat-Affected Zone .............................................................................................38
4.2.
As-Bonded Specimens .........................................................................................39
4.3.
Heat-Treated Specimens......................................................................................43
III
4.3.1.Specimens Heat-Treated for 2 Hours ..................................................................43
4.3.2.Specimens Heat-Treated for 24 Hours ................................................................48
4.3.3.Specimens Heat-Treated for 96 Hours ................................................................52
4.3.4.Specimens Heat-Treated for 200 Hours...............................................................58
4.4.
Microstructural and Compositional Evolution of the Wire-Bonds...........................64
CHAPTER 5 :
Discussion ....................................................................... 66
5.1.
Copper Free Air Balls ...........................................................................................66
5.2.
As-Bonded Specimens .........................................................................................66
5.2.1.Morphology of the Bonded Copper-Ball and Aluminium Metallization ...................66
5.2.2.Intermetallic Formation .......................................................................................68
5.3.
Heat-Treated Specimens......................................................................................69
5.3.1.Morphology of the Heat-Treated Specimens ........................................................69
5.3.2.Composition of the Al-Cu Intermetallics...............................................................69
5.3.2.1.Imprecision in Composition Measurements ......................................................69
5.3.2.2.Oxidation of the Heat-Treated Specimens ........................................................72
5.3.3.Morphology of the Al-Cu Intermetallics ................................................................73
5.3.3.1.Voids Found at the Al-Cu Interface ...................................................................73
5.3.3.2.Al2Cu Intermetallic Phase Growth.....................................................................74
CHAPTER 6 :
Summary and Conclusions ............................................... 76
CHAPTER 7 :
Future Outlook ................................................................. 79
7.1.
Recommendations for the Process Improvement .................................................79
7.2.
Recommendations for Future Work ......................................................................79
Appendix ................................................................................................... 81
I.
Crystallographic Data Used for Compositional Analysis .......................................81
II.
Evaluation of the EFTEM for Composition Measurements .....................................81
III.
Computation of the Limit of Detection of Al in Cu ..................................................82
IV.
Calculation of the Formation Energy for the Al2Cu Phase ......................................83
References ................................................................................................ 84
IV
LIST OF FIGURES
Figure 1: Schematic representation of the wire-bonding process. ............................................. 6
Figure 2: Schematic representation of the microstructure of the wire after FAB formation [1]. .. 8
Figure 3: The aluminum-copper binary phase diagram [18]. .................................................... 11
Figure 4: Schematic diagram of in-situ phase formation in thin Al-Cu film couples as a function
of annealing temperatures [31]............................................................................... 17
Figure 5: A photograph of an experimental setup for Cu wire-bonding. An electric spark is
transferred to the wire via the EFO wand, and a reducing gas (95%N2, 5%H2) is
supplied via a small tube. ....................................................................................... 31
Figure 6: A photograph of a characteristic device strip. The wiring directions are indicated. .. 32
Figure 7: Secondary electron (SE) SEM micrograph of a copper wire-bond cross-section.
Extensive curtaining can be seen at the periphery of the bond................................ 33
Figure 8: Schematic drawing of (0 0 1) plane in an FCC structure (a) in Bragg conditions, and
(b) tilted. It is visible, that more extensive channeling will occur for the (0 0 1) planes
which are in Bragg conditions with the incident beam. ........................................... 34
Figure 9: A section of the Al-Cu phase diagram in the temperature range of 300-400°C. The
labeled regions referred to in the EDS results are marked as I, II and III. ................ 37
Figure 10: SE SEM micrograph of copper free air balls produced in air. The scale bar is
equivalent for all three micrographs. The asymmetrical FABs’ shape is visible. ...... 37
Figure 11: SE FIB micrograph of a representative copper FAB produced in a reducing
atmosphere. The FAB is completely symmetric with no visible flaws. ...................... 38
Figure 12: SE FIB micrographs of representative FABs formed (a) in air and (b) in a protective
environment. The heat-affected zone of the FABs produced in air is as twice as large
as that of FABs produced in a reducing environment. The scale bar is equivalent for
both micrographs. .................................................................................................. 39
Figure 13: SE SEM micrograph of as-bonded Cu wires. The arrow indicates aluminum forced
out from below the Cu ball during the bonding process. ......................................... 39
Figure 14: (a) SE SEM micrograph and (b) SE FIB micrograph of the cross-section of an asbonded wire-bond. ................................................................................................. 40
Figure 15: SE SEM micrograph taken from the periphery of the as-bonded Al-Cu wire-bond. ... 40
Figure 16: SE SEM micrograph taken from the center of the as-bonded Al-Cu wire-bond.
Smaller copper grains at the very center of the interface can be detected. ............. 41
Figure 17: HAADF-STEM micrograph acquired from the periphery of an as-bonded specimen.
The STEM-EDS results of the annotated regions are indicated in Figure 9............... 42
Figure 18: BF STEM micrograph taken from the center of an as-bonded Al-Cu wire-bond. The
STEM-EDS results of the annotated regions are indicated in Figure 9. .................... 42
Figure 19: SE SEM micrograph of the copper wires heat-treated for 2 hours at 175ºC in argon.
Partial oxidation of the wires is visible. ................................................................... 43
Figure 20: SE SEM micrograph of a cross-section of a copper wire-bond heat-treated for 2
hours at 175ºC in argon. ......................................................................................... 44
Figure 21: SE HRSEM micrograph taken from the periphery of a copper wire-bond heat-treated
for 2 hours at 175ºC in argon. Discontinuous non-uniform intermetallics are
detected. ............................................................................................................... 44
Figure 22: SE HRSEM micrograph taken from the periphery of a copper wire-bond heat-treated
for 2 hours at 175ºC in air. A discontinuous non-uniform intermetallic layer is
detected. ............................................................................................................... 45
V
Figure 23: SE HRSEM micrograph taken from the central region of a copper wire-bond heattreated for 2 hours at 175ºC in argon. A continuous intermetallic layer is visible. .... 45
Figure 24: SE HRSEM micrographs acquired from (a) the center and (b) the periphery of a
copper wire-bond heat-treated for 2 hours at 175ºC in argon. The difference in the
intermetallic morphology is visible. The scale bar is equivalent for both micrographs.
.............................................................................................................................. 46
Figure 25: HAADF-STEM micrograph of a peripheral region of a wire-bond heat-treated at
175ºC for 2 hours in argon. The STEM-EDS results of the annotated regions are
indicated in Figure 9. .............................................................................................. 46
Figure 26: BF STEM micrograph of a central region of a wire-bond heat-treated at 175ºC for 2
hours in argon. The insets are SAD patterns from the indicated grains. The STEMEDS results of the annotated regions are indicated in Figure 9. .............................. 47
Figure 27: BF STEM micrograph of a peripheral region of a wire-bond heat treated at 175ºC for
2 hours in air. The insets are SAD patterns from the indicated grains. The STEM-EDS
results of the annotated regions are indicated in Figure 9. The scale bar is equivalent
for both micrographs. ............................................................................................ 48
Figure 28: BF STEM micrograph of a central region of a wire-bond heat-treated at 175ºC for 2
hours in air. The insets are SAD patterns from the indicated grains. The STEM-EDS
results of the annotated regions are indicated in Figure 9. ...................................... 48
Figure 29: SE SEM micrograph taken from a cross-section of a copper wire-bond heat-treated
for 24 hours at 175ºC in argon. Non-uniform thickness of aluminum metallization is
detected. ............................................................................................................... 49
Figure 30: SE SEM micrographs of (a) the center and (b) the periphery of a copper wire-bond
heat-treated for 24 hours at 175ºC in argon. A discontinuous intermetallic layer is
detected. The scale bar is equivalent for the both micrographs. ............................. 49
Figure 31: BF STEM micrograph of the peripheral region of a wire-bond heat-treated at 175ºC
for 24 hours in argon. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9. .................... 50
Figure 32: BF STEM micrographs of the central region of a wire-bond heat treated at 175ºC for
24 hours in argon. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9. The scale bar
is equivalent for both micrographs. ........................................................................ 51
Figure 33: BF STEM micrograph of the peripheral region of a wire-bond heat treated at 175ºC
for 24 hours in air. The insets are SAD patterns from the indicated regions. The
STEM-EDS results of the annotated regions are indicated in Figure 9. .................... 51
Figure 34: BF STEM micrograph of the central region of a wire-bond heat treated at 175ºC for
24 hours in air. The insets are SAD patterns from the indicated grains. The STEMEDS results of the annotated regions are indicated in Figure 9. .............................. 52
Figure 35: SE SEM micrograph of the periphery of a copper wire-bond heat-treated for 96 hours
at 175ºC in argon. A discontinuous intermetallic layer is detected. Voids at the Al-Cu
interface can also be seen...................................................................................... 53
Figure 36: SE SEM micrograph of the center of a copper wire-bond heat-treated for 96 hours at
175ºC in argon. More continuous and uniform intermetallics are detected. ............. 53
Figure 37: SE SEM micrograph of the periphery region of copper wire-bond heat-treated for 96
hours at 175ºC in argon. A discontinuous void-line between the intermetallics and
copper ball is visible............................................................................................... 53
Figure 38: SE HRSEM micrograph of the central region of a copper wire-bond heat-treated for
96 hours at 175ºC in argon. More continuous and uniform intermetallics than the
intermetallics found at the bond periphery are visible. Small voids between the
copper ball and intermetallic layer are seen. .......................................................... 54
Figure 39: SE SEM micrographs taken from (a) peripheral and (b) central regions of a copper
wire-bond heat-treated for 96 hours at 175ºC in air. A discontinuous intermetallic
layer is visible. A void line at the bond periphery can also be detected. .................. 55
VI
Figure 40: BF STEM micrograph of the peripheral region of a wire-bond heat treated at 175ºC
for 96 hours in argon. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9. Two
discontinuous intermetallic layers of different contrasts and shapes can be
distinguished. ........................................................................................................ 56
Figure 41: BF STEM micrograph of the central region of a wire-bond heat treated at 175ºC for
96 hours in argon. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9. A continuous
intermetallic layer can be distinguished. ................................................................ 56
Figure 42: BF STEM micrograph of a peripheral region of a wire-bond heat-treated at 175ºC for
96 hours in air. The insets are SAD patterns from the indicated grains. The STEMEDS results of the annotated regions are indicated in Figure 9. A discontinuous
intermetallic layer and voids can be found on the Al-Cu interface. .......................... 57
Figure 43: BF STEM micrograph of the central region of a wire-bond heat treated at 175ºC for
96 hours in air. The insets are SAD patterns from the indicated grains. The STEMEDS results of the annotated regions are indicated in Figure 9. An oxide layer is
found between the copper ball and intermetallics................................................... 58
Figure 44: SE SEM micrographs of copper wires heat-treated for 200 hours at 175°C (a) in
argon and (b) in air. Deformation of the copper wire heat-treated in air is detected.59
Figure 45: SE SEM micrographs taken from (a) the center and (b) periphery of a copper wirebond heat-treated for 200 hours at 175ºC in argon. The scale bar is equivalent for
both micrographs. A discontinuous intermetallic layer is detected.ected. .............. 59
Figure 46: SE SEM micrograph taken from the periphery of a copper wire-bond heat-treated for
200 hours at 175ºC in argon. Two intermetallic layers are detected. ....................... 60
Figure 47: SE SEM micrograph taken from the center of a copper wire-bond heat-treated for
200 hours at 175ºC in argon. Two intermetallic layers of different contrast are
detected. ............................................................................................................... 60
Figure 48: SE SEM micrographs taken from (a) the center and (b) the periphery of a copper
wire-bond heat-treated for 200 hours at 175ºC in air. Voids at the Cu-IM interface and
between two intermetallic layers are detected. The scale bar is equivalent for both
micrographs. ......................................................................................................... 61
Figure 49: A BF STEM micrograph taken from the periphery of a wire-bond heat treated at
175ºC for 200 hours in argon. The insets are SAD patterns from the indicated grains.
The STEM-EDS results of the annotated regions are indicated in Figure 9............... 62
Figure 50: A BF STEM micrograph taken from the center of a wire-bond heat treated at 175ºC
for 200 hours in argon. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9. .................... 62
Figure 51: A BF STEM micrograph taken from the periphery of a wire-bond heat treated at
175ºC for 200 hours in air. The insets are SAD patterns from the indicated grains.
The STEM-EDS results of the annotated regions are indicated in Figure 9. Two
intermetallic layers with different contrasts can be distinguished. .......................... 63
Figure 52: A BF STEM micrograph taken from the periphery of a wire-bond heat treated at
175ºC for 200 hours in air. The insets are SAD patterns from the indicated grains.
The STEM-EDS results of the annotated regions are indicated in Figure 9. Nonuniform intermetallic grains are visible. .................................................................. 63
Figure 53: A BF STEM micrograph taken from the center of a wire-bond heat treated at 175ºC
for 200 hours in air. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9. .................... 64
Figure 54: Schematic representation of the intermetallic morphology and evolution at Al-Cu
wire-bonds as a function of a heat-treatment time. ................................................. 65
Figure 55: Finite element analysis of mechanical stress applied on the copper ball during the
bonding process. Maximum stress is noted in red. Dashed lines mark the location of
the capillary. .......................................................................................................... 67
VII
Figure 56: Schematic representation of an experimental STEM-EDS measurement setup. When
the specimen is tilted towards the EDS detector, the region from which the
information is received is not symmetrical around the incident electron beam. ....... 70
Figure 57: BF STEM micrograph of Al-Cu interface obtained from a reference specimen
prepared by PVD. The TEM specimen was prepared by FIB. No intermetallics were
found at the Al-Cu interface. ................................................................................... 71
Figure 58: The square of intermetallic size as a function of heat-treatment time. The purple line
is a linear fit of the represented data. ..................................................................... 75
Figure 59: (a) BF TEM and (b) EFTEM micrographs taken from a central region of as-bonded AlCu wire-bond.......................................................................................................... 82
LIST OF TABLES
Table 1: Selected properties of Al, Au and Cu metals [,]. ......................................................... 10
Table 2: Selected characteristics of Al-Cu intermetallic compounds [5]. ................................. 13
Table 3: Energies of formation (ΔG) for selected Cu and Al oxides........................................... 20
Table 4: Atomic packing factors of Al, Cu and intermetallic phases found at the Al-Cu interface.
.............................................................................................................................. 74
Table 5: Crystallographic data of Al-Cu intermetallics, used for SAD patterns identification []. 81
Table 6: Values of points used for the computation of limit if detection of Al in Cu for the current
experimental setup................................................................................................. 82
Table 7: Computation of energy of formation of the Al2Cu phase.phase. .................................. 83
VIII
ABSTRACT
Wire-bonding is a main interconnection process in the packaging industry. Gold wires
are the most commonly used for wire-bonding, while copper is being evaluated as a
gold substitute. Very little reliable analytical data regarding the time-dependent
behavior of Al-Cu wire-bonds is available. This work focuses on the Al-Cu wirebonding process, and the time-dependent microstructural behavior of Al-Cu wirebonds.
Prior to the bonding process, the tip of the copper wire is melted by an electrical spark
to form a metal ball. In this research, copper wires melted in air and wires melted in a
protective environment were studied. It was found that the protective environment
enables acquisition of symmetrical and flawless wire-balls. It was also found that
wires melted in the protective environment exhibit a shorter region of copper grain
growth zone (in comparison with copper wires melted in air), which improves the
mechanical reliability of the Al-Cu wire-bonds. The investigations of the copper wire
grain size were conducted using secondary electrons (SE) induced by focused ion
beam (FIB).
To understand the time-dependant behavior of Al-Cu wire-bonds, as-bonded and heattreated bonds were investigated. In order to simulate encapsulation in inert versus
corrosive adhesives, Al-Cu wire-bonds heat-treated at 175ºC up to 200 hours in air
were compared with those heat-treated in argon. Detailed information regarding the
morphology and composition of the Al-Cu interface was obtained from advanced
microscopy techniques (scanning and transmission electron microscopy). FIB enabled
relatively artifact-free and site-specific specimen preparation.
Discontinuous and non-uniform intermetallics were found in as-bonded and heattreated specimens. The intermetallics were found to form in regions of high localized
stress introduced into the interface during the wire-bonding process. The main
intermetallic phase found at the Al-Cu interface was Al2Cu. This phase was found to
grow via bulk solid state diffusion. In specimens heat-treated for 96 and 200 hours,
Al4Cu9 was found. Voids found at the Al-Cu interface of specimens heat-treated for
1
96 and 200 hours were probably created via volume reduction during growth of the
Al4Cu9 phase.
Void formation at the heat-treated Al-Cu interfaces was not found to be the source of
failure. The failure of Al-Cu wire-bonds heat-treated for 200 hours in air occurred via
creep of the copper wires.
2
LIST OF SYMBOLS AND ACRONYMS
x
Intermetallic thickness
D
Diffusion coefficient
t
Diffusion time
Q
Activation energy for intermetallic growth
kb
Boltzmann constant
ΔG
Energy of formation
4N
99.99% purity
C
Atomic concentration
I
Signal intensity [counts]
σ
Standard deviation
OM
Optical microscopy
SEM
Scanning electron microscopy
EDS
Energy dispersive spectroscopy
XRD
X-ray diffraction
WB
Wire bond
TEM
Transmission electron microscopy
EF-TEM
Energy filtered transmission electron microscopy
FIB
Focused ion beam
EFO
Electronic flame-off
HAZ
Heat affected zone
FAB
Free air ball
RBS
Rutherford backscattering spectroscopy
DSC
Differential scanning calorimetry
SIMS
Secondary ion mass spectrometry
CTE
Coefficient of thermal expansion
SE
Secondary electron
STEM
Scanning transmission electron microscopy
FEG
Field emission gun
IM
Intermetallics
SAD
Selected area diffraction
HAADF
High angle annular dark field
BF
Bright field
HM
High magnification
UHR
Ultra High Resolution
APF
Atomic Packing Factor
PVD
Physical Vapor Deposition
TIA
Tecnai Instrumental Analysis
3
CHAPTER 1 : INTRODUCTION
Wire-bonding is one of the main technologies used for electronic chip
interconnections. Highly conductive metals (mainly aluminum and gold) are
connected to the conductive pads (usually made of copper, silver or aluminum) in
order to make an electronic connection of semiconductor devices to the macroscopic
circuits [1]. In the process of bonding the wire to the metallization pad and to the lead,
a reliable bond with good adhesion is needed. This is achieved by applying ultrasonic
and thermal energies that assist the bonding process.
The mechanical properties of joins between two different metals are strongly affected
by the presence of intermetallic phases. These phases can ensure direct connection
and a strong uniform bond between the metals. However, intermetallic phases may
also have lower electrical and thermal conductivity, thus damaging the chip-package
electrical connection [2]. Moreover, the intermetallic phases are usually more brittle
than the initial metals, thus limiting the connection’s mechanical reliability. In
addition, the intermetallic phases differ in density from the initial metals, which can
introduce voids into the intermetallic region during intermetallic formation.
Recently, due to its superior thermal and electrical conductivity and higher
metallurgical stability, usage of copper in advanced wire-bonding processes is of
interest. In order to incorporate the copper wire into the packaging industry, the
reliability of the Al-Cu wire-bond during the service period should be evaluated and
reliable techniques for quality assurance of the bond should be established. Among
the many methods developed by the industry to test the reliability of copper wirebonding are: microscopic evaluation of the damage introduced into the Al
metallization and Si substrate during the bonding process; mechanical tests for the
evaluation of the adhesion strength of the bond; calculation of an actual bonded area
of the wire-bond and examination of the intermetallic Al-Cu phase formation after
heat treatment for different periods of time; and comparison of all of the above with
Al-Au phase formation. The morphology of the Al-Cu interface is usually examined
by means of optical microscopy (OM) and scanning electron microscopy (SEM),
4
while the intermetallic composition is evaluated via energy dispersive spectroscopy
(EDS) embedded on SEM and x-ray diffraction (XRD).
The resolution achieved by the characterization techniques used in the industry is
insufficient for the detailed morphological and compositional analysis of the thin
intermetallic layers found at the Al-Cu interface. Therefore, in this research more
detailed investigation of the Al-Cu interface will be conducted.
The goal of this research is to obtain reliable information regarding time-dependent
Al-Cu intermetallic growth at Al-Cu wire-bonds. In order to conduct detailed
investigation of the wire bond (WB) interface, it was studied by transmission electron
microscopy (TEM), TEM-EDS and energy filtered (EF) TEM, while the site specific
specimen preparation was carried out using dual-beam focused ion beam (FIB). The
tunneling effect of ion-induced secondary electrons in FIB was used for qualitative
grain size measurements of the copper wires before and after the bonding process.
The use of advanced microscopy techniques enabled a better understanding of the
influence of the wire-bonding process variables, and the elevated temperatures
expected during device operation on the microstructural and compositional evolution
of the Al-Cu wire-bonded interface.
5
CHAPTER 2 : LITERATURE SURVEY
2.1. The Wire-Bonding Process
Wire-bonding is one of the main technologies used for electronic chip
interconnections. During the process, combined mechanical, ultrasonic and thermal
energies are applied. Figure 1 schematically illustrates a simplified wire-bonding
process.
Figure 1: Schematic representation of the wire-bonding process.
The wire-bonding process can be described by the following steps:
1. The wire is fed into the capillary, and an electric spark transferred via
electronic flame-off (EFO) wand melts the wire. Due to surface energy, a
spherical ball is formed at the end the wire.
2. The ball is positioned against the bottom of the capillary
3. The capillary is lowered to the bond pad, and the ball is pressed against the
pad by a gentle impact force. The wire ball is compressed against the pad
metallization. The wire-pad interface is heated, ultrasonic energy is applied
through the capillary, and the bond is formed.
4. The capillary is raised, leaving the ball joined to the surface.
6
5. The wire loop is formed, and the wire moves with the capillary towards the
second bond position. In the wire-ball bonding process, due to the symmetry
of the bond, there are no restrictions on the location of the second bond.
6. The capillary is lowered to make the second bond, also known as the “stitch
bond”. In a similar way to the first bond, mechanical, thermal and ultrasonic
energies are applied.
7. The capillary is raised and a wire clamp above the capillary tool (not shown)
pulls and breaks the capillary free. The tool rises up, the wire is lowered, and
the bonding cycle is ready to be repeated [1,3].
The applied mechanical stress couples the wire and the metallization together,
insuring a maximum connection area between the two metals. The thermal energy
applied to the metallic pad softens the metals in order to achieve a better join and as a
result, a better bond between the two metals. The ultrasonic energy is introduced into
the bonding process in order to increase the total amount of energy transferred to the
bonded wire, thus reducing the flow stress, without noticeable increase in the system
temperature [4,5]. Ramsey and Alfaro state that during ultrasonic activation, the area
around dislocations is thermosonically heated allowing dislocation movement. In
addition, the free energy transferred to the dislocations enhances the dislocations’
activity [4]. It is agreed that ultrasonic activation increases to some extent the
temperature of the bonded interface. However, to the best of our knowledge no
quantified or measured data regarding this temperature increase was published.
Application of the ultrasonic force during the wire-bonding process also enables to
clean the bonded surfaces from oxides, thus providing a more intimate contact
between the two bonded metals [6,7].
One of the initial stages of the wire-bonding process is the melting of a wire by an
electrical spark (Figure 1 step 1). During melting, as a result of the heat transfer from
the ball tip, the part of the wire located above the formed ball is exposed to an
elevated temperature. This part of the wire goes through annealing and is referred to
as a heat-affected zone (HAZ). This zone (see Figure 2) has a larger grain size and
lower hardness in comparison to the rest of the wire. Thus the HAZ is weaker than the
7
rest of the wire and is prone to failure [1]. A shorter HAZ with a smaller grain size
can increase the bond reliability.
Figure 2: Schematic representation of the microstructure of the wire after FAB
formation [1].
2.2. Copper Wires in the Wire-Bonding Process
While aluminum wires are used in wedge-bonding processes, gold wires are the
commonly used and well characterized media for chip interconnection via wirebonding. Recently, the use of copper wires has become a point of interest to the
industry due to some advantages over Au wires:
1. Copper wires have a higher thermal conductivity, lower electrical resistance
and higher pull strength (see Table 1) which allows reduction of the wire
diameter. This consequently allows reduction of the pad size [8].
2. The stiffness of copper allows to work with a lower loop height [9] and
reduces wire sag [10], which is preferable for multi-tiered wire-bonded
structures, thus also allowing pad size reduction.
3. Systems with copper wires also have improved reliability due to fewer failures
that occur at the copper wire neck [8,11]. Singh et al. states that the reason for
fewer wire neck failures in Cu is due to the increased strength of the copper
HAZ [9].
8
4. Al-Cu intermetallics are of higher metallurgical stability than those in the AlAu system, i.e. Al-Cu intermetallic growth (at elevated temperatures) is much
slower than the growth of Al-Au intermetallics. This phenomena is sometimes
explained by the difference in atomic radii between Al and Cu (10.5%
difference, compared to 0.3% for the Al-Au system) [7,11,12,13].
5. The electrical performance of the Al-Cu system is better than that of the AlAu system due to the higher electrical conductivity of copper aluminides. The
thermal expansion mismatch between the pure metals and their intermetallics
is lower in the Al-Cu system, thus reducing the danger of void formation due
to temperature perturbations and increasing the system reliability.
6. Due to increased Cu hardness, the wire sweep during plastic encapsulation is
reduced [10,11]. Moreover, copper wire bonds are found to be more resistant
to operator mishandling during transportation [10].
Despite all the benefits of the Al-Cu wire-bonding system, copper wires are not as
commonly used as gold wires, and this is due to the following drawbacks of the Al-Cu
wire-bonded system:
1. Copper has a higher hardness than gold, and the use of copper wires might
cause damage to the underlying Al metallization or even the Si layer. The
damage might appear as micro cracks in the underlying layers, thus increasing
the electrical resistivity of the bond, or as complete wire disconnection due to
micro crack evolution, thus affecting the bond reliability. Copper is also a
readily-oxidized metal, which can cause unsymmetrical ball formation, and
consequentially stress concentration at the wire-bond. Moreover, oxidation
hardens the copper ball, increasing the possible damage to the aluminum
metallization and the Si layer [8,12].
2. In order to avoid copper oxidation and the damage that might be caused to the
Al metallization and Si pad, modified equipment (such as a reducing gas
atmosphere and improved capillaries) and bonding methodology (such as
variation of impact strength applied during the bonding process) must be used
[8,14].
9
3. Due to the high copper hardness, the Al-Cu interconnection might be attacked
by stress corrosion during the life-span of the device. Moreover, chlorine and
bromine outgasing, which can originate from polymer encapsulate
decomposition, catalyzes Cu wire corrosion, thus decreasing the bond
reliability [5,15]. Therefore, designated encapsulates should be used for
copper wire-bonding.
Table 1: Selected properties of Al, Au and Cu metals [16,17].
Element
Cu
Property
Atomic radius (Å)
Lattice constant (Å)
Melting point (ºC )
Density (g/cm3)
Electrical resistivity (mΩcm)
Thermal conductivity (W/mK)
Coefficient of linear expansion (10-6/K)
Tensile strength (MPa)
Young’s modulus (GPa)
Elongation (%)
1.28
3.6147
1083.45
8.93
1.69
394
16.6
240
136
15.0
Au
Al
1.44
4.0785
1063.15
19.32
2.3
293
14.2
230
88
4.5
1.43
4.050
660.4
2.70
2.67
238
23.5
2.3. The Al-Cu Phase Diagram
As noted above, intermetallic phases formed during the bonding of two different
metals can strongly affect the join between metals. On one hand, it is agreed that
intermetallics indicate direct contact between the two metals and assist in achieving a
strong uniform bond between them [2,3]. On the other hand, Al-Cu intermetallics are
more brittle and have a lower thermal and electrical conductivity than pure copper or
aluminum, which can decrease the electrical and mechanical reliability of the bond. In
order to achieve a better understanding of the Al-Cu bond in the wire-bonding
process, knowledge of the intermetallic phases in the given system is needed.
The Al-Cu phase diagram, as given by Murray, is presented in Figure 3 [18]. It can be
seen that at temperatures below 300ºC, six near-stoichiometric phases exist.
10
Figure 3: The aluminum-copper binary phase diagram [18].
The possible Al-Cu intermetallic phases according to the phase diagram at 300°C are
as follows: θ, which is of the Al2Cu cubic structure, η2 which is based on monoclinic
AlCu, ξ2 which is based on monoclinic Al3Cu4, δ which is based on the Al2Cu3 cubic
structure (noted by Wulff et al. [5] as trigonal), γ1 that is based on cubic Al4Cu9 phase
(sometimes noted as AlCu2 phase), and the rarely noted α2 cubic phase (that is
possibly based on AlCu7) [5,18,19,20].
2.4. Some Properties of Intermetallic Al-Cu Phases
The exact influence of the intermetallic phases on the Al-Cu bond integrity is
unknown. It is agreed that Al-Cu intermetallic phases are of higher resistivity than the
pure metals. However, Wallach et al. found that at very thin intermetallic widths (2μm
and lower), the resistivity of the Al-Cu couple is not decreased due to intermetallic
formation [20]. Also, no changes in current paths were detected at small intermetallic
dimensions. At higher intermetallic widths the current path changes, resulting in
increased resistivity.
11
Macchioni et al. stated that the γ2 and δ phases have high resistivity, while the ξ2, η2
and θ phases have relatively low resistivity (similar to pure Cu) [19]. However, Lee et
al. noted that Al-Cu intermetallics (AlCu- η2 and Al2Cu- θ) have a high electricallyresistant nature, since these compounds have a nonmetallic covalent bond. Lee et al.
also showed that the electrical resistivity increased with increased intermetallic
dimension such that the measured electrical resistivity of the total intermetallic
compound layer represented a much higher value compared with the sum of resistivity
values of the individual phases. This may be explained by microstructural factors of
the intermetallic layer such as porosity, cracks, grain size changes, and increased
dislocation density in the diffusion zone, all of which might reduce the effective
contact area [21].
The mechanical properties of Al-Cu intermetallics also differ from those of the pure
metals. Braunovic et al. reported that if the total thickness of the intermetallic
compound (at a friction-welded Al-Cu couple) was less than 2μm, the Al-Cu joins
maintained their mechanical properties [2]. However, for samples with thicker
intermetallics, the tensile strength of the Al-Cu couple decreases. For example,
Abbasi et al. found that when the intermetallic phase thickness of a welded Al-Cu
couple exceeds 2.5μm the metal bond strength is sharply reduced and failure is of a
brittle nature [22]. These results can also be explained by the change in the
microstructural factors of the intermetallic layer. Voropai et al. also noted that
intermetallic Al-Cu phases have a lower tensile strength than the initial metals, and
the failure of the joins occur in the intermetallics [23].
Microhardness can be proportional to brittelness. The microhardness of the Al-Cu
intermetallics is found to be higher than that of the pure metals [2]. Peng et al.
compared the microhardness of various Al-Cu intermetallics and concluded that
copper-rich phases are harder than aluminum-rich phases [24]. Table 2 summarizes
select properties of the major Al-Cu intermetallic compounds [5].
12
Table 2: Selected characteristics of Al-Cu intermetallic compounds [5].
Phase
Crystal
structure
Al
cubic
θ
Al2Cu
tetragonal
η2
AlCu
monoclinic
ξ2
Al3Cu4
monoclinic
δ
Al2Cu3
trigonal
γ1
Al4Cu9
cubic
Cu
cubic
Notation
Color
silverish
whiteyellow
grey
whiteyellow
white-gray
white-light
yellow
red
Resistivity ρ
(mΩ-cm)
Termal Expantion
Coefficient (CTE)
(ppm/ºC)
2.4
2.35
7.0-8.0
1.61
11.4
1.19
12.2
1.61
13.4
1.51
14.2-17.3
1.76
2.0
1.73
2.5. Phase Formation in Bulk Al-Cu Welded Joins
Braunovic et al. found the following sequence at the Al-Cu interface in bulk frictionwelded samples: Cu-AlCu2-Al2Cu3-Al3Cu4-AlCu-Al2Cu-Al [2]. According to
Braunovic et al., the growth rate of intermetallic phases can be determined by
measuring the thicknesses of the interdiffusion layers at different intervals. They also
stated that the rate of formation of intermetallic phases can only be described by
considering two separate temperature regions: 200-300ºC and 360-525ºC, due to
different growth mechanisms: 1) diffusion via structural defects such as grain
boundaries or dislocations (at lower temperatures) and 2) bulk diffusion (at higher
temperatures). The results of Braunovic’s study showed that after annealing at 100ºC
the growth of the intermetallic layers is negligible. However, after as little as one hour
at 150ºC, the thickness of the layer grows beyond 2 μm, thus enhancing the area’s
resistivity (according to Wallach et al.[2]).
Lainer et al. [25] investigated diffusion formation of Al-Cu intermetallics at Al-Cu
hot-welded joins with Al excess by means of XRD. They found that diffusion
between Cu and Al begins at 150ºC by formation of Al2Cu. At 350ºC, in addition to
the Al2Cu phase, Al4Cu9 phase was found. The AlCu intermediate layer appeared at
400ºC. In samples annealed at 350ºC for 2 hours the Al3Cu4 phase was detected.
13
The following phase sequence was observed by Abbasi et al. in cold roll-welded and
annealed (250ºC) Al-Cu specimens: Cu-Al2Cu3-Al3Cu4-saturated solution of 913%Al-AlCu-Al2Cu-Al [22]. No explanation of the non-monotonic concentration
change was suggested. Abbasi et al. also noted that the intermetallic growth rate
constant for roll welding is lower than for friction welding. This is probably due to the
lack of heat input during cold roll welding as apposed to the heat generated during
friction welding.
The following phase sequence was observed by Lee et al. at friction welded and
annealed (100-500ºC) Al-Cu specimens: Al-Al2Cu-AlCu-Cu [21]. Lee et al. also
showed that the formation of different intermetallics at the Al-Cu couple depends on
the annealing time and annealing temperature of the welded couple.
Peng et al. studied Al-Cu laminates prepared by roll bonding at 430ºC. The samples
were aged at 300-500ºC for different periods of time. Mainly Al2Cu and Al4Cu9
intermetallic phases were found, while the Al4Cu9 phase was dominant at higher
temperatures and longer annealing times. It was also found that at high temperatures
and long annealing times the Kirkendall effect becomes a significant factor in
determining the intermetallic bond strength [24].
Hannech et al. found that in welded (600ºC) Al-Cu couples the morphology and phase
sequence were: Cu-α phase- Pores (k)- α2 phase- Al2Cu3(δ phase)- Al3Cu4 (ζ2 phase)Al. The pores were explained by the Kirkendall effect [26].
Wallach et al. described the formation of AlCu, Al2Cu, Al3Cu4, Al2Cu and Al4Cu9
phases in cold pressure-welded joins after annealing at 300, 400 and 500ºC [27]. As
opposed to Peng and Hannech et al., Wallach reported that no deterioration due to
pore formation was observed at the Al-Cu joins.
It can be seen that the identified intermetallic phase formation depends on the
preparation method as well as the annealing time and temperature (and perhaps the
characterization method). Not all of the intermetallic phases appear in all of the
14
experiments noted above. Moreover, Braunovics’ conclusion that intermetallic
formation occurs due to different diffusion mechanisms increases the probability of
different routes of intermetallic formation in thin Al-Cu couples. The nonhomogeneous concentration changes during the Al-Cu intermetallic phase formation
can be explained by diffusion-driven processes or by a decomposition of lowtemperature intermetallic phases at higher temperatures.
2.6. Phase Formation in Al-Cu Thin Films
Gershinskii et al. investigated phase growth kinetics in Al-Cu thin films using
electrochemical anode etching and confirmed his findings by TEM. Using
electrochemical anode etching, the Al2Cu and AlCu phases were found in specimens
annealed at 150-250ºC (annealing time varied from 2 to 120 min). In addition, the
Al4Cu9 phase was found by TEM. Gershinskii found that below 150ºC the Al2Cu
phase grows first, and the AlCu phase appears after longer annealing. At higher
temperatures, AlCu decomposes to Al2Cu and free copper, which grows with a
pearlite-like microstructure. The limiting stage of intermetallic growth in specimens
with Cu excess was found to be diffusion of the initial material through the forming
layer. Activation energies for intermetallic formation were found to be 1.36eV and
1.08eV for Al2Cu and AlCu formation, respectively [28].
Campisano et al. investigated Al-Cu thin film bi-layers by Rutherford backscattering
spectroscopy (RBS) and XRD. The specimens were annealed at 160-200ºC for
different times (up to 4.5h) after which the Al2Cu phase was found. The Al4Cu9 phase
was also found, but only at annealing temperatures above 200ºC and for long
annealing times. The growth rate of the Al4Cu9 phase was found to be lower than that
of the Al2Cu phase [29]. Campisano et al. also stated that the formation of Al2Cu
occurs by a diffusion-limiting process. The activation energy of the AlCu2 phase
formation was found to be 1.02eV.
Rajan and Wallach used TEM to investigate phase formation in Al-Cu thin film
couples. The specimens, which contained ~53nm thick Al films and ~24nm thick Cu
15
films, were annealed for 15, 30 and 90 min. at 500ºC. Out of 5 possible main
equilibrium phases only the Al2Cu and Al4Cu9 phases were found [30]. This did not
match studies carried out by Gershinskii and Campisano, probably due to the higher
annealing temperatures that Rajan and Wallach used. The following explanations as to
the absence of 3 more intermetallics were provided by the authors: 1) the presence of
intermetallic phases depends on relative amounts of the initial metals. This
explanation does not agree with the results presented in their article and with other
works cited here, since at Al-rich couples the Al4Cu9 phase was found. 2) Phase
nucleation is kinetically controlled, since high diffusivity can be expected in thin
films. 3) The compositions of intermetallics that are formed move from both parent
metal compositions, so that the intermetallics that are in the center of the equilibrium
diagram are the last to form. The last explanation, however, disregards the possibility
of reactions on the intermetallic interfaces, the stress/strain factors, the fact that
kinetic factors might be important, and the influence of the heat of formation of the
intermetallics (the intermetallics with lower heats of formation are expected to be the
first to nucleate).
Vanderberg and Hamm studied phase formation in Al-Cu thin film couples by X-ray
analysis during in-situ annealing in the temperature range of 35-500ºC [31]. The
authors presented a schematic diagram of the temperature-dependent phase formation
for Al- and Cu- rich regions in Al-Cu thin films, and a phase diagram of the
crystalline phases observed by X-ray analysis in the temperature range of 35-400ºC.
The schematic diagram of phase formation as described by Vanderberg and Hamm is
given in Figure 4.
16
Figure 4: Schematic diagram of in-situ phase formation in thin Al-Cu film couples as
a function of annealing temperatures [31].
It can be seen that for temperatures up to 170ºC, the phase formation is identical for
the Al-rich and Cu-rich regions. The phases formed are the Al2Cu phase (at 140ºC),
and β-AlCu3 (which appears at 170ºC together with Al2Cu). Above 200ºC AlCu,
Al2Cu and β-AlCu3 were found for both Al- and Cu- rich regions. Also, in the Cu-rich
region the Al4Cu9 phase was detected. At higher temperatures, the AlCu and β-AlCu3
phases prevailed in the Al-rich regions, when the final phases (at 350ºC) in the Al-rich
regions were found to be AlCu and Al2Cu. In the Cu-rich regions at temperatures
above 250ºC the Al4Cu9 phase prevailed, which was also the final phase (at 350ºC).
17
In a later work Hamm and Vanderberg studied in-situ intermetallic phase formation in
Al-Cu thin films annealed in the temperature range of 157-220ºC by RBS and XRD
[32]. The specimens were heated at a rate of 2ºC /min. Only two major phases were
found: Al2Cu which grew with a planar interface, and Al4Cu9 which did not grow as a
planar layer but rather as isolated grains. It was also found that these two phases grow
simultaneously. However, the growth rate of Al2Cu was found to be much higher than
that of Al4Cu9. This was confirmed by RBS and by calculated growth parameters of
the phases. The activation energy for the Al2Cu phase was found to be 1.21eV (via
XRD) and 1.24eV (via RBS). For Al4Cu9 the activation energy was found to be
1.59eV. Intermetallic growth was found to be controlled by volume diffusion.
Comparing their results with previous works on the subject [28, 29], Hamm and
Vanderberg also noted that the composition of the phases depends on the rate of
annealing temperature increase.
Hentzell et al. also studied interdiffusion in Al-Cu thin film couples by means of XRD
and RBS [33]. The samples were heat-treated at 160-300ºC. In Al-rich couples, the
Al2Cu phase was the main phase observed during all stages of annealing. In Cu-rich
couples, the sequence of phase formation at each annealing temperature was as
follows: Al2Cu-AlCu-Al4Cu9. The activation energy of Al2Cu formation was
calculated to be 1.31eV, which does not agree with previous data obtained for thin
films, but is in good agreement with the data obtained for bulk couples [28, 29, 33].
Shearer et al. investigated interfacial reactions in Al-Cu thin couples by means of
contact resistance. They found that intermetallic reactions occurred both by grain
boundary diffusion (of Al in Cu and vice versa) and by bulk interdiffusion. The
samples were heat-treated at temperatures above 70ºC [34].
Jiang et al. investigated reactions of Al-Cu multilayered thin films by differential
scanning calorimetry (DSC) and TEM. The multilayer film investigated by DSC
consisted of 15 Al-Cu bilayers, with an excess of copper. The TEM samples were
prepared as a tri-layer of Cu-Al-Cu. The samples were heated at a rate of 160K/min
18
up to 347ºC. The sequence of phase formation was found to be: Cu-Al2Cu-Al4Cu9.
The activation energies for the Al2Cu phase and for the Al4Cu9 phase were calculated
to be 0.78eV and 0.83eV, respectively [35].
Usually, the dependence of the intermetallic thickness (x) on the heat-treatment time
(t) is in accordance with Equation 1. This enables the description of the temperature
dependence of the growth rate of Al-Cu intermetallics using the Arrhenius’ equation
(Equation 2),
x = Dt
(1)
Q
)
k bT
(2)
D = D0 exp(−
which also provides activation energies (noted as Q) for intermetallic formation. D
and D0 are the diffusion coefficients at the experiment temperature and at 0K
respectively and kb is Boltzmann’s constant.
The Al-Cu intermetallic formation activation energies calculated for welded joins
differ from those calculated for thin films. Moreover, different authors computed
different values of activation energies for similar thin film joins [28,29,33,35]. The
reason for the differences may lie in different annealing rates, the amounts of pure
metals, and the characterization methods used in the studies.
2.7. Influence of Oxygen on Al-Cu Intermetallics
Both copper and aluminum are readily-oxidized metals. Since the Al-Cu interfaces are
exposed to oxygen prior to chip encapsulation, copper and aluminum oxides might
form. Although one of the roles of ultrasonic activation during the wire-bonding
process is to force the oxidized regions away from the bond region, some oxidation
might remain on the Al-Cu interface. Oxidation at the bond interface can deteriorate
the electrical and thermal conductivity of the bond and reduce the mechanical
reliability of the connection.
19
Oxidation of aluminum and copper might also occur during the life-span of the wirebonded device. Thomas et al. studied surface oxidation of Al-Cu thin films using the
resistivity method and by XRD and SEM. The samples consisted of layers with equal
resistivity. Samples with and without heat treatment were investigated. The timedependent phase formation on top of the Al-Cu layers (capped with Cu) without heat
treatment was as follows: CuO on samples aged for one week; CuO and Al2O3 on
samples aged in air for one month. On samples annealed at 150ºC for one month the
following phases were found: Al2Cu3, Cu2O and Al2O. From optical microscopy it
was concluded that a capping layer of Al on Al-Cu thin films provides better
corrosion protection due to Al2O3 formation [36].
In order to understand the effect of oxidizing encapsulates on phase formation in AlCu wire-bonded systems, phase formation at the Al-Cu systems subjected to oxygen
should be studied. However, there is little data on Al-Cu thin film interaction in an
oxidizing environment. Cai et al. investigated Al-Cu-O films in order to achieve
optimization for p-type semiconductors and found that the possible combinations that
may coexist in these films are AlCuO2, AlCu2O4, Cu2O, CuO and Al2O3 [37]. Another
interesting work was carried out by Chang who, in his studies of intermetallic
formation in Al-Cu and Pt-Al-Cu thin film systems after annealing in forming gas and
in N2-O2, found that in thin Al-Cu couples, formation of the Al-Cu intermetallic is
inhibited by the presence of oxygen. For the case of a Pt layer covering the Al-Cu
couple, no oxygen inhibition of the intermetallic formation was observed [38].
Table 3 provides energies of formation for aluminum and copper oxides at room
temperature and at 227ºC [16].
Table 3: Energies of formation (ΔG) for selected Cu and Al oxides.
Compound
-ΔG (27ºC)
-ΔG (227ºC)
Al2O3
1584.0
1520.8
Cu2O
144.9
129.8
CuO
127.3
108.9
20
However, it is clear that aluminum and copper oxide formation depends not only on
the oxide formation energy, but also (mainly) on interdiffusion of the metals and
diffusion of oxygen into the metals.
2.8. Phase Formation in Al-Cu Wire Bonds
As noted above, the composition and activation energies of Al-Cu intermetallics
investigated for bulk samples and for thin Al-Cu layers are strongly dependant on the
process of join formation. Therefore, in order to fully understand intermetallic
formation processes during wire-bonding, investigations of the ball bonding process
itself needs to be considered.
It is commonly accepted, that no intermetallics are formed during the Al-Cu wirebonding process. Studies conducted on ultrasonic Al-Cu wire bonds repeatedly report
that no intermetallics are found at as-bonded Al-Cu wire-bonds [5,39,40,11].
Tomlinson et al. also did not find any intermetallic compounds in their investigation
of Al-Cu thermosonic ball-bonded interfaces which were thermally aged at 300ºC for
up to 1000 hours, prepared by “standard metallographic methods” and examined by
SEM [41].
Onuki et al. investigated intermetallic growth in copper ball bonds by dissolving the
Cu wires and identifying the remaining phases by micro X-ray diffraction. AlCu and
Al2Cu phases were found at the Al-Cu wire-bond interface annealed in air at 120200ºC. Empirical activation energy for the Al/Cu bond was found to be 1.26eV [12].
Kim et al. investigated Al-Cu intermetallic formation after annealing at 150-300ºC for
relatively long periods of time (up to 250 hours). The main intermetallic phase
identified was Al4Cu9. The activation energy obtained by Kim et al. for intermetallic
formation in the Al-Cu system was 1.13eV. Comparing the reaction rate of Al-Cu
intermetallic formation with literature data of Au-Al intermetallic formation, Kim et
21
al. showed that the reaction rate for Al-Cu was up to 100 times slower than the
reaction rate for Au-Al intermetallics [42].
Ratchev et al. found local intermetallic formation at Al-Cu bonds aged for 60 days at
150ºC by optical and scanning electron microscopy. Intermetallic compositions,
which were measured by energy dispersive spectroscopy (EDS), cannot be taken into
consideration since, according to the authors’ statement, the probe size exceeded the
intermetallics’ width [11].
Murali et al. investigated Al-Cu wire-bonded specimens heat-treated at 175°C for
100-1000 hours. The specimens were prepared by dissolution of the Al layer, and the
chemical analysis was conducted using SEM-EDS. The intermetallic phase that was
found to correspond with the measured composition was Al2Cu3 [7].
By conducting finite element simulations, Wulff et al. found that in regions with a
maximal stress gradient, the intermetallic phase formation is enhanced. Wulff et al.
also investigated the Al-Cu intermetallic phase composition and found that after 200
hours at 175ºC the Al metallization layer diffused completely into the Cu ball. The
intermetallic phases Wulff et al. identified by EDS mounted on SEM were Al2Cu,
CuAl and Al4Cu9. After 1000 hours at 175ºC, only the copper α phase was detected,
while the etched samples led the authors to believe that grain boundary diffusion was
the dominant diffusion process. In contradiction with the fact that Al-Au intermetallic
growth is much faster than growth of the Al-Cu intermetallics phase, Wulff et al.
stated that after heat treatments conducted at 250ºC the growth rate of the Cu-Al
phases and the overall phase thickness is larger than in the Au-Al system [5].
Moreover, although no intermetallics were found at as-bonded Al-Cu wire-bonds,
Wulff et al. concluded that there must be a metallurgical bond of some kind
(intermetallic compound, intermediate phases or random alloys) between Cu and Al
[5].
22
2.9. Influence of Al-Cu Wire-Bonding Process Parameters on
Bond Reliability
As noted previously, despite many advantages of the Al-Cu system over the Al-Au
system for the wire-bonding process, copper wires still have several drawbacks which
can reduce the bond reliability. In order to create a reliable Al-Cu wire-bond, special
bonding techniques should be developed, and in order to do so, the effects of the wirebonding process parameters on the bond reliability should be understood.
2.9.1.
Factors Affecting Al Cracks, Si Cratering and Bond Strength
According to Toyozawa et al., the occurrence rate of aluminum pad damage increases
as the bonding load is increased, and reaches a maximum at a bonding load of about
100 g. On the other hand, the rate of silicon cratering decreases as the bonding load is
increased. Toyozawa et al. explained this by the fact that at higher bonding loads the
Cu ball is “crushed” before shear is applied on the Al metallization by ultrasonic
power. At lower bonding loads the Al is displaced while applying ultrasonic power
via the Cu ball, which comes into direct contact with the silicon oxide and possibly
damages it [8]. Tomlinson et al. also found that the wire bond strength increased with
increased bonding load. However, they found that the silicon damage in this process
was excessive and unacceptable. Annealing at high temperatures or for a long time
reduced the number of silicon failures. Tomlinson et al. also stated that the reduced
strength of the bond made at a lower bonding load is due to insufficient mechanical
cleaning of the aluminum oxide on the Al metallization [41].
According to Toyozawa et al. the rate of underpad cracking decreases as ultrasonic
power is increased. However, the size and volume of underpad cracks becomes larger
as ultrasonic power is increased [8].
23
2.9.2.
Factors Affecting the Copper Ball Hardness
The hardness of the copper FAB is in direct correlation with the damage induced to
the aluminium metallization and silicon underpad during wire-bonding. Thus, the
factors that affect the Cu ball hardness should be considered. Kurtz et al. noted that
oxide layers harden the copper ball, thus requiring more bonding energy, which
consequently damages the underlaying silicon [43]. In addition to the fact that
elimination of copper oxide contamination on the Cu free air ball decreases the total
ball hardness, Onuki et al. found that when the bonding process (at elevated
temperatures up to 175ºC) is conducted in a reducing environment, the copper ball
cooling rate is slowed, thus reducing the dislocation density and eliminating gaseous
impurities from the ball during solidification, which decreases the copper ball
hardness [12].
Onuki et al. found that while the ball hardness decreases with an increase in the wire
purity, gaseous impurities that might be introduced into the Cu wire increase the
dislocation density and thus increase the ball hardness [12]. Tomlinson et al. found
that the microhardness of the copper ball increased after annealing the ball bond in
air, which suggests lattice strengthening by internal oxidation [41].
Zhong et al. studied factors influencing the FAB hardness, and found that in addition
to the ultrasonic current that can be optimized, the temperature which the copper wire
reaches during FAB formation affects the Cu ball hardness. The higher the
temperature induced by the electrical spark the lower the hardness of the Cu FAB
[44].
2.10. Kirkendall Effect
The Kirkendall effect is a phenomenon that occurs in a system which contains atoms
of different mutual diffusivities. The difference in diffusivity can result in voids
appearing in the bulk of the faster-diffusing atoms. Since Kirkendall voiding is
24
sometimes considered as being one of the failure modes for Al-Au wire bonds [1,45],
the effect should be taken into account for the Al-Cu system.
It is known that for copper-capped Al-Cu couples, aluminum tends to segregate
through the copper to the surface [46,47]. Therefore, theoretically, voiding in the Al
layer might occur. However, Peng et al. and Hannech et al. found voiding in the Cu
phase presumably related to Kirkendall effects [24,26]. This data correlates with the
fact that the diffusion coefficient of copper into bulk aluminum is higher than the
diffusion coefficient of aluminum into bulk copper [48]. Comparing Al-Cu with the
Al-Au system, Wei and Daud concluded that the Kirkendall effect is more explicit in
the Al-Au system than in the Al-Cu system. This is due to the fact that while in the
Al-Au system the Kirkendall effect was detected by the authors after 600h at 150ºC
and was held responsible for rapid degradation in conductivity, in the Al-Cu system
conductivity degradation related to Kirkendall voiding was detected only after 1000h
[49]. Moreover, Kim et al. and Ratchev et al. found no Kirkendall voiding in Al-Cu
wire-bonded samples annealed at 300ºC for 250h and in Al-Cu samples annealed for
120 days at 150ºC respectively [11,42]. Ratchev et al. explained the absence of
Kirkendall voiding in the Al-Cu samples as due to a low rate of intermetallic
formation in the system [11].
2.11. Characterization of Al-Cu Wire-Bonds
Over the past 16 years many methods were developed to test the reliability of copper
wire-bonds. The following is a brief summary of the test methods used:
1. In order to evaluate the effective bond area and the adhesion strength of the
bond, pull-tests followed by inspection of the bond area (on the pulled wire
and also on the Al metallization) is used.
2. Silicon cratering refers to failure in which part of the silicon underpad is
removed together with the copper ball during a pull-stress. Silicon cratering is
often characterized by the pull test in which the force needed to pull up the Cu
wire is measured, and failure at the interface is examined [8, 50].
25
3. Cracks in the silicon substrate that might be generated during Cu bonding are
usually characterized by dissolution of copper balls and the Al metallization,
followed by examination of the silicon underpad [3, 8, 50]. Dissolution of only
the Cu wires is also used to evaluate intermetallic formation at the ball-bond
interface [15].
4. Copper ball hardness can affect the underpad. The bonded ball microhardness
can be measured after mechanical polishing and etching of the ball. Impurities
and the microstructure of the copper bonds that affect the copper ball hardness
can be measured by SIMS and TEM respectively [12].
5. Optical microscopy is used in preliminary microstructure inspection [11]. Kim
et al. also used optical microscopy in order to obtain information about Al-Cu
intermetallic phase formation through color contrast [42].
6. The intermetallic chemical composition is often identified by XRD or microXRD which is performed at the failure site after a ball shear [42]. Singh et al.
remarked that Al-Cu ball bonds created on top of Cu metallization are not
visible via X-ray inspection, and this is opposed to the Al-Au bonds created on
top of Cu metallization [51].
7. The composition of Al-Cu intermetallic phases is sometimes inspected by
means of SEM-EDS and SIMS although the spatial resolution of EDS
mounted on a SEM might be insufficient for quantitative analysis, since the
interaction volume of the beam (for thick samples) might be larger than the
intermetallic width. Since SIMS is not a site-specific technique, it cannot
provide information regarding specific intermetallic phase formation [7,11].
8. The above-mentioned techniques are also used to test wire-bonds after aging
at different temperatures and for different heat-treatment durations.
All of the techniques mentioned above are used mainly for comparative analysis, and
reflect the needs of the industry. The Al-Cu system was compared with the Al-Au
wire-bonded system and optimization for the Al-Cu wire-bonding process parameters
was conducted. Moreover, not all the methods used to characterize the Al-Cu wire
bonds are reliable. For instance, the identification of Al-Cu intermetallic phases by
color is subjective, and conducting SEM-EDS analysis, when the interaction volume
26
in the sample is larger than the length-scale of the intermetallics, is unreliable [11]. In
addition, during metallographic preparation of SEM specimens (which consists of
polishing the bond in the area of interest) plastic deformation at the Al-Cu interface
occurs, making it difficult to distinguish between the pure metals and Al-Cu
intermetallics.
2.12. Investigation of Al-Cu Intermetallics Using Focused Ion
Beam
In order to produce artifact-free cross-sections of the interface of interest for SEM
investigations, specimen preparation methods beyond traditional polishing should be
used. In order to conduct quantitative analysis of the Al-Cu wire-bond intermetallic
composition, TEM and analytical TEM should be used, and site-specific techniques
for TEM specimen preparation are essential.
A dual beam focused ion beam (FIB) can be used for the preparation of SEM and
TEM samples. A dual beam FIB consists of two columns (electrons and ions)
providing for specimen preparation by FIB while conducting almost simultaneous
specimen analysis with a high resolution SEM probe (for SEM specimens and TEM
specimens at the initial stage). Crystalline grain analysis of the specimens can be
performed by ion channeling contrast [52,53,54].
During specimen preparation by FIB, gallium ions impinge onto the specimen at
30KeV parallel to the examined interface, and (at the final stages of TEM specimen
preparation) at 2-5KeV and an angle of 5-7º to the examined interface.
The exact damage which can be introduced by FIB into the Al-Cu interface is
unknown. In studies conducted on FeAl-WC and steel samples, it was found that
amorphization of the electron transparent region, redeposition of the milled material,
and Ga ion implantation might take place during TEM specimen preparation by FIB
[55,56]. However, Hutchinson et al. found that the main FIB-induced damage is an
increase of dislocation density of the irradiated material. Hutchinson stated that the
27
redeposition of the milled material can be reduced by increasing the initially-milled
section beyond the area of interest. He also stated that the changes in the chemistry of
the sample are specific for each investigated sample [57]. Bowden and Brandon also
found, that after a bombardment of gold, copper and nickel foils with 100eV argon
ions, dislocation loops were formed by an agglomeration of interstitial point defects.
The bombardment by the argon ions was conducted in a normal direction to the foils
[58]. Ishitani and Yaguchi state that for samples with high thermal conductivity, the
temperature during FIB milling increases by less than 10K [59].
No information was found considering phase formation in the Al-Cu system during
FIB milling. However, Campisano et al. [60] investigated Al-Cu phase formation in
Al-Cu thin films under ion beam bombardment. The ion bombardment was carried
out by Kr ions at low temperatures (below 100K) at energies of 60-240KeV. The only
phase formed was Al4Cu9. The phase formation occurred in the direction of the ion
beam.
28
2.13. Research Goals
The main goal of this research was to understand the process of formation and
microstructural evolution of Al-Cu wire bonds, which is essential for improving the
reliability of Al-Cu bonds.
In order to understand the wire-bonding process, the following parts of the process
were investigated:
1. The microstructure of copper free air balls (FABs), which are formed in the
first stage of the wire-bonding process, were investigated as a function of the
bonding environment.
2. The microstructure of as-bonded Al-Cu wire-bonds was correlated with the
stresses applied during bond formation.
3. The morphology and chemistry of as-bonded and heat-treated wire-bonds was
investigated and compared.
29
CHAPTER 3 : EXPERIMENTAL PROCEDURES &
CHARACTERIZATION TECHNIQUES
3.1. Specimen Preparation
3.1.1.
Free Air Balls
Two sets of FABs were examined; one set was prepared (using an electric spark as
commonly done in wire-bonding) in a reducing environment composed of 95% N2
and 5% H2 (forming gas), and another set was prepared by melting the copper wire in
air. The FABs were prepared in cross-section for microstructural investigations using
the FIB.
3.1.2.
Al-Cu Reference Specimen
In order to evaluate damage introduced into the Al-Cu interface during FIB specimen
preparation, a reference specimen that consists of an Al thin layer capped with a Cu
thin layer was prepared. This was done by physical vapor deposition (PVD) on the
Von Ardenne model NS320S mashine. During the sputtering process the chamber was
filled with argon and vacuumed down to 6.3mtorr (for Al deposition) and 5mtorr (for
Cu deposition). In order to avoid oxidation of the aluminum layer, copper deposition
was conducted immediately after the aluminum deposition without breaking the
vacuum between the two depositions. The reference TEM specimen was prepared
using FIB in a manner similar to preparation of the as-bonded and heat-treated wirebonded specimens.
3.1.3.
Al-Cu Wire-Bonds
Si wafers coated with a uniform Al metallization on top of thermally-grown SiO2,
were used for the ball bonding process. The Al metallization consisted of 0.5 wt.% Cu
and 1 wt.% Si. A 25.4µm diameter Cu wire of 99.99% purity (4N) was used for ball
bonding. The ball bonding was performed on a K&S model 8028 PPS automatic wire
bonder, using a capillary with an inner chamfer of 90º (see Figure 5). The ball bond
30
diameter was kept within 52 ± 3µm. Thermosonic ball bonding of each Si die was
performed at 220°C for an approximate time of 18 seconds per device with a pre-heat
and post-heat at 150°C. The pre-heat and post-heat times of the device were also
approximately 18 seconds each. The bonding parameters were optimized to ensure
zero pad peals, and the resulting average bond shear force was 13.3kg/mm2. Melting
of the Cu wire tip to prepare a FAB was conducted in a reducing atmosphere (95%N2,
5%H2).
Figure 5: A photograph of an experimental setup for Cu wire-bonding. An electric
spark is transferred to the wire via the EFO wand, and a reducing gas (95%N2, 5%H2)
is supplied via a small tube.
Devices may be subjected to elevated temperatures during their life-spans. It is
currently accepted that the maximum temperature that a wire-bond is expected to
withstand is 175°C. Therefore, in order to investigate time-dependent accelerated
aging of the Al-Cu wire-bonds, the devices were heat-treated at 175°C for 2, 24, 96
and 200 hours. In order to simulate encapsulation in a corrosive and nominally inert
adhesive, the specimens were heat-treated in air and in argon. Samples that underwent
a heat treatment in argon were placed in Pyrex tubes. The tubes were evacuated to a
vacuum of 5*10-1 torr and filled with 99.999% pure Argon. This vacuum-flush
procedure was repeated three times before the tubes were hermetically sealed.
31
In order to ensure investigation of the bonds subjected to a similar stress direction
during the ultrasonic force application, all the investigated samples were taken from
the “y-up” direction of the device which is perpendicular to the direction of the strip
movement in the wire-bonding machine (see Figure 6).
Figure 6: A photograph of a characteristic device strip. The wiring directions are
indicated.
The site-specific cross-sections of the Al-Cu wire-bonds prior to SEM and FIB
characterization were prepared using FIB ion beam milling. During FIB milling a
“curtaining effect” can take place. This effect appears as thicker perpendicular lines
through the specimen (Figure 7) which are commonly called “curtains”. These lines
are caused by different etching rates of copper grains depending on the grain
orientation and surface irregularities. Although platinum deposition on top of the
copper layer can reduce the “curtaining effect”, sometimes the remains of surface
irregularities result in an uneven SEM and TEM specimen thickness.
32
Figure 7: Secondary electron (SE) SEM micrograph of a copper wire-bond crosssection. Extensive curtaining can be seen at the periphery of the bond.
TEM specimens were prepared using the dual-beam FIB, while specimen examination
during the preparation was carried out by SEM and (at the final stages of the
preparation) by scanning transmission electron microscopy (STEM). During specimen
preparation, gallium ions enter the specimen at 30kV in a direction parallel to the
examined interface, while the specimen cleaning routine was carried out at 2-5kV at
an incident angle of 5-7º to the TEM specimen.
3.2. Characterization
3.2.1.
Initial Characterization
Micrometer length-scale characterization of the FABs and of copper bonds was
conducted using secondary electrons (SE) under a scanning incident electron beam in
the FIB. Characterization of the copper grain morphology of FABs and wire-bonded
cross-sections was conducted using the channeling effect in both SEM and FIB.
The channeling effect occurs when an incident ion or electron beam is aligned parallel
to a low-index zone axis of the crystalline sample. Under these conditions, the beam
does not collide with target atoms near the surface, but rather penetrates deeper into
the sample (Figure 8.a). When an inelastic collision occurs, the secondary electrons
emitted from the “channeled” grain have a smaller probability of escaping to the
surface, so these grains will appear darker than unaligned grains, or grains aligned
33
parallel to a higher index zone axes. Due to the fact that ion-induced SE’s are of a
higher energy than electron-induced SE’s, the channeling effect is more pronounced
in the FIB mode [61,62,63,64].
Figure 8: Schematic drawing of (0 0 1) plane in an FCC structure (a) in Bragg
conditions, and (b) tilted. It is visible, that more extensive channeling will occur for
the (0 0 1) planes which are in Bragg conditions with the incident beam.
Measurements of the HAZ after preparation FAB’s were conducted by measuring the
number of Cu grains per copper wire thickness. The micrographs used were acquired
by FIB in the ion-induced SE mode. The Al-Cu intermetallic grain size was measured
from micrographs obtained by TEM.
3.2.2.
Morphological Characterization
Initial characterization of the copper bonds was conducted using SEs in SEM mode of
the dual-beam FIB, and an immersed-lens mode of SE acquisition was used for better
spatial resolution.
STEM mounted on the dual-beam FIB was used for initial characterization of the
TEM specimens. This provided qualitative information on the thickness of the TEM
specimens during thinning, and initial morphological characterization of the
specimens.
Extensive morphological characterization of the Al-Cu bonds was carried out using
TEM on a number of different instruments: JEOL 3010UHR (ultra high resolution)
TEM, FEI Tecnai G2 F20 field emission gun (FEG)-TEM with a high angle annular
34
dark field (HAADF) STEM detector and a retractable EDAX detector; FEI Tecnai G2
20 with a retractable EDAX detector, and the FEI Titan 80-300kV monochromated
and aberration corrected S/TEM.
3.2.3.
Compositional Analysis
In order to obtain reliable information regarding the Al-Cu interface composition, the
interfaces were studied using combined EDS and selected area diffraction (SAD).
Energy-filtered (EF) TEM was also evaluated as a complimentary technique for
compositional analysis.
The EDS analysis was done on the FEI Tecnai F20 and Tecnai T20 microscopes. The
standardless procedure was chosen. In order to ensure acquisition of the signal from a
chosen region, the specimen was tilted 15° towards the EDS detector. The limit of
detection of Al in Cu was calculated at a constant thickness, assuming that for low Al
concentrations, the atomic concentration of Al in Cu (C) is a direct function of the Al
intensity given in counts (I) (see Equation 3).
C = const. * I
(3)
In order to detect a signal with a 95% probability the signal intensity should be larger
than twice the standard deviation of the signal background (σ) (Equation 4).
I Al ≥ 2σ
(4)
The standard deviation is calculated as a function of the background intensity,
sampling time (τ), and number of samples taken from the same region (m) (Equation
5).
σ2 =
2 I BG
mτ
(5)
The limit of detection is calculated as concentration at a point of I=2σ.
In addition, the intermetallic composition was determined by SAD. Selected area
diffraction patterns were acquired from the intermetallic grains found adjacent to the
interface region, which were also characterized by TEM-EDS. The diffraction
patterns were acquired using the Tecnai T20 TEM.
35
In EFTEM the electrons transmitted from the specimen are separated according to
their energy loss due to inelastic interaction with the specimen. Due to the high
sensitivity of the technique, it allows for acquisition of elemental maps in small time
periods (up to one minute), enabling a relatively high spatial resolution. This high
spatial resolution is impossible to achieve during elemental map acquisition via EDS,
due to specimen drift that occurs during the scan which lasts at least 30 minutes. The
elemental map presented in this research was obtained by EFTEM on the Titan 80300kV S/TEM.
36
CHAPTER 4 : RESULTS
In this chapter the morphology of FABs and the HAZ of copper wires will be
presented. In addition, the structure, morphology, and compositional analysis of asbonded and heat-treated Al-Cu wire-bonded specimens will be presented. The EDS
results will be presented as a concentration range of copper relative to aluminum
(wt.%), where the region of 0-33% copper is labeled as I, the region of 33-50% is
labeled as II, and the region of 50-70% as III (see Figure 9).
Figure 9: A section of the Al-Cu phase diagram in the temperature range of 300400°C. The labeled regions referred to in the EDS results are marked as I, II and III.
4.1. Free Air Balls
4.1.1.
Morphology
Copper is a highly oxidative metal. During melting of the copper wire tip in a nonprotective environment, oxidation of the wire occurs. As can be seen from Figure 10,
oxidation of the copper wire results in unsymmetrical ball formation, voids, and
flaws.
Figure 10: SE SEM micrograph of copper free air balls produced in air. The scale bar
is equivalent for all three micrographs. The asymmetrical FABs’ shape is visible.
37
The copper FABs formed in a reducing atmosphere have a uniform shape. No
asymmetries or flaws of the FABs were detected (Figure 11).
Figure 11: SE FIB micrograph of a representative copper FAB produced in a reducing
atmosphere. The FAB is completely symmetric with no visible flaws.
4.1.2.
Heat-Affected Zone
During melting of the copper wire by an electrical spark, a large amount of energy is
transferred into the wire. Among other paths, the thermal energy escapes via the wire
above the melted tip, creating the heat-affected zone. In order to understand the effect
of the bonding environment on HAZ evolution, the heat-affected zones of free air
balls produced in air and in a reducing environment were examined. The measured
HAZ length for FABs melted in air was 98±19μm, while the measured HAZ length
for FABs melted in a reducing environment was 46±4μm. Figure 12 presents crosssections of representative FABs formed in air and in a reducing environment. The
HAZ length for FABs formed in air is up to twice as large as that of the FABs formed
in a reducing environment.
38
Figure 12: SE FIB micrographs of representative FABs formed (a) in air and (b) in a
protective environment. The heat-affected zone of the FABs produced in air is as
twice as large as that of FABs produced in a reducing environment. The scale bar is
equivalent for both micrographs.
4.2. As-Bonded Specimens
Figure 13 presents a SE SEM micrograph of as-bonded wires. Minimal corrosion
damage is visible on the wires. Aluminum which flows out from below the copper
ball during the bonding process is indicated by an arrow.
Figure 13: SE SEM micrograph of as-bonded Cu wires. The arrow indicates
aluminum forced out from below the Cu ball during the bonding process.
SEM and ion-induced FIB SE micrographs of as-bonded wire cross-section are
presented in Figure 14. The morphology of the copper grains is visible. It should be
39
noted that in the bond periphery the copper grains are smaller, while at the center of
the bond the copper grains are larger. The copper grains at the periphery are elongated
and aligned in parallel to the Al metallization. The copper grains at the center of the
bond are more isotropic in shape. Smaller grains at the very center of the bond
adjacent to the Al-Cu interface can be detected in the FIB micrograph (Figure 14b).
The aluminum metallization thickness is not uniform, where at the very center of the
bond and at the bond periphery the aluminum layer is thinner than at the rest of the
bond.
Figure 14: (a) SE SEM micrograph and (b) SE FIB micrograph of the cross-section of
an as-bonded wire-bond.
Figure 15 presents a SE SEM micrograph of the peripheral region of an as-bonded
wire-bond. A seemingly continuous void-line propagating from the bond edge can be
seen. A region assumed to be composed of intermetallic grains (IM) region is noted.
Pores are detected in the copper ball. At the bond periphery, aluminum that was
forced out during the bonding process is visible.
Figure 15: SE SEM micrograph taken from the periphery of the as-bonded Al-Cu
wire-bond.
40
Figure 16 presents a micrograph taken from the center of the as-bonded Al-Cu wirebond. The non-uniform thickness of the aluminum metallization is visible. The
smaller copper grains at the center of the bond adjacent to the Al-Cu interface are also
visible. No intermetallics are visible at this magnification.
Figure 16: SE SEM micrograph taken from the center of the as-bonded Al-Cu wirebond. Smaller copper grains at the very center of the interface can be detected.
Figure 17 presents HAADF STEM micrograph acquired from the periphery of the asbonded wire-bond. The micrograph is taken from the region noted as intermetallic in
Figure 15. The labels (I, III) on the micrograph present acquired EDS results noted in
accordance with Figure 9. It should be noted that the investigated intermetallic region
consists of 2 regions; a region adjacent to the copper layer which is richer in copper
and its composition corresponds to region II in Figure 9, and a non-homogeneous
intermetallic region that is located adjacent to the void. This region contains less
copper (region I on the Figure 9) and in its center is a grain which has it’s a
composition similar to the aluminum metallization.
41
Figure 17: HAADF-STEM micrograph acquired from the periphery of an as-bonded
specimen. The STEM-EDS results of the annotated regions are indicated in Figure 9.
Figure 18 is a bright-field (BF) STEM micrograph of a central region of an as-bonded
Al-Cu wire-bond. Larger and more uniform intermetallic grains are found at the AlCu interface in this region, than at the bond periphery. The composition of the
intermetallic grains measured by EDS corresponds to the regions II and III on the AlCu phase diagram (Figure 9). The dark contrast located at the bottom left corner of the
micrograph is organic contamination and not related to the specimen’s microstructure.
Figure 18: BF STEM micrograph taken from the center of an as-bonded Al-Cu wirebond. The STEM-EDS results of the annotated regions are indicated in Figure 9.
42
4.3. Heat-Treated Specimens
In this section, morphological analysis of bonds heat-treated in argon and air will be
presented. The morphology and composition of specimens heat-treated in air
resembled that of specimens heat-treated in argon. Therefore, the specimens heattreated in air will be presented in a partial way, only showing results which are
different than specimens heat-treated in argon.
4.3.1.
Specimens Heat-Treated for 2 Hours
Copper wires heat-treated for 2h in argon are presented in Figure 19. Partial oxidation
of the wires can be observed. A small amount of aluminum which was forced out
from below the Cu ball during the bonding process is also visible.
Figure 19: SE SEM micrograph of the copper wires heat-treated for 2 hours at 175ºC
in argon. Partial oxidation of the wires is visible.
Figure 20 presents a cross-section of a wire heat-treated for 2 hours at 175ºC in argon.
The thickness of the aluminum metallization in not uniform, the aluminum is thinner
at the bond center and at the bond periphery. Smaller copper grains can be detected at
the bond periphery, and at the bond center adjacent to the Al-Cu interface. No
intermetallics can be detected at this magnification.
43
Figure 20: SE SEM micrograph of a cross-section of a copper wire-bond heat-treated
for 2 hours at 175ºC in argon.
Figure 21 is a SE HRSEM micrograph acquired from the periphery of a specimen
heat-treated for 2 hours in argon. Discontinuous intermetallics are detected on the AlCu interface. Voids adjacent to the intermetallic region are also found. Pores found in
the copper ball are indicated.
Figure 21: SE HRSEM micrograph taken from the periphery of a copper wire-bond
heat-treated for 2 hours at 175ºC in argon. Discontinuous non-uniform intermetallics
are detected.
Figure 22 presents a HRSEM micrograph taken from the peripheral region of a
specimen heat-treated for 2 hours in air. A discontinuous intermetallic layer is
detected at the Al-Cu interface. The intermetallic grains are not uniform; at the left
side of the micrograph smaller intermetallic grains are visible. Voids are also detected
44
at the Al-Cu interface. The aluminum metallization thickness is not uniform over the
length of the interface.
Figure 22: SE HRSEM micrograph taken from the periphery of a copper wire-bond
heat-treated for 2 hours at 175ºC in air. A discontinuous non-uniform intermetallic
layer is detected.
Figure 23 presents a HRSEM micrograph acquired from the center of a bond heattreated for 2 hours in argon. A continuous layer of small, relatively uniform,
intermetallic grains is detected at the Al-Cu interface. Pores at the copper ball are also
visible.
Figure 23: SE HRSEM micrograph taken from the central region of a copper wirebond heat-treated for 2 hours at 175ºC in argon. A continuous intermetallic layer is
visible.
Higher magnification micrographs of the center and of the periphery of the wire-bond
heat-treated for 2 hours in argon emphasize the difference between intermetallics
found in those regions (Figure 24). The intermetallic layer at the bond center (Figure
45
24a) is more uniform and continuous, while at the bond periphery voids are found
adjacent to the non-uniform intermetallics (Figure 24b).
Figure 24: SE HRSEM micrographs acquired from (a) the center and (b) the periphery
of a copper wire-bond heat-treated for 2 hours at 175ºC in argon. The difference in the
intermetallic morphology is visible. The scale bar is equivalent for both micrographs.
Figure 25 presents a BF STEM micrograph of a specimen taken from the peripheral
region of a wire-bond heat-treated for 2 hours in argon. Discontinuous and nonuniform intermetallic regions are found at the Al-Cu interface. A distorted region is
visible at the Al-Cu interface on the right side of the micrograph. The change of
copper concentration in this region is not uniform.
Grains with a low copper
concentration located near the copper ball are found adjacent to the copper-rich grains
located closer to the aluminum layer. In the distorted region, intermetallic grains with
a copper concentration up to 50 wt.% are found. A solid solution of aluminum in
copper (noted as α-Cu) is also present at the Al-Cu interface. Some voids are also
found at the interface. The copper grains located just above the distorted region are
significantly smaller than those located further away from the region.
Figure 25: HAADF-STEM micrograph of a peripheral region of a wire-bond heattreated at 175ºC for 2 hours in argon. The STEM-EDS results of the annotated
regions are indicated in Figure 9.
46
In micrographs acquired from the central region of a bond heat-treated for 2 hours in
argon (Figure 26) more uniform intermetallic grains are found. The coverage of the
Al-Cu interface by the intermetallics is also more uniform. The copper concentration
measured by STEM-EDS at the intermetallic grains matched regions I and II of the
Al-Cu phase diagram. The SAD patterns from the indicated grains confirmed the
presence of the Al2Cu (θ) phase. A dark grain noted on the micrograph by a black
arrow was found to contain 7 wt.% copper, while three grains adjacent to it contained
no more than 1 wt. % copper.
Figure 26: BF STEM micrograph of a central region of a wire-bond heat-treated at
175ºC for 2 hours in argon. The insets are SAD patterns from the indicated grains.
The STEM-EDS results of the annotated regions are indicated in Figure 9.
BF STEM micrographs acquired from the periphery of a bond heat-treated for 2 hours
in air are presented in Figure 27. The intermetallic coverage of the interface is nonuniform, and the morphology of the intermetallic grains is not uniform. The
composition measurements by STEM-EDS showed the presence of the Al2Cu (θ)
phase, and copper concentrations matching regions I and II on the Al-Cu phase
diagram. The copper-rich Al4Cu9 phase and solid solution of aluminum in copper
were also detected. The SAD patterns of the indicated grains confirmed the presence
of the Al2Cu (θ) phase.
47
Figure 27: BF STEM micrograph of a peripheral region of a wire-bond heat treated at
175ºC for 2 hours in air. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9. The scale bar is
equivalent for both micrographs.
At the center of a bond heat-treated for 2 hours in air, more uniform intermetallic
grains were found. The coverage of the interface is also more homogeneous. The
compositions of the intermetallic grains were found by STEM-EDS to match regions I
and II of the Al-Cu phase diagram. Adjacent to the copper ball, copper-rich
intermetallics were also found. SAD patterns of the large indicated grains confirmed
the presence of the Al2Cu phase.
Figure 28: BF STEM micrograph of a central region of a wire-bond heat-treated at
175ºC for 2 hours in air. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9.
4.3.2.
Specimens Heat-Treated for 24 Hours
A SE SEM micrograph of part of a cross-section of a specimen heat-treated for 24
hours in argon is presented in Figure 29. Excessive oxidation of the copper surface is
visible. Smaller copper grains at the periphery of the specimen are also visible. The
48
aluminum metallization thickness is non uniform, while at the bond periphery and at
the bond center the aluminum thickness is smaller.
Figure 29: SE SEM micrograph taken from a cross-section of a copper wire-bond
heat-treated for 24 hours at 175ºC in argon. Non-uniform thickness of aluminum
metallization is detected.
Higher magnification SEM micrographs of the central and peripheral regions of the
wire-bond heat-treated for 24 hours in argon are presented in Figure 30. The
intermetallic coverage of the Al-Cu interface is discontinuous, while more uniform
and continuous intermetallics are found at the bond center (Figure 30a), while the
intermetallics found at the bond periphery are discontinuous and less uniform (Figure
30b). The aluminum metallization is thinner in regions of intermetallic growth at both
the bond center and at the bond periphery.
Figure 30: SE SEM micrographs of (a) the center and (b) the periphery of a copper
wire-bond heat-treated for 24 hours at 175ºC in argon. A discontinuous intermetallic
layer is detected. The scale bar is equivalent for the both micrographs.
Figure 31 is a BF-STEM micrograph taken from the peripheral region of a wire-bond
heat-treated for 24 hours in argon. The intermetallic grains do not cover the entire AlCu interface. STEM-EDS analysis showed that the large intermetallic grains’
49
composition fits region II of the Al-Cu phase diagram, while some of the grains were
identified as the Al2Cu phase. In regions adjacent to the copper ball, intermetallics
with higher copper concentrations were found. The oxidation on the copper ball was
identified by STEM-EDS and SAD as CuO. Analysis of the SAD patterns acquired
from the indicated grains confirmed the presence of the Al2Cu phase. Al2Cu3 was also
found by SAD.
Figure 31: BF STEM micrograph of the peripheral region of a wire-bond heat-treated
at 175ºC for 24 hours in argon. The insets are SAD patterns from the indicated
grains. The STEM-EDS results of the annotated regions are indicated in Figure 9.
Figure 32 is a BF STEM micrograph taken from the center of a bond heat-treated for
24 hours in argon. The large voids at the center of the micrograph were introduced
during TEM specimen preparation. The intermetallic coverage of the central region of
the bond is more complete. The composition of the large intermetallic grains was
found to fit the second region (II) of the Al-Cu phase diagram. Smaller intermetallic
regions above the large intermetallic grains were identified by STEM-EDS as regions
of higher copper concentration. SAD patterns acquired from the indicated grains
confirm the presence of the Al2Cu phase.
50
Figure 32: BF STEM micrographs of the central region of a wire-bond heat treated at
175ºC for 24 hours in argon. The insets are SAD patterns from the indicated grains.
The STEM-EDS results of the annotated regions are indicated in Figure 9. The scale
bar is equivalent for both micrographs.
Figure 33 presents a BF STEM micrograph of the peripheral region, taken from a
bond heat-treated for 24 hours in air. Similar to the specimen heat-treated in argon,
the coverage of the Al-Cu interface is discontinuous. The larger grains were identified
by STEM-EDS and by SAD as the Al2Cu phase. A solid solution of aluminum in
copper was also found. The oxide on the copper ball was identified as Cu2O.
Figure 33: BF STEM micrograph of the peripheral region of a wire-bond heat treated
at 175ºC for 24 hours in air. The insets are SAD patterns from the indicated regions.
The STEM-EDS results of the annotated regions are indicated in Figure 9.
Figure 34 presents a BF STEM micrograph of the central region of a bond heattreated for 24 hours in air. At the bond center the intermetallic grains are more
uniform and the intermetallic layer is more continuous than at the bond periphery. The
concentration of larger intermetallic grains was found by STEM-EDS to fit regions I
and II on the Al-Cu phase diagram. Some of the grains were found by STEM-EDS to
have the same copper concentration as the Al2Cu phase. The SAD patterns also
identified the presence of Al2Cu for the indicated grains. An intermetallic layer with a
51
darker contrast between the large intermetallic grains and the copper ball was found
(by STEM-EDS) to contain a higher concentration of copper.
Figure 34: BF STEM micrograph of the central region of a wire-bond heat treated at
175ºC for 24 hours in air. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9.
4.3.3.
Specimens Heat-Treated for 96 Hours
Cross-sections of the peripheral and central regions of a specimen heat-treated for 96
hours in argon are presented in Figure 35 and Figure 36, respectively. The copper
grains at the bond periphery are significantly smaller than in other regions of the
copper ball (Figure 35). The copper grains at the very center of the bond adjacent to
the Al-Cu interface are slightly smaller than those located above the Al-Cu interface
(Figure 36). The thickness of residual aluminum metallization is not uniform, since
part of the aluminum was “consumed” by the growing intermetallics. Discontinuous
intermetallic regions can be observed at the bond interface, while intermetallics
observed at the bond center appear to be more continuous with a more homogeneous
morphology. Small voids are detected at the Al-Cu interface at the bond periphery.
52
Figure 35: SE SEM micrograph of the periphery of a copper wire-bond heat-treated
for 96 hours at 175ºC in argon. A discontinuous intermetallic layer is detected. Voids
at the Al-Cu interface can also be seen.
Figure 36: SE SEM micrograph of the center of a copper wire-bond heat-treated for
96 hours at 175ºC in argon. More continuous and uniform intermetallics are detected.
Figure 37 is a higher magnification SEM micrograph of the peripheral region of a
specimen heat-treated for 96 hours in argon. The non-uniform intermetallics that do
not completely cover the Al-Cu interface are visible. A discontinuous void-line
between the intermetallics and the copper ball is also detected.
Figure 37: SE SEM micrograph of the periphery region of copper wire-bond heattreated for 96 hours at 175ºC in argon. A discontinuous void-line between the
intermetallics and copper ball is visible.
53
Figure 38 presents a high magnification HRSEM micrograph of the center of a bond
heat-treated for 96 hours in argon. The intermetallic layer found at the bond center is
more continuous and uniform that the one found at the bond periphery. Small voids
are also found between the copper ball and the intermetallic layer.
Figure 38: SE HRSEM micrograph of the central region of a copper wire-bond heattreated for 96 hours at 175ºC in argon. More continuous and uniform intermetallics
than the intermetallics found at the bond periphery are visible. Small voids between
the copper ball and intermetallic layer are seen.
SEM micrographs of the cross-section of the periphery and center of a bond heattreated for 96 hour in air are presented in Figure 39. The copper grain size is
significantly smaller at the bond periphery and at the very center of the bond than for
the rest of the copper ball, while at the bond center the small copper grains are
elongated and aligned parallel to the Al-Cu interface. Discontinuous intermetallics are
found at the interface. At the bond periphery, the intermetallics are not uniform, while
at the bond center the intermetallic layer is more uniform and continuous. Part of the
aluminum metallization at the bond center is consumed by the intermetallics. No
voids are visible between the intermetallics and the SiO2 layer. A void line located at
the bond periphery between the copper ball and intermetallics/aluminum metallization
is visible. Small voids are also visible at the bond center.
54
Figure 39: SE SEM micrographs taken from (a) peripheral and (b) central regions of a
copper wire-bond heat-treated for 96 hours at 175ºC in air. A discontinuous
intermetallic layer is visible. A void line at the bond periphery can also be detected.
A BF STEM micrograph taken from the periphery of a bond heat-treated for 96 hours
in argon is presented in Figure 40. The large voids visible at the micrograph were
expanded during the specimen preparation. In the micrograph, a discontinuous
intermetallic layer is visible. Intermetallics of two different contrasts can be
distinguished: large intermetallic grains with a brighter contrast and smaller
intermetallic regions with a darker contrast located between the copper ball and large
intermetallic grains. The composition of the large intermetallic grains as measured by
STEM-EDS fits region II of the Al-Cu phase diagram, while the concentration of the
annotated grain matched the Al2Cu phase. The SAD patterns confirmed the presence
of Al2Cu in the large grains indicated on the micrograph. The smaller intermetallic
regions with a darker contrast were found by STEM-EDS to contain a larger atomic
percent of copper (fitting region III on the Al-Cu phase diagram). The presence of
Al4Cu9 in the smaller intermetallic grains was confirmed by SAD patterns. Voids
found at the Al-Cu interface are located mainly between the copper ball and the
intermetallics, and between the two intermetallic phases.
55
Figure 40: BF STEM micrograph of the peripheral region of a wire-bond heat treated
at 175ºC for 96 hours in argon. The insets are SAD patterns from the indicated
grains. The STEM-EDS results of the annotated regions are indicated in Figure 9.
Two discontinuous intermetallic layers of different contrasts and shapes can be
distinguished.
Figure 41 is a BF STEM micrograph acquired from the center of a bond heat-treated
for 96 hours in argon. A continuous intermetallic layer consisting of relatively
uniform grains is seen at the Al-Cu interface. The large intermetallic grains were
identified by STEM-EDS as matching region II of the Al-Cu phase diagram. The
darker intermetallic layer found between the large intermetallic grains and the copper
ball was found to consist of more than 70 wt.% copper. The SAD patterns acquired
from the large grains indicated on the micrograph, confirmed the presence of Al2Cu.
Figure 41: BF STEM micrograph of the central region of a wire-bond heat treated at
175ºC for 96 hours in argon. The insets are SAD patterns from the indicated grains.
The STEM-EDS results of the annotated regions are indicated in Figure 9. A
continuous intermetallic layer can be distinguished.
56
Figure 42 is a BF STEM micrograph acquired from the peripheral region of a
specimen heat-treated for 96 hours in air. The intermetallics found at the bond
interface are discontinuous. Voids were found at the Al-Cu interface between the
copper ball and the intermetallics, and between the copper and aluminum (in regions
without intermetallic grains). The large intermetallic grains were identified by STEMEDS as mainly matching region II on the Al-Cu phase diagram, while grains that
match a concentration of the Al2Cu phase were also detected. Intermetallic areas
located between the copper ball and large intermetallic grains were identified as
copper-rich regions. SAD patterns confirmed the presence of Al2Cu and Al4Cu9
phases in grains marked on the micrograph.
Figure 42: BF STEM micrograph of a peripheral region of a wire-bond heat-treated at
175ºC for 96 hours in air. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9. A discontinuous
intermetallic layer and voids can be found on the Al-Cu interface.
Figure 43 presents a BF STEM micrograph acquired from the central region of a
specimen heat-treated for 96 hours in air. Some large intermetallic grains which
totally consume the aluminum metallization are visible. Smaller intermetallic grains
having a darker contrast are found between the copper ball and large intermetallic
grains. A discontinuous layer of very small and non-uniform grains is found between
the copper ball and intermetallics. No voids are found between the intermetallic grains
and the SiO2 layer. Results from STEM-EDS analysis show that the composition of
the large intermetallic grains fits region II of the Al-Cu phase diagram and the Al2Cu
intermetallic phase. Smaller intermetallic grains with a darker contrast were found to
57
fit the Al4Cu9 phase, and the layer between the copper ball and intermetallics was
found to consist of copper, aluminum and a large weight percent of oxygen. SAD
patterns confirmed the presence of the Al2Cu phase in large intermetallic grains, and
the presence of the Al4Cu9 phase in smaller intermetallic grains adjacent to the copper
ball.
Figure 43: BF STEM micrograph of the central region of a wire-bond heat treated at
175ºC for 96 hours in air. The insets are SAD patterns from the indicated grains. The
STEM-EDS results of the annotated regions are indicated in Figure 9. An oxide layer
is found between the copper ball and intermetallics.
4.3.4.
Specimens Heat-Treated for 200 Hours
Copper wires located on specimens heat-treated for 200 hours in argon (a) and in air
(b) are presented in Figure 44. There is a higher degree of oxidation for wires heattreated in air than for those heat-treated in argon. This phenomenon is common for all
specimens heat-treated in argon and air for the same period of time. However, another
difference between wires heat-treated in argon and in air appears only for wires heattreated for 200 hours. The loop height of the wires heat-treated in air (Figure 44.b) is
significantly lower than that of the wires heat-treated in argon (Figure 44.a), and
consequently, the wire heat-treated in air is more deformed.
58
Figure 44: SE SEM micrographs of copper wires heat-treated for 200 hours at 175°C
(a) in argon and (b) in air. Deformation of the copper wire heat-treated in air is
detected.
Figure 45 presents SEM micrographs taken from (a) the center and (b) periphery of a
bond heat-treated for 200 hours in argon. The intermetallic coverage of the Al-Cu
interface is not complete. Some regions at the interface with no intermetallics are still
found. The intermetallics found at the bond periphery seem less uniform than those
found at the bond center. The intermetallics found at the bond center are more
continuous. Voids between the copper ball and the intermetallics located at the bond
periphery can be detected.
Figure 45: SE SEM micrographs taken from (a) the center and (b) periphery of a
copper wire-bond heat-treated for 200 hours at 175ºC in argon. The scale bar is
equivalent for both micrographs. A discontinuous intermetallic layer is detected.ected.
A higher magnification micrograph taken from the periphery of a bond heat-treated
for 200 hours in argon is presented in Figure 46. A void line propagating from the
bond edge is discontinuous and does not unite with voids found at the Cu-IM
interface. Two intermetallic layers with different contrast can be distinguished (noted
59
on the micrograph as IM 1 and IM 2). The voids found at the Cu-IM interface are
located between copper and IM 2, and between the two intermetallic layers.
Figure 46: SE SEM micrograph taken from the periphery of a copper wire-bond heattreated for 200 hours at 175ºC in argon. Two intermetallic layers are detected.
A high resolution micrograph taken from the center of a bond heat-treated for 200
hours in argon is presented in Figure 47. More continuous intermetallics, which are
uniform in thickness are visible. Two intermetallic layers with different contrasts can
be distinguished. The few visible voids are located on the IM1-IM2 interface.
Figure 47: SE SEM micrograph taken from the center of a copper wire-bond heattreated for 200 hours at 175ºC in argon. Two intermetallic layers of different contrast
are detected.
Figure 48 presents SEM micrographs acquired from (a) the central and (b) the
peripheral regions of a wire-bond heat-treated for 200 hours in air. The detected
intermetallics are non-uniform and discontinuous. Similar to intermetallics of the
specimens heat-treated in argon, two intermetallic layers with of a different contrast
60
can be distinguished. At the bond center a discontinuous line of pores can be detected,
while the pores are located at the IM 1-IM 2 interface. At the bond periphery, two
discontinuous void lines can be seen. A larger void line is observed at the Cu-IM
interface, while smaller voids are found at the IM 1-IM 2 interface.
Figure 48: SE SEM micrographs taken from (a) the center and (b) the periphery of a
copper wire-bond heat-treated for 200 hours at 175ºC in air. Voids at the Cu-IM
interface and between two intermetallic layers are detected. The scale bar is
equivalent for both micrographs.
Figure 49 presents a BF STEM micrograph taken from the periphery of a specimen
heat-treated for 200 hours in argon. A discontinuous intermetallic layer is detected at
the Al-Cu interface. A discontinuous void line can be seen at regions of intermetallic
growth. An intermetallic grain is also detected at the Al-SiO2 interface. The larger
intermetallic grains noted on the micrograph were identified by STEM-EDS as the
Al2Cu phase, while the smaller grains adjacent to the copper ball were found to
contain a higher copper concentration. The majority of the large grains were identified
as being Al2Cu using SAD. The smaller grains adjacent to the copper ball and one
larger grain noted on the micrograph were identified using SAD as Al4Cu9. The
intermetallic grain found at the Al-SiO2 interface was identified as Al2Cu by STEMEDS and SAD, while the two grains located above it were identified by STEM-EDS
as aluminum grains containing ~1 wt.% copper.
61
Figure 49: A BF STEM micrograph taken from the periphery of a wire-bond heat
treated at 175ºC for 200 hours in argon. The insets are SAD patterns from the
indicated grains. The STEM-EDS results of the annotated regions are indicated in
Figure 9.
Figure 50 presents a BF STEM micrograph acquired from the center of a bond heattreated for 200 hours in argon. A continuous intermetallic layer is detected. The large
intermetallic grains found are also relatively uniform in size. Smaller intermetallic
grains of a darker contrast are found adjacent to the copper layer. The large
intermetallic grains were sampled by STEM-EDS and were found to fit region II of
the Al-Cu phase diagram. Analysis of SAD patterns of the noted large intermetallic
grains confirmed the presence of Al2Cu. Smaller intermetallic grains were identified
using SAD patterns as the copper-rich Al4Cu9 phase.
Figure 50: A BF STEM micrograph taken from the center of a wire-bond heat treated
at 175ºC for 200 hours in argon. The insets are SAD patterns from the indicated
grains. The STEM-EDS results of the annotated regions are indicated in Figure 9.
Figure 51 and Figure 52 are BF STEM micrographs taken from the periphery of a
specimen heat-treated for 200 hours in air. A discontinuous intermetallic layer can be
62
seen. Void lines are found at the Cu-IM interface. The intermetallic layer consists of
intermetallics having two different contrasts (seen best in Figure 51). The larger
grains with a brighter contrast were examined by STEM-EDS and their concentration
was found to fit the Al2Cu intermetallic phase (Figure 51) and region II on the Al-Cu
phase diagram (Figure 52). The presence of Al2Cu in the large intermetallic grains
was also confirmed by SAD (Figure 52). Smaller intermetallic grains with a darker
contrast found adjacent to the copper ball were identified by STEM-EDS as Al4Cu9.
The presence of this phase in the noted grains was confirmed by SAD (Figure 51).
The oxide region found at the bond edge (Figure 52) was identified as Cu2O by
STEM-EDS, and as CuO using SAD.
Figure 51: A BF STEM micrograph taken from the periphery of a wire-bond heat
treated at 175ºC for 200 hours in air. The insets are SAD patterns from the indicated
grains. The STEM-EDS results of the annotated regions are indicated in Figure 9.
Two intermetallic layers with different contrasts can be distinguished.
Figure 52: A BF STEM micrograph taken from the periphery of a wire-bond heat
treated at 175ºC for 200 hours in air. The insets are SAD patterns from the indicated
grains. The STEM-EDS results of the annotated regions are indicated in Figure 9.
Non-uniform intermetallic grains are visible.
63
Figure 53 presents a BF STEM micrograph acquired from the center of a bond heattreated for 200 hours in air. A discontinuous intermetallic layer is detected at the AlCu interface. Isolated voids are found at the Cu-IM interface. The large intermetallic
grains having a brighter contrast were sampled by STEM-EDS and were found to fit
region II of the Al-Cu phase diagram. Analysis of SAD patterns of the noted large
intermetallic grains confirmed the presence of Al2Cu. Smaller grains with a darker
contrast incorporated into the intermetallic layer adjacent to the copper grain were
determined to be the copper-rich Al4Cu9 phase by SAD. A region found above the
intermetallic grains was found by STEM-EDS to consist of more than 70 wt.%
copper.
Figure 53: A BF STEM micrograph taken from the center of a wire-bond heat treated
at 175ºC for 200 hours in air. The insets are SAD patterns from the indicated grains.
The STEM-EDS results of the annotated regions are indicated in Figure 9.
4.4. Microstructural and Compositional Evolution of the
Wire-Bonds
The evolution of the microstructure and intermetallic composition of the Al-Cu wirebond as a function of heat-treatment time (in argon) is schematically represented in
Figure 54. The length of the intermetallic layer was measured at the bond center using
STEM micrographs.
64
Figure 54: Schematic representation of the intermetallic morphology and evolution at
Al-Cu wire-bonds as a function of a heat-treatment time.
Discontinuous intermetallics were found in as-bonded specimens. The thickness of
the intermetallic layer increases with heat-treatment time. Up to heat-treatment
duration of 200 hours at 175°C the intermetallic layer remains discontinuous. The
main intermetallics were found to be the Al2Cu phase, while after heat-treatment for
96 and 200 hours the Al4Cu9 intermetallic phase also appears adjacent to the copper
ball.
Voids were found at the IM-Cu interface, and between the two intermetallics which
were detected. After heat-treating for 96 and 200 hours, in several regions the Al layer
was consumed by intermetallic growth. No voids were found at the Al2Cu-SiO2
interface.
65
CHAPTER 5 : DISCUSSION
5.1. Copper Free Air Balls
The FAB formation is a very important step in the wire-bonding process.
Unsymmetrical ball formation might cause stress concentration in the bonded area.
Moreover, unsymmetrical ball bonds may cause contacts between leads, leading to an
electrical shortcut. Symmetrical copper FABs are formed in an inert atmosphere.
After FAB formation, energy transmitted to the wire tip is expelled from the wire
FAB by conduction (through the wire), convection and radiation. The heat transfer
coefficient for conduction via the wire for identical copper wires is constant
regardless of the environment. Thus, the heat flux through the wire is dependent only
on the residual heat in the FAB, after part of the heat is expelled via convection and
radiation. The heat transfer via radiation is two orders of magnitude lower than the
heat transfer via conduction and convection, and thus can be disregarded [65]. During
melting of the copper wire tip in a reducing atmosphere (95%N2 and 5% H2) the
reducing gas flows over the wire tip with a low velocity. The thermal conductivity
coefficient of air at 300K is lower than that of the reducing atmosphere: 26.2 MW/mK
for air and 34 MW/mK for the reducing atmosphere [66]. In this way, regardless of
the inert gas flow, more energy is transferred via convection and less via the wire
during FAB formation in an inert atmosphere. The flow of the inert gas around the
copper FAB also increases the heat transfer away from the wire tip. Since less energy
is transferred via the wire, smaller HAZ lengths for FABs formed in an inert
atmosphere are observed.
5.2. As-Bonded Specimens
5.2.1.
Morphology of
Metallization
the
Bonded
Copper-Ball
and
Aluminium
The size of the copper grains in the bonded copper ball can be evaluated using
electron and ion-induced SE micrographs (Figure 14). It can be seen, that copper
66
grains found at the bond periphery are considerably smaller than those found in the
rest of the copper ball. The grains found at the bond periphery are located adjacent to
the bond free surface. The grains are elongated and inclined with respect to the Al-Cu
interface. At the center of the Al-Cu join, copper grains smaller than the grains
located above the interface are found (Figure 39.a).
Figure 55 presents a map of the mechanical stress applied during the wire-bonding
process on the copper ball [5]. This finite elemental analysis was conducted for a
capillary with an inner chamfer of 90°. The dashed lines mark the location of the
capillary. The maximum stresses applied during the bonding process are concentrated
in regions strained by the capillary i.e. at the periphery of the copper ball (marked in
red). In these regions smaller copper grains, inclined with respect to the capillary,
were found. A region at the very center of the bond adjacent to the Al-Cu interface is
marked on the map of mechanical stresses with a blue color, meaning a smaller
amount of mechanical stress is applied during the bonding process.
Figure 55: Finite element analysis of mechanical stress applied on the copper ball
during the bonding process. Maximum stress is noted in red. Dashed lines mark the
location of the capillary.
Aluminum metallization thickness was found to decrease at the bond periphery
(underneath regions of maximum stress applied during the bonding process) and at the
very center of the bond (in a region of minimum mechanical stress applied according
to Figure 55), which imply that higher strains develop in these regions during the
bonding process. Thick aluminum layer found beyond the bond edge was forced out
67
of the bonding region during the bonding process, probably due to the ultrasonic
power application.
5.2.2.
Intermetallic Formation
Despite reoccurring reports of no intermetallic formation during the Al-Cu wire-bond
formation, intermetallics at the Al-Cu interface were detected for the as-bonded
samples. The detected intermetallics were discontinuous, and found at the periphery
of the bond as well as at the bond center. The bond periphery sustained the largest
applied mechanical stress during the bonding process, and the bond center sustained
the smallest applied stress during the bonding process (in accordance with Figure 55).
However, smaller oval grains elongated in parallel to the Al-Cu interface found at the
bond center may originate from deformation induced by the ultrasonic power, which
appear in a more pronounced way in the absence of mechanical stress.
The structure and composition of the intermetallic grains found at the bond periphery
suggest that intermetallic formation occurred during application of combined
mechanical and ultrasonic forces. Thus incorporation of an aluminum grain into the
intermetallic region can also be explained (Figure 17). Smaller intermetallic grains
found in the periphery region (relative to the intermetallic grains found at the bond
center) also suggest that extensive mechanical stress was applied during intermetallic
formation at the bond edge.
The composition of the intermetallics as measured by STEM-EDS at the bond center
differs from that found at the bond periphery: more copper-rich intermetallics were
found at the bond center. However, the examined intermetallic grains are relatively
small, and the STEM-EDS measurements could be influenced by nearby grains,
increasing inaccuracy of the measurements.
68
5.3. Heat-Treated Specimens
5.3.1.
Morphology of the Heat-Treated Specimens
For the specimens which underwent heat-treatments up to 200h at 175°C,
intermetallic growth was discontinuous. The intermetallics grew mainly in regions of
intermetallic formation. It is assumed that the energy of intermetallic formation is
only attained during the wire-bonding process. The aluminum metallization layer,
which in the wire-bonding process is forced out in the regions of maximum applied
stress, is consumed by intermetallic growth during ageing of the wire-bond in the
same regions.
5.3.2.
Composition of the Al-Cu Intermetallics
The main intermetallic phase found in as-bonded and heat-treated specimens is Al2Cu.
In specimens heat-treated for 96 and 200 hours, the Al4Cu9 phase was also detected
and was found adjacent to the copper wire. The Al4Cu9 phase first appeared in
specimens heat-treated for 96 hours at the bond periphery. No Al4Cu9 phase was
detected in the centre of specimens heat-treated for 96 hours in argon. This fact allows
for the assumption that one of the reasons for Al4Cu9 formation is discontinuity of the
Al2Cu grains: voids between the Al2Cu grains may inhibit its growth and may be
responsible for Al4Cu9 formation. However, the Al4Cu9 phase was also found at the
centre of a wire-bonds heat-treated for 200 hours, in a regions without voids (Figure
50 and Figure 53), proving that voids found adjacent to the Al2Cu grains at bonds’
periphery are not the sole reason for Al4Cu9 formation.
5.3.2.1. Imprecision in Composition Measurements
It should be noted that the intermetallic composition measured by STEM-EDS can
differ by as much as 15% from results obtained by SAD analysis (Figure 28, Figure
32, Figure 40, Figure 41, Figure 42, Figure 50 and Figure 53). STEM-EDS analysis
usually leads to a higher copper concentration than that measured by the SAD. There
are many reasons for this difference. The three main reasons will be discussed here:
69
1. The thickness of the investigated TEM specimen might be greater than the
thickness of the investigated grain. Thus the SAD pattern is obtained from the
investigated grain (which is in Bragg conditions), and the STEM-EDS results
are received from the total specimen thickness.
2. In the current research a regular TEM specimen holder rather then one which
is specifically designed for “low background” elemental analysis was used for
STEM-EDS measurements. In order to receive a strong signal from a specific
region, the specimen holder was tilted by 15° towards the EDS detector.
Figure 56 presents the untilted and tilted STEM-EDS experimental setup. In
the case of a specimen tilted towards the detector (Figure 56b) the region from
which the information is obtained (noted with green on the specimen) is not
distributed uniformly around the incident beam. In the case of a specimen
tilted towards the detector, the region located on the specimen which is closer
to the detector “donates” more signal than the region which is located far from
the detector. In this research in order to reduce the signal received from the
grid, the copper side of the specimen was always located closer to the EDS
detector. In this way, the EDS detector constantly received a stronger copper
signal than it normally would.
Figure 56: Schematic representation of an experimental STEM-EDS measurement
setup. When the specimen is tilted towards the EDS detector, the region from which
the information is received is not symmetrical around the incident electron beam.
70
3. The wire-bonding process is conducted in non-equilibrium conditions. Thus,
the investigated intermetallic regions are not necessarily stoichiometric. SAD
patterns are not sensitive to minor deviations from stoichiometry. In this case
the STEM-EDS analysis can provide more precise results concerning the real
concentrations of the investigated regions. However, since the STEM-EDS
results are less consistent (for the same heat-treatment times) than results
obtained by SAD investigations, it can be safely assumed that no large
deviations from stoichiometry are found in heat-treated specimens.
During specimen preparation by FIB, heating of the region subjected to the ion
milling might occur. Regardless of the fact that in highly conductive specimens the
heating probably does not exceed 10K [60], the influence of Ga ions on intermetallic
formation at the Al-Cu couple should be understood. Figure 57 presents a reference
specimen that consists of aluminum and copper thin layers produced by PVD
sputtering. The TEM specimen was prepared using FIB, with the same conditions
used for all the TEM specimens prepared for the current research. No intermetallic
formation is visible at the Al-Cu interface. This allows us to believe that specimen
preparation using FIB in the current research is not the source for intermetallic
formation.
Figure 57: BF STEM micrograph of Al-Cu interface obtained from a reference
specimen prepared by PVD. The TEM specimen was prepared by FIB. No
intermetallics were found at the Al-Cu interface.
71
5.3.2.2. Oxidation of the Heat-Treated Specimens
In order to simulate time-dependent ageing of specimens encapsulated in corrosive
and non-corrosive adhesives, the investigated specimens underwent heat-treatments in
air and in tubes filled with argon. It was found that specimens which were heat-treated
in argon also underwent partial oxidation (Figure 19 and Figure 29). The oxides were
identified as copper oxides. It is reasonable to assume that in 99.999% purity argon,
the maximum partial pressure of oxygen (P(O2)) is ~10-5 atm. Ellingham phase
diagrams show that the P(O2) threshold for Cu2O formation at 500K is 7.1x10-28 atm.,
and the P(O2) threshold for CuO formation is 4x10-19atm. [67]. Thus in the current
experimental setup it is impossible to prevent wire oxidation. However, the oxidation
of copper wires is less extensive in specimens heat-treated in argon (see for example
Figure 44) excluding the specimen heat-treated in argon for 24 hours, which was
probably insufficiently rinsed prior to sealing with argon.
No difference in intermetallic composition was found for specimens heat-treated in air
compared to specimens heat-treated in argon. However, it was found that copper wire
creep occurred in specimens heat-treated in air for 200 hours. No similar wire creep
was found in specimens heat-treated in argon (Figure 44). This phenomenon can be
explained by extensive corrosion of the wire which accelerated the copper wire creep
at a very low temperature (relative to the copper melting point). During heattreatment, copper wires are still connected to the package by the second bond, thus
creep occurs on the strained copper wire.
In most samples heat-treated in air, no oxidation of the Al-Cu interface was found.
However, oxides were detected at the interface of a wire-bond heat-treated in air for
96 hours (Figure 43). The oxide line was detected between the intermetallics and the
copper ball, which means that the region was oxidized after the wire-bonding process.
However, due to the fact that no oxide regions were found in other specimens heattreated in air for 96 and 200 hours, it is safe to assume that the phenomenon observed
in Figure 43 is singular and presents local failure of the wire-bonding process. We
assume that in this case, voids introduced during the wire-bonding evolved into
72
continuous cracks, which became oxidized by oxygen entering from the bond edge.
The propagation of the crack is best visible in Figure 39.
5.3.3.
Morphology of the Al-Cu Intermetallics
5.3.3.1. Voids Found at the Al-Cu Interface
Two types of voids were found at the Al-Cu interfaces of heat-treated specimens;
voids that were formed during the wire-bonding process, and voids that evolved
during the heat-treatment of the bond. The voids that were formed during the bonding
process are relatively large, continuous, located at the bond periphery and found in asbonded and heat-treated specimens (see for example Figure 21, Figure 22 and Figure
30). In general, these voids are not the source of bond failure (see for example Figure
46). Voids formed during the wire-bonding process were found at the IM-Cu interface
(Figure 43), while voids at the IM-Al interface and voids in the aluminium
metallization were also found (Figure 17, Figure 25). These voids may shrink during
the heat-treatment of the specimens [68,69]. Voids formed during heat-treatments
were usually found to be discontinuous, relatively small, detected mainly in
specimens heat-treated for 96 and 200 hours and located not only at the bond
periphery but also in the centre of the investigated bonds. These voids were mainly
found at the Al4Cu9-Cu and at the Al4Cu9-Al2Cu interfaces.
In order to understand processes that occur at the Al-Cu interface during intermetallic
growth, atomic packing factors (APFs) of the Al2Cu and Al4Cu9 intermetallics were
calculated and compared to those of pure Al and Cu. The atomic packing factors of
the intermetallic phases were calculated by dividing the total volume of atoms
occupying the unit cell by the total volume of the unit cell (Equation 6):
APF =
ΣN atomsVatoms
Vunit .cell
(6)
Table 4 presents atomic packing factors of pure metals and of the Al2Cu and Al4Cu9
intermetallics. While the APF of Al2Cu is very close to those of the pure metals, the
73
APF of Al4Cu9 is 5% higher than those of Al2Cu and of the pure metals. Therefore,
voids formed at the Al4Cu9-Cu and Al4Cu9-Al2Cu interface can be explained by
volume reduction during Al4Cu9 phase formation.
Table 4: Atomic packing factors of Al, Cu and intermetallic phases found at the AlCu interface.
Phase
APF
Al
Al2Cu
Al4Cu9
Cu
0.74
0.739
0.776
0.74
5.3.3.2. Al2Cu Intermetallic Phase Growth
Bonding of a single wire in the wire-bonding process (in this research) took less than
0.2 seconds. The temperatures that evolve during the ultrasonic activation at the AlCu interface are unknown and might approach melting points of the metals of interest
[45]. In order to understand grain growth of the Al-Cu intermetallics and to determine
in which state intermetallic formation occurs, the grain growth of the Al2Cu
intermetallic phase was analysed. Due to lack of a uniform Al4Cu9 layer, the timedependent growth of this phase was not analyzed.
The thickness of the Al2Cu layer was measured at the centre of as-bonded and heattreated specimens (results are noted on Figure 54). The specimens heat-treated for 200
hours were not included in the calculations since for those specimens it was difficult
to distinguish between the Al2Cu and the Al4Cu9 intermetallics. Figure 58 presents the
square of the Al2Cu intermetallic size as a function of heat-treatment time. The direct
correlation between the square of the intermetallic size and the heat-treatment time
allows for the use of Equation 1 for calculation of the diffusion coefficient of copper
in aluminium at 175°C.
74
Figure 58: The square of intermetallic size as a function of heat-treatment time. The
purple line is a linear fit of the represented data.
After extraction of the diffusion coefficient of copper in aluminium at 175°C, the
activation energy for Al2Cu formation is calculated using the Arrhenius equation
(Equation 2). The activation energy for Al2Cu formation was found to be ~1.24eV,
which correlates well with data acquired by Hamm and Vanderberg for Al2Cu
formation via bulk solid state diffusion in Al-Cu thin films at 157-220ºC (1.211.24eV) [32].
A few intermetallic grains were found to grow towards a grain boundary between two
aluminum grains in a manner similar to grain boundary diffusion (Figure 18, Figure
32 and Figure 41). Moreover, in two occasions copper was found to diffuse through
aluminum grain boundaries and intermetallic grains were created (see Figure 26 and
Figure 49). However, examples of grain boundary diffusion are singular and at heattreatments up to 200 hours have little influence on the Al2Cu intermetallic layer
thickness. Thus, the main diffusion mechanism observed in this system is bulk solid
state diffusion.
75
CHAPTER 6 : SUMMARY AND CONCLUSIONS
The purpose of this research was to understand the time-dependent behavior of Al-Cu
wire-bonds. In order to achieve this goal, the stages of the wire-bonding process
(formation of the FAB, growth of the HAZ, Al-Cu intermetallic growth at elevated
temperatures) were studied. To receive maximum information regarding the studied
processes, high resolution microscopy techniques were used.
Specimens were
prepared using a site-specific and relatively artifact-free technique (FIB), and the
morphological analysis of the specimens was conducted by HRSEM, FIB and
TEM+SAD. Compositional analysis of the specimens was conducted by STEM-EDS.
Prior to the wire-bonding process, the copper wire is melted by an electrical spark to
form a symmetric FAB. To prevent copper FAB oxidation, which hardens the copper
ball and might damage the aluminum metallization during the bonding process, the
FAB was formed in a reducing gas. FABs produced in a reducing gas are uniform in
shape and flawless, in contrast to the FABs produced in air. Moreover, it was found
that due to the higher heat transmittance of the reducing gas (that consists of 95 wt.%
N2 and 5 wt.% H2) in comparison with air, the length of the HAZ in copper wires
melted in a reducing gas is shorter than the HAZ of copper wires melted in air. This
is important, since a shorter HAZ is associated with higher mechanical stability of the
copper ball-bond.
Qualitative analysis of the bonded copper ball grains revealed smaller copper grains at
the periphery of the bond, in a region where the highest mechanical stresses were
applied during the bonding process. Smaller copper grains were also found at the
center of copper ball, in a region where low mechanical stresses were applied during
the bonding process. This implies that at the center of the wire-bond high residual
stresses exist. These stresses may originate from a more pronounced influence of the
ultrasonic vibrations.
As apposed to the data available in the literature, intermetallics were found in asbonded specimens. The intermetallics formed were discontinuous, and were found
adjacent to smaller copper grains, i.e. in regions subjected to higher stresses. This
76
indicates that localized stress incorporated into the Al-Cu interface during the bonding
process catalyze intermetallic formation. The intermetallics found in heat-treated
specimens were also discontinuous, probably due to the fact that the energy sufficient
for intermetallic formation is only achieved during the wire-bonding process. In
regions subjected only to the ultrasonic strains (i.e. center of the bond) more uniform
intermetallic grains were found (in comparison with intermetallic grains found at the
bond periphery).
The main intermetallic phase in as-bonded and heat-treated specimens was found to
be the Al2Cu phase. After longer heat-treatment times Al4Cu9 was found. Al2Cu grain
growth occurred via solid state diffusion. The growth of the Al2Cu phase at the bond
periphery was partially inhibited by voids adjacent to the Al2Cu grains, a phenomenon
that is also responsible for enhanced Al4Cu9 growth at the bond periphery.
Two types of voids were found at the Al-Cu interface; voids that were created during
the wire-bonding process, and voids that appear after the heat-treatments. Voids that
were introduced during the wire-bonding process are usually located at the IM-Cu
interfaces. These voids do not result in failure of the bond, and partially “heal” during
the heat-treatments. Voids that were found in heat-treated specimens were found at
IM-Cu and IM-IM interfaces, and are probably created by a mechanism of volume
reduction during Al4Cu9 intermetallic formation.
Almost no continuous void-lines were found in specimens heat-treated at 175ºC for
up to 200 hours. Thus it can be stated that void formation in specimens heat-treated
for 200 hours is not a main source for bond failure. For specimens heat-treated for 200
hours in air, copper wire deformation was observed, which can be regarded as a
failure of the wire-bonds.
Specimens that underwent heat-treatment in argon underwent partial oxidation of the
copper wires. This occurred due to the fact that in the experimental setup the
oxidation of the copper specimens could not be avoided.
77
In general, at least two independent techniques should be used for composition
characterization. The accuracy of STEM-EDS analysis depends on specimen
thickness and experimental working conditions. During this research, those conditions
were not optimal for acquiring accurate results. Thus, in this research SAD analysis
was essential for precise intermetallic identification.
78
CHAPTER 7 : FUTURE OUTLOOK
7.1. Recommendations for the Process Improvement
Intermetallics formed at the Al-Cu interface indicate removal of surface contaminants
during the wire-bonding process and assure direct contact between the metals.
Continuous and flawless intermetallic formation (similar to the intermetallics found at
the center of the bonds) can improve the mechanical and electrical reliability of Al-Cu
wire-bonds. However, voids at Al-Cu interfaces may compromise mechanical and
electrical properties of the bond. While voids formed by volume reduction during
intermetallic formation cannot be eliminated, voids introduced into the Al-Cu
interfaces during the wire-bonding process by application of high mechanical strains
should be avoided.
We believe that reduction of the mechanical load applied during the wire-bonding
process will assure a more uniform intermetallic coverage at the Al-Cu interface.
Reduction of the mechanical load should also reduce the density of voids introduced
into the Al-Cu interface during the wire-bonding process. In order to maintain the
total energy applied during the wire-bonding process with lower mechanical loads, the
integral time of ultrasonic application should be increased. Increasing the time of
ultrasonic activation should also reduce void density at the Al-Cu interface by forcing
oxides and other surface contaminations out from the bonded region.
7.2. Recommendations for Future Work
No quantitative data regarding the reliability of heat-treated bonds was obtained in
this research. In order to correlate between the intermetallic structure and the bond
reliability, extensive mechanical and electrical tests of as-bonded and heat-treated
specimens should be performed.
79
To understand the influence of mechanical and ultrasonic energies applied during the
wire-bonding process on Al-Cu intermetallic growth and bond reliability, the
morphology and composition of Al-Cu wire-bonds prepared using different loading
procedures should be investigated. Comparative mechanical and electrical tests should
also be preformed on heat-treated Al-Cu wire-bonds prepared by different loading
procedures.
In order to obtain full understanding of the wire-bond reliability, second bonds (i.e.
bonds between the copper wire and the package) should be investigated.
80
APPENDIX
I. Crystallographic Data Used for Compositional Analysis
Table 5: Crystallographic data of Al-Cu intermetallics, used for SAD patterns
identification [70].
Structure
Pearson
Lattice
Phase
Space Group
Type
Symbol
Parameter (nm)
a=0.6067
Al2Cu
Al2Cu
tI12
I4/mcm
c=0.4877
a=0.4146
Al2Cu3
InNi2
hP6
P63/mmc
c=0.5063
Al4Cu9
Al4Cu9
cP52
P43m
a=0.87039
II. Evaluation of the EFTEM for Composition Measurements
In this research, STEM-EDS and SAD techniques were used for compositional
analysis of intermetallics found at the Al-Cu interface. However, these techniques
cannot map regions that underwent oxidation. Oxides can be only recognized by SAD
in case of large amorphous regions (see for example Figure 31and Figure 33) or if
relatively large crystalline grains (80nm or more) are formed. Large oxide regions can
be also recognized by STEM-EDS, however as it has been discussed above, the
precision of this technique in a current setup is relatively low. Moreover, high
resolution elemental maps can be obtained only by the EFTEM technique.
Figure 59 presents TEM micrographs taken from a center of an as-bonded specimen.
On the BF TEM micrograph (Figure 59.a) the intermetallic regions are noted. Voids
are also visible. Figure 59.b is an EFTEM micrograph, which provides an elemental
map of the examined interface. Aluminum, oxygen and copper are marked in red,
green and blue respectively. The oxygen is found at the edges of the aluminum and
copper regions, around the voids and in regions of no intermetallic formation at the
Al-Cu interface. It is unknown whether the oxidation occurred during the wirebonding process or after the specimen preparation. However, it is evident, that
EFTEM technique is capable of providing qualitative information regarding the Al-Cu
wire-bond oxidation.
81
Figure 59: (a) BF TEM and (b) EFTEM micrographs taken from a central region of
as-bonded Al-Cu wire-bond.
III. Computation of the Limit of Detection of Al in Cu
According to Equation 3, at very low concentrations of aluminum in copper, the
dependence of aluminum concentration (as given by a computational program
embedded in the Tecnai TEM- Tecnai Instrumental Analysis (TIA)) with the number
of counts is linear. Three measurements taken from the copper regions with low
aluminum concentration were used. The thickness of the copper region for these
measurements was constant (~200 nm). The standard deviation (σ) used, was the
standard deviation of the background as received from TIA. The values used for the
computation of the limit of detection are registered in Table 6, while CAl is the
concentration of Al in copper as measured by TIA, IAl is a number of Al counts (per
second), constant is calculated using Equation 3, and limit of detection is calculated
using Equation 7, which is derived from Equations 3 and 4:
Clim,0.95 ≥ const. * 2σ
(7)
Table 6: Values of points used for the computation of limit if detection of Al in Cu for
the current experimental setup.
IAl
Limit of
Measurement CAl (%)
σBG
Constant
(counts/sec)
detection
1
0.245
180.1
129.4
1.36*10-3
0.35
-3
2
0.282
223.1
150. 3
1.26*10
0.38
3
0.126
163.2
114.3
0.77*10-3
0.18
-3
4
0.181
269.6
147.6
0.67*10
0.2
82
It is evident, that the assumption of linear dependence of aluminum concentration
with count number at low concentrations is not precise. At lower aluminum
concentrations the assumption of linear dependence is more exact. Therefore, it is safe
to assume that the limit of detection of aluminum in copper in a current experimental
setup with certainty of 95% is 0.2%.
IV. Calculation of the Formation Energy for the Al2Cu Phase
The diffusion coefficient of copper in aluminum at 0K was obtained from the CRC
handbook [48]. The values used for calculations of energy of formation for the Al2Cu
phase are listed in Table 7, while the diffusion coefficient at 175°C (D) was calculated
using Equation 1. The formation energy of the Al2Cu phase (Q) was calculated using
Equation 2.
Table 7: Computation of energy of formation of the Al2Cu phase.phase.
Heat-Treatment
Time (hr)
2
24
96
IM Thickness
(nm)
80
200
500
D (cm/hr)
Q(eV)
3.2*10-11
1.7*10-11
2.6*10-11
1.23
1.26
1.24
83
REFERENCES
1.
G. Harman, "Wire Bonding in Microelectronics Materials, Processes, Reliability and Yield". 2 ed. Electronic
Packaging and Interconnection, ed. C.A. Harper. 1997: McGraw-Hill.
2.
M. Braunovic and N. Alexandrov, Intermetallic Compounds at Aluminum-to-Copper Electrical Interfaces - Effect
of Temperature and Electric-Current, IEEE Transactions on Components Packaging and Manufacturing
Technology Part A, 17[1]: 78-85, 1994.
3.
S. Murali, N. Srikanth and C.J. Vath, Grains, Deformation Substructures, and Slip Bands Observed in
Thermosonic Copper Ball Bonding, Materials Characterization, 50[1]: 39-50, 2003.
4.
T.H. Ramsey and C. Alfaro, High-Frequency Enhancement for Ambient Temperature Ball Bonding,
Semiconductor International, 20[9]: 93-94, 96, 1997.
5.
F.W. Wulff, C.D. Breach, D. Stephan, Saraswati, K.J. Dittmer and M. Garnier, Further Characterisation of
Intermetallic Growth in Copper and Gold Ball Bonds on Aluminium Metallisation. 2005. www.kns.com.
6.
Y.R. Jeng and J.H. Horng, A Microcontact Approach for Ultrasonic Wire Bonding in Microelectronics, Journal of
Tribology-Transactions of the Asme, 123[4]: 725-731, 2001.
7.
S. Murali, N. Srikanth, Y.M. Wong and C.J. Vath, Fundamentals of Thermo-Sonic Copper Wire Bonding in
Microelectronics Packaging, Journal of Materials Science, 42[2]: 615-623, 2007.
8.
K. Toyozawa, K. Fujita, S. Minamide and T. Maeda, Development of Copper Wire Bonding Application
Technology, IEEE Transactions on Components, Hybrids, and Manufacturing Technology, 13[4]: 667-72, 1990.
9.
I. Singh, Sr., J.Y. On and L. Levine, Sr., Enhancing Fine Pitch, High I/O Devices with Copper Ball Bonding,
Proceedings - Electronic Components & Technology Conference, 55 [Vol. 1]: 843-847, 2005.
10. S.L. Khoury, D.J. Burkhard, D.P. Galloway and T.A. Scharr, A Comparison of Copper and Gold Wire Bonding on
Integrated-Circuit Devices, IEEE Transactions on Components Hybrids and Manufacturing Technology, 13[4]:
673-681, 1990.
11. P. Ratchev, S. Stoukatch and B. Swinnen, Mechanical Reliability of Au and Cu Wire Bonds to Al, Ni/Au and
Ni/Pd/Au Capped Cu Bond Pads, Microelectronics and Reliability. Microelectronic Reliability : 2005 ROCS
Workshop, 46[8]: 1315-1325, 2006.
12. J. Onuki, M. Koizumi and I. Araki, Investigation of the Reliability of Copper Ball Bonds to Aluminum Electrodes,
IEEE Transactions on Components Hybrids and Manufacturing Technology, 10[4]: 550-555, 1987.
13. S. Murali, Formation and Growth of Intermetallics in Thermosonic Wire Bonds: Significance of Vacancy-Solute
Binding Energy, Journal of Alloys and Compounds, 426[1-2]: 200-204, 2006.
14. M. Deley and L. Levine, The Emergence of High-Volume Copper Ball Bonding, IEEE/CPMT/SEMI International
Electronics Manufacturing Technology Symposium, 29th, San Jose, CA, United States, July 14-16, 2004: 186190, 2004.
15. F.W. Wulff, C.D. Breach, D. Stephan, Saraswati and K.J. Dittmer, Characterisation of Intermetallic Growth in
Copper and Gold Ball Bonds on Aluminium Metallisation, Proceedings of Electronics Packaging Technology
Conference , 6th, Singapore, Singapore, Dec. 8-10, 2004: 348-353, 2004.
16. E.A. Brandes and G.B. Brook, Smithells Metals Reference Book. 5th. ed. 1992.
17
www.ksn.com, as on 30 August 2007.
18. J.L. Murray, The Aluminum-Copper System, International Metals Reviews, 30[5]: 211-33, 1985.
19. C. Macchioni, J.A. Rayne and C.L. Bauer, Low-Temperature Resistivity of Bulk Copper-Aluminum Alloys,
Physical Review B, 25[6]: 3865-3870, 1982.
20
E.R. Wallach and G.J. Davies, Joint Resistances and Current Paths in Heat-Treated Aluminium/Copper SolidPhase Welds, Metal Science, 11[3]: 97-102, 1977.
21. W.B. Lee, K.S. Bang and S.B. Jung, Effects of Intermetallic Compound on the Electrical and Mechanical
Properties of Friction Welded Cu/Al Bimetallic Joints During Annealing, Journal of Alloys and Compounds,
390[1-2]: 212-219, 2005.
22. M. Abbasi, A.K. Taheri and M.T. Salehi, Growth Rate of Intermetallic Compounds in Al/Cu Bimetal Produced by
Cold Roll Welding Process, Journal of Alloys and Compounds, 319[1-2]: 233-241, 2001.
23. N.M. Voropai, Y.Y. Gretskii, A.V. Lozovskaya and T.I. Tsarenko, Strength and Structure of Aluminum-Copper
Welded Joints, Tsvetnye Metally (Moscow, Russian Federation), 40[1]: 86-8, 1967.
24. X.K. Peng, R. Wuhrer, G. Heness and W.Y. Yeung, On the Interface Development and Fracture Behaviour of
Roll Bonded Copper Aluminium Metal Laminates, Journal of Materials Science, 34[9]: 2029-2038, 1999.
84
25. D.I. Lainer and A.K. Kurakin, Diffusion During Production of Aluminum-Copper Bimetal, Tsvetnye Metally
(Moscow, Russian Federation), 37[10]: 63-6, 1964.
26. E.B. Hannech, N. Lamoudi, N. Benslim and B. Makhloufi, Intermetallic Formation in the Aluminum-Copper
System, Surface Review and Letters, 10[4]: 677-683, 2003.
27. E.R. Wallach and G.J. Davies, Mechanical Properties of Aluminum-Copper Solid-Phase Welds, Metals
Technology (London), 4, Pt. 4: 183-90, 1977.
28. A.E. Gershinskii, B.I. Fomin, E.I. Cherepov and F.L. Edelman, Investigation of Diffusion in the Copper-Aluminum
Thin Film System, Thin Solid Films, 42[3]: 269-75, 1977.
29. S.U. Campisano, E. Costanzo, F. Scaccianoce and R. Cristofolini, Growth Kinetics of the Q Phase in AluminumCopper Thin Film Bilayers, Thin Solid Films, 52[1]: 97-101, 1978.
30. K. Rajan and E.R. Wallach, A Transmission Electron Microscopy Study of Intermetallic Formation in AluminumCopper Thin-Film Couples, Journal of Crystal Growth, 49[2]: 297-302, 1980.
31. J.M. Vandenberg and R.A. Hamm, An Insitu X-Ray Study of Phase Formation in Al-Cu Thin-Film Couples, Thin
Solid Films, 97[4]: 313-323, 1982.
32. R.A. Hamm and J.M. Vandenberg, A Study of the Initial Growth-Kinetics of the Copper-Aluminum Thin-Film
Interface Reaction by Insitu X-Ray-Diffraction and Rutherford Backscattering Analysis, Journal of Applied
Physics, 56[2]: 293-299, 1984.
33. H.T.G. Hentzell, R.D. Thompson and K.N. Tu, Interdiffusion in Copper-Aluminum Thin-Film Bilayers .1.
Structure and Kinetics of Sequential Compound Formation, Journal of Applied Physics, 54[12]: 6923-6928,
1983.
34. M.P. Shearer, S.K. Sen and C.L. Bauer, Investigation of Interfacial Reactions in Bimetallic Thin-Film Couples of
Al/Cu and Pb/Cu by Measurement of Contact Resistance, Physica Status Solidi a-Applied Research, 69[1]: 139149, 1982.
35. H.G. Jiang, J.Y. Dai, H.Y. Tong, B.Z. Ding, Q.H. Song and Z.Q. Hu, Interfacial Reactions on Annealing Cu/Al
Multilayer Thin-Films, Journal of Applied Physics, 74[10]: 6165-6169, 1993.
36. P.V. Thomas, V.K. Vaidyan and J.T. Abraham, Surface Oxidation Studies of Al/Cu Thin Films, Vacuum, 47[1]:
83-5, 1996.
37. J.L. Cai and H. Gong, The Influence of Cu/Al Ratio on Properties of Chemical-Vapor-Deposition-Grown P-Type
Al-Cu-O Transparent Semiconducting Films, Journal of Applied Physics, 98[3], 2005.
38. C.A. Chang, Interaction in Al-Cu and Pt/Cu/Al Thin-Films - Competing Oxygen Effects and Preferential Reaction
between Pt and Al-Cu Alloy, Applied Physics Letters, 44[3]: 310-312, 1984.
39. S. Murali, N. Srikanth and C.J. Vath, III, An Analysis of Intermetallics Formation of Gold and Copper Ball
Bonding on Thermal Aging, Materials Research Bulletin, 38[4]: 637-646, 2003.
40. S. Murali, N. Srikanth and C.J. Vath, An Evaluation of Gold and Copper Wire Bonds on Shear and Pull Testing,
Journal of Electronic Packaging, 128[3]: 192-201, 2006.
41. W.J. Tomlinson, R.V. Winkle and L.A. Blackmore, Effect of Heat-Treatment on the Shear-Strength and Fracture
Modes of Copper Wire Thermosonic Ball Bonds to Al-1-Percent Si Device Metallization, Ieee Transactions on
Components Hybrids and Manufacturing Technology, 13[3]: 587-591, 1990.
42. H.-J. Kim, J.Y. Lee, K.-W. Paik, K.-W. Koh, S. Choe, J. Lee, J.-T. Moon and Y.-J. Park, Effects of Cu/Al
Intermetallic Compound (Imc) on Copper Wire and Aluminum Pad Bondability, IEEE Transactions on
Components and Packaging Technologies, 26[2]: 367-374, 2003.
43. J. Kurtz, D. Cousens and M. Dufour, Copper Wire Ball Bonding, Proceedings - Electronic Components
Conference, 34th: 1-5, 1984.
44. Z.W. Zhong, H.M. Ho, Y.C. Tan, W.C. Tan, H.M. Goh, B.H. Toh and J. Tan, Study of Factors Affecting the
Hardness of Ball Bonds in Copper Wire Bonding, Microelectronic Engineering, 84[2]: 368-374, 2007.
45 . A. Karpel, G. Gur, Z. Atzmon and W.D. Kaplan, Tem Microstructural Analysis of as-Bonded Al-Au Wire-Bonds,
Journal of Materials Science, 42[7]: 2334-2346, 2007.
46. P.J. Ding, W.A. Lanford, S. Hymes and S.P. Murarka, Thermal Annealing of Buried Al Barrier Layers to
Passivate the Surface of Copper Films, Applied Physics Letters, 65[14]: 1778-80, 1994.
47. P.-I. Wang, S.P. Murarka, D.A. Kaminski, S. Bedell and W.A. Lanford, Surface Segregation of Al of the Bilayers
of Pure Cu and Cu-Al Alloy Films, Journal of the Electrochemical Society, 148[9]: G481-G486, 2001.
48 J.F. Shackelford and W. Alexander, eds. Crc Materials Science and Engineering Handbook. 3 ed., ed. J.F.
Shackelford and W. Alexander. 2001, CRC Press LLC.
49. T.C. Wei and A.R. Daud, Mechanical and Electrical Properties of Au-Al and Cu-Al Intermetallics Layer at Wire
Bonding Interface, Journal of Electronic Packaging, 125[4]: 617-620, 2003.
85
50. C.W. Tan and A.R. Daud, Bond Pad Cratering Study by Reliability Tests, Journal of Materials Science: Materials
in Electronics, 13[5]: 309-314, 2002.
51. I. Singh, Sr., J.Y. On and L. Levine, Sr., Enhancing Fine Pitch, High I/O Devices with Copper Ball Bonding,
Proceedings - Electronic Components & Technology Conference, 55th[Vol. 1]: 843-847, 2005.
52
http://www.s3.infm.it/fib.html
53. L.A. Giannuzzi and F.A. Stevie, eds. Introduction to Focused Ion Beams. 2005, Springer.
54. B.W. Kempshall, S.M. Schwarz, B.I. Prenitzer, L.A. Giannuzzi, R.B. Irwin and F.A. Stevie, Ion Channeling
Effects on the Focused Ion Beam Milling of Cu, Journal of Vacuum Science & Technology B, 19[3]: 749-754,
2001.
55. J.M. Cairney and P.R. Munroe, Redeposition Effects in Transmission Electron Microscope Specimens of FealWc Composites Prepared Using a Focused Ion Beam, Micron, 34[2]: 97-107, 2003.
56. J.S. Yu, J.L. Liu, J.X. Zhang and J.S. Wu, Tem Investigation of Fib Induced Damages in Preparation of Metal
Material Tem Specimens by Fib, Materials Letters, 60[2]: 206-209, 2006.
57. C.R. Hutchinson, R.E. Hackenberg and G.J. Shiflet, A Comparison of EDS Microanalysis in Fib-Prepared and
Electropolished Tem Thin Foils, Ultramicroscopy, 94[1]: 37-48, 2003.
58. P. Bowden and D.G. Brandon, The Production of Interstitial Point Defects by Low Energy Ion Bombardment,
Journal of Nuclear Materials, 9[3]: 348-354, 1963.
59. T. Ishitani and T. Yaguchi, Cross-Sectional Sample Preparation by Focused Ion Beam: A Review of Ion-Sample
Interaction, Microscopy research and technique, 35[4]: 320-33, 1996.
60. S.U. Campisano, T. Chu, S. Cannavo and E. Rimini, Energy Dependence of Compound Growth in GoldAluminum and Copper-Aluminum Bilayer Systems During Ion Beam Mixing, Materials Research Society
Symposium Proceedings, 27[Ion Implant. Ion Beam Process. Mater.]: 97-102, 1984.
61. G. Wang, S.D. Wu, C. Esling, G.Y. Li, L. Zuo and Z.G. Wang, Microstructural Characteristics by Ebsd and Ecc in
Ecae Processed Pure Cu Samples, Advanced Engineering Materials, 5[8]: 593-597, 2003.
62. G. Wang, S.D. Wu, L. Zuo, C. Esling, Z.G. Wang and G.Y. Li, Microstructure, Texture, Grain Boundaries in
Recrystallization Regions in Pure Cu Ecae Samples, Materials Science and Engineering a-Structural Materials
Properties Microstructure and Processing, 346[1-2]: 83-90, 2003.
63. B.W. Kempshall, S.M. Schwarz, B.I. Prenitzer, L.A. Giannuzzi, R.B. Irwin and F.A. Stevie, Ion Channeling
Effects on the Focused Ion Beam Milling of Cu, Journal of Vacuum Science & Technology B, 19[3]: 749-754,
2001.
64. J.D. Casey, M. Phaneuf, C. Chandler, M. Megorden, K.E. Noll, R. Schuman, T.J. Gannon, A. Krechmer, D.
Monforte, N. Antoniou, N. Bassom, J. Li, P. Carleson and C. Huynh, Copper Device Editing: Strategy for
Focused Ion Beam Milling of Copper, Journal of Vacuum Science & Technology B, 20[6]: 2682-2685, 2002.
65. I.M. Cohen, L.J. Huang and P.S. Ayyaswamy, Melting and Solidification of Thin Wires - a Class of PhaseChange Problems with a Mobile Interface .2. Experimental Confirmation, International Journal of Heat and Mass
Transfer, 38[9]: 1647-1659, 1995.
66. “CRC Handbook of Chemistry and Physics”, 87th Edition, http://www.hbcpnetbase.com/, as on 30 August 2007.
67. “Ellingham phase diagrams”, www.engr.sjsu.edu, as on 30 August 2007.
68. A. Hill and E.R. Wallach, Modeling Solid-State Diffusion Bonding, Acta Metallurgica, 37[9]: 2425-2437, 1989.
69. Y. Takahashi, K. Inoue and K. Nishiguchi, Identification of Void Shrinkage Mechanisms, Acta Metallurgica Et
Materialia, 41[11]: 3077-3084, 1993.
70. P. Villaris, Pearson's Handbook Desk Edition. Vol.1. 1997: ASM International.
86