Large Cu Wire Wedge Bonding Process for Power Devices

Jamin Ling, Tao Xu, Christoph Luechinger

Kulicke & Soffa Industries, Inc., Wedge Bond Business Unit
16700 Red Hill Ave., Irvine, CA 92606 USA

jling@[Link]

Abstract size, with limitations to optimize bond quality. A Cu-only

Due to the high cost of gold, Cu-wire bonding is being bond head was therefore developed.
used more and more on fine-pitch, high-pin-count IC devices. Table 1 lists the specifications of the new, rear-cut Cu
Medium-size Cu wire (>1mil) has been used for bonding on bond head, which was used on K&S’s Orthodyne 3600Plus
power devices for years but has not become a popular bonder throughout the study mentioned in this paper.
interconnect option.
Recently, however, successful bonding of large (16-mil) Table 1: Cu Bondhead Specification
Cu wire on specifically metallized IGBTs was reported [1, 2]. 3600Plus Large Al Wire 3600Plus Large Cu Wire
Some of these devices showed 10 times the power-cycling
Bondhead Type Front-cut/Rear-cut Rear-cut ALC
life-span of standard Al wire bonded to Al die metallization.
Three force ranges:
In order to develop and make this process production Programmable 50 to 1,500gf
300 to 2,200gf
capable, K&S has developed a new Cu bond head that delivers Bond Force Range
1,500 to 3,500gf
bond forces, wire clamp forces, and ultrasonic power that are 3,000 to 5,000gf
significantly higher than those typically achieved with Al
Transducer Type 60kHz or optional 80kHz 80kHz High Power
wire. This rear-cut head is compatible with most current
power-module designs. A modified V-groove bond tool for Digital Generator 65 watt 150 watt
use with the new head enables bond shear and pull strengths Power Output
that are 1.0 and 1.5 times the wire’s tensile strength Wire Clamp Force 500gf maximum 1,500gf maximum
respectively.
Wire Size <12mil /300um (Cu) 20mil /500um (Cu)
20mil /500um (Al)
Introduction
Large Cu-wire bonding on a Cu substrate was first studied
and shown to be possible in 1984 using an Orthodyne Model
20 bonder with 2-mil Cu wire on 2-oz. Cu-plated PCB [3].
Bondability was reported as highly dependent on the quality
of the substrate (and the wire). It was shown that bond
parameters must be optimized at lower power settings, as bond
lifts and heel cracks resulted at higher settings.
As demand for longer-lasting, higher-density power
devices (mainly power modules) grows, bonds that are highly Figure 1: Rear-cut Cu Bondhead with 5,000gf Bond Force,
conductive, carry high levels of current, and transfer heat 1,500gf Clamp Force (left); Bonding on Cu Metallized Wafer;
efficiently are essential to the reliable operation of a device.
Following the significant success of the improvements of die
Bondability tests were conducted using 8- to 20-mil Cu
attach heat conductivity [4] and the revolution of Cu die
wire on Cu plates, DBC, and Cu metallized wafers. Cu wire
metallization [5], a critical link now becomes the die-substrate
from several sources and of various tensile strengths and
interconnect, namely, wire bonding.
elongations was used for comparison. The Cu plate used in the
In the first phase of an extended feasibility study, 12- and
study was a 40-mil thick flat plate typically used for bonder
20-mil 4N uncoated Cu wire was bonded to direct-bonded
setup before experiments. DBC with a 12-mil, fine-grain Cu
copper (DBC) substrates, solid bare Cu plates, and silicon
layer was used as the primary test vehicle to compare and
wafers with various in-house-developed metallization
assess the bondability of large Cu wire. Cu-top wafers with
structures. A broad set of bond-quality criteria was developed
various under-layers deposited on a Si wafer with a thin SiO2
based on experiences with Al wire bonding. Bond-process
layer were also used to assess their sturdiness and bondability
windows were then generated to determine the needed
with large-Cu wire.
equipment specifications for large-Cu wire bonding.
The geometric structure of a bond tool’s tip greatly
influences bond quality, which in turn, affects bond width
Experiments (deformation severity), pull strength, and shear strength
A standard large wire bond head on a K&S’s Orthodyne (Figure 2). A modified bond tool (SG-Std) provided a much
3600Plus bonder was used for the first phase of the studies lower standard deviation and somewhat higher pull and shear
[6]. This bond head enabled Cu-Cu bonds up to a certain wire strengths compared to a standard bond tool (Std). Hence, the
SG-Std bond tool was used for this analysis.
978-1-4577-1982-0/11/$26.00 ©2011 IEEE 2011 13th Electronics Packaging Technology Conference

1
 Pull force comparison: SG-Std & Std tools
6000.0

5500.0

5000.0

4500.0
Pull force (g)

4000.0

3500.0

3000.0
SG-Std tooll
2500.0 Std tool
Figure 4: Shear Mode with Nugget Coverage Comparison
2000.0
3000 4000 5000 6000 7000 8000 9000
Bond force (g) Though the bonding process window could have reached a
(A) Pull Strength bond force of 8,250gf before the bond over-deformed, optimal
window was achieved up to 5,250gf. The bond pull strength
Shear force comparison: SG-Std and Std tools matched the tensile strength of the wire, and the shear strength
14000.0
was over 1.5 times the tensile strength, which was a general
13000.0
target for the bond-quality index. The failure mode of the pull
12000.0
strengths showed a mid-span break in the loop. The sheared
11000.0
interface could be reasonably covered by the nugget (the
Shear force (g)

10000.0
bonded portion of wire that remains after shearing), which
9000.0
implied a stronger bond interface relative to the wire yield
8000.0

7000.0
strength. However, whether or not this nugget coverage (shear
6000.0
remains) should be included in the bond-quality index needs
SG-Std tool
5000.0 Std tool to be resolved because Cu wire responds differently than Al or
4000.0
Au wire to compression stress on a given substrate during
3000 4000 5000 6000 7000 8000 9000 bonding.
Bond force (g)
The bond-force comparison (using various wire sizes) of
(B) Shear Strength Cu wire to Al wire is shown in Figure 5. For 8- to 20-mil
sizes, the required bond force for Cu wire (on Cu substrate)
Figure 2: Pull Test Comparison with Different Bondtools for was about 50% higher than its tensile strength, and in turn, its
20mil Wire Bonding
yield strength. Bond force for Al wire (on Al substrate),
however, was just about 30% higher than its tensile strength.
Results and Discussions
1) Cu Wire Bonding Process Window
The bonding process window was examined in a design of
experiments (DOE) by first screening the ultrasonic power of
each wire size at a given bond force to define the experimental
ranges. The DOE was then conducted to establish a workable
bonding window, which conformed to the pre-defined criteria
(pull strength, shear strengths and their failure modes, bond
width, and other bond visual defects). The bonding window
for 20mil Cu wire on DBC is shown in Figure 3; the sheared
interface is shown in Figure 4.

Figure 5: Bonding Force Relative to Tensile Strength

2) Copper Vs. Aluminum Wire Bonding

Generally, Cu has 3 times the density of Al, 1.7 times the
modulus, 2 times the tensile strength for fine wires, and more
than 2 times the tensile strength as wire size increases. Under
compression during bonding, it would yield more slowly,
needing greater force to deliver equal deformation. Depending
on its properties, the substrate could be yielded accordingly
under stress. Therefore, the severity of compression into the
Figure 3: Bonding Window for 20mil Cu Wire on DBC substrate could differ with the work hardening of the wire.
Figure 6 shows two wire samples of different hardness.
With a 300mN test load, the Hv (Viker scale hardness) of the
wire’s core was compared to that of the wire next to the bond.

2011 13th Electronics Packaging Technology Conference

2
An estimated 13% increase in hardness was noted from the
strain of bond deformation, while the DBC hardness was
>50% from compression by the wire. A softer wire (Sample
2), paired with a harder substrate produced less substrate
deformation than Sample 1 (20μm vs. 48μm compression into
the DBC’s Cu substrate).

Figure 8: Pull and Shear Strength Comparison of Cu

wires and Al wires in Various Sizes

3) Bond Process Development

The bond process is critical to improving Cu-Cu bonding.
Other testing (using 12-mil Cu wire on Cu substrate) that
Figure 6: Wire and DBC Material Properties for Bonding included plasma cleaning and force ramping, was performed
to determine their effectiveness.
In this case, depending on the shear height, the wire core As shown in Figure 9A, shear strengths increased and the
may be stronger than the bond with less nugget remaining bond window widened significantly when 5 min. of Ar + (5%)
after shearing. Therefore, unlike Al-wire bonding, Cu bond H2 plasma cleaning was applied to a Cu substrate prior to
quality should not necessarily be based on nugget coverage. bonding. Many studies on the effects of plasma cleaning on
Pull strength vs. wire deformation (strain hardened) in Cu wire bonding have been performed [7, 8] in which the oxide
wire was shown in Figure 7. The Cu deformation severity was layer or other contaminants were removed to enhance bonding
defined as the ratio of deformed bond width to wire diameter. efficiency. Transducer current traces, which showed coupling
Over-deformed Cu wire bonds >1.7 times the wire diameter between the bond tool and the Cu wire as well as its
had lower pull strengths due possibly to weakened heels, deformation profile, were consistent as shown in Figure 9B.
while under-deformed bonds <1.3 times the wire diameter
would have insufficient support strength during pulling. Shear strength comparison: cleaned vs. uncleaned substrate
95% CI for the Mean
C leaning
0.9 N
Pull strength vs. wire deformation
Normalized shear strength

Y
2500 0.8
0.7
2400 0.6
0.5
2300
0.4
2200 0.3
0.2
Pull strength (g)

2100
0.1
2000 0.0
Cleaning N Y N Y N Y N Y N Y N Y
1900 Time (ms)
Power
40
55
60 40
55
60 40
75
60
(A)
1800 Force (g) 1000 1500

1700

1600

1500
1.2 1.25 1.3 1.35 1.4 1.45 1.5 1.55 1.6 1.65 1.7
Wire deformation (bonded wire width / wire diameter)

Figure 7: Copper Wire Working Window

(12mil Cu Wire on DBC)

Al wire pull- and shear-strength results were lower than

those of Cu wires for all tested sizes. The larger the wire, the
higher the strength deviation was. Figure 8 shows Cu-wire
pull strengths were generally 60-78% higher and shear (B)
strengths 64-69% higher than Al wire, correlated with a 70%
increase of the Young’s modulus of Cu from Al.

Figure 9: Plasma Clean Effects on Cu-Cu Bonding; (A) Bond

Shear Strengths; (B) Transducer Current and Z-Deformation

2011 13th Electronics Packaging Technology Conference

3
 Also, as shown in Figure 10, force-ramping experiments 90° forced-angle bonding was performed using 12-mil Cu
were performed for Cu-Cu bonding. Increasing the force- wire on a DBC substrate at full bonding speed. Figure 12
ramping range by adjusting the ultrasonic power, power shows these forced-angle bonds, which were made using a
ramping, and bond time, appeared to increase bond shear touch force of 2,850gf and a reduced twist height of 500μm.
strengths. However, the standard deviation of the shear
strengths widened as the strengths increased.

Shear strength for different force ramping

4000.0

3800.0 1st bond shear strength

3600.0 2nd bond shear strength
3400.0
Shear strength (g)

3200.0

3000.0

2800.0

2600.0
Figure 12 90° Forced-angle Bonding with 12mil Wire
2400.0

2200.0
c. Cu Metallization Wafer Bonding
2000.0
50g-1500g 500g-1500g 1000g-1500g 1500g-1500g Cu-wire bonding on an active device as opposed to a
Force ramp range passive DBC or Cu plate may be difficult. A bond pad’s
Figure 10 Force Ramp Effects on Cu-Cu Bonding design and its metallurgical ability to accept Cu bonds are
crucial factors. Pad-structure management has been written
4) Other Developments about in many publications for Cu wire bonding, probing, and
a. Looping Au wire bonding on the low-k dielectric structures. [9, 10, 11,
The looping capabilities of large (12mil) Cu wire were 12]
also investigated. Figure 11 shows a stitch-loop profile for a Because large Cu-wire bonding is greatly influenced by
first loop with a step-back of 1.6mm and a loop height of improvements in reliability, a Cu die-pad metallization was
0.9mm followed by a second loop with a step-back of 3.5mm needed to accept the Cu wires. Therefore, creating a robust Cu
and a loop height of 1.9mm. All first, stitch, and last bonds die-pad structure and metallization may be the key to
were sheared to determine their strengths. Equivalent shear successful bonding. To date, much work has been done to
strengths were achieved for all three groups of bonds, develop such a die pad, but little success has been reported in
revealing no degradation due to the stitch bonding. the market or technical literature. K&S, however, has shown
that bonding 12- and 20-mil Cu wire to Cu-topped metal
wafers is feasible, simulating pad metallization without
Shear strength comparison for different bonds pattern, as shown in Figure 13. Efforts to develop a robust
6000 metal pad structure are ongoing.
5000
Shear strength (g)

4000

3000

2000

1000

0
1st bonds 2nd bonds Stitch bonds
Bonds
Figure 13: Cu Wire Bonding on a Cu Metallization Wafer

Conclusions
The study yielded these conclusions:
• A Cu bond head with a bond force of 5,000gf and a clamp
force of 1,500gf successfully bonded Cu wire sizes up to
20-mil on Cu-metallized substrates (DBC, plates, wafers).
• Bonding windows for 8-, 12-, and 20-mil Cu wires on
DBC, Cu plates, and Cu-metallized wafers were defined
according to industrial general requirements per Al wire
Figure 11 Stitch Wire Bonding Using 12mil Wire
bonding experience.
b. Force Angle Bonding
2011 13th Electronics Packaging Technology Conference

4
• Bond force was proportional to wire tensile strengths. A
50% higher tensile strength was needed for Cu-wire
bonding as opposed to 30% higher for Al-wire bonding.
• The Cu-wire (deformation) working window of 1.3 to 1.7
times the wire diameter was observed for 12mil Cu wire
bonding on DBC. Over-deformed bonds reduced the pull
strengths, while under-deformed bonds did not stick, so the
bond lifted.
• Pre-bonding plasma cleaning on a Cu substrate and force-
ramp bonding may increase the shear strengths of Cu-Cu
bonding.
• Forced-angle bonding and stitch-bond looping were
successfully tested for specific applications of Cu-wire
wedge bonding.

References
1. P. Luniewski, K. Guth, D. Siepe, ‘Cu Bonds and Chip-to
Substrate Joints Beyond Silver Sintering’, Bodo’s Power
Systems, ([Link]), Aug 2010, pp.32-33;
2. D. Siepe, R. Bayerer, R. Roth, ‘The Future of Wire
Bonding Is? Wire Bonding!’, Proceeding of CIPS 2010,
Mar 16-18, 2010, Nuremberg, Germany, Paper 3.7;
3. J. Ling, C.E. Albright, ‘The Influence of Atmospheric
Contamination on Copper to Copper Ultrasonic Welding’,
Proc. ECTC, 1984, pp.209;
4. G. Lu, J.N. Calata, Z. Zhang, J.G. Bai, A’ lead-free, low-
temperature sintering die-attach technique for high-
performance and high-temperature packaging’, Proc.
IEEE HDP, 2004, pp.42-46;
5. C. Chen, S. Zhang, S. Lee, L. Mohamed, ‘Investigation
on copper diffusion depth in copper wire bonding’,
Microelectronics Reliability, 51 (2010), p.166-170;
6. KNS Orthodyne 3600 Plus Large Wire Bonder Bulletin
36-305;
7. R. Rust, D. Doane, I. Sawchyn, ‘Improvements in wire
bonding and solderability of surface mount components
using plasma cleaning techniques’, IEEE Trans. CHMT,
Vol.14, No.5, 1991, pp.573;
8. J.M. Nowful, S.C. Lok, S. Lee, ‘Effects of plasma
cleaning on the reliability of wire bonding,‘ IEEE Int’l
Symp. Electronic Material and Packaging, 2001, pp.39;
9. J. Rohan, G. O’Riordan, J. Boardman, ‘Selective electro-
less nickel deposition on copper as a final barrier bonding
layer material for microelectronics applications’, App
Surf Sc 185, 2002, 289-297;
10. H. Clauberg, J. Ling, A. Hashmonai, T. Thieme, B.
Chylak, ‘Nickel-Palladium Bond Pads for Copper and
Gold Wire Bonding’, Semi Semicon West, Tech. Symp.
2009;
11. Y. Takahashi, M. Inoue, ‘Numerical Study of Wire
Bonding—Analysis of Interfacial Deformation Between
Wire and Pad’, ASME J. Elect. Pkg. Vol.124, 2002,
pp.27;
12. D. Degryse, B. Vandevelde, E. Beyne, ‘Mechanical FEM
Simulation of Bonding Process on Cu Low-k Wafers’,
IEEE Trans. Component and Packaging, Vol.27, No.4,
2004, pp. 643.

2011 13th Electronics Packaging Technology Conference