Ultrasonic Bonding Understanding How Process Parameters Determine The Strength of Au-Al Bonds

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Proc. Intl. Symposium on Microelectronics, IMAPS, Denver, CO, USA, pp. 626-631, 2002.

Ultrasonic Bonding: Understanding How Process Parameters Determine the


Strength of Au-Al Bonds

Michael Mayer1, Jürg Schwizer2


1ESEC SA, 6330 Cham, Switzerland
2Physical Electronics Laboratory, ETH Zurich, 8093 Zurich, Switzerland

Abstract
The physics describing ultrasonic wire bonding is only partly known. A better understanding how bond strength
depends on process parameters could increase productivity by speeding up process ramp-up and increasing process
reliability. In this work, the influence of normal force and ultrasonic amplitude on the ball bond quality is investigated.
The determination of the ultrasonic friction energy dissipated at the bond interface during bonding is reported. The
analysis is based on real-time measurements of the ultrasonic tangential force FT(t) obtained in situ at the bond zone.
A state-of-the-art test chip containing integrated piezoresistive xyz-force microsensors is used for the measurements.
Considering the stick-slip behavior of a harmonically driven friction oscillator, several characteristic coefficients were
identified to determine the friction power during bonding at the interface: the ultrasonic frequency f, the normal
clamping force FN, the ultrasonic amplitude of the capillary tip measured in air A, and the compliance of the bonding
system c. The time dependent friction power P(t) is found to be P(t) = 4 f FT(t) [A - c FT(t) ]. The integral of P(t) over
time equals to the friction energy E. Experimental values of the friction energy density E / S are reported, where S is
the interface area. The friction energy density does not correlate well with the bond strength. Obviously, a varying part
of E is dissipated as heat and does not directly contribute to the bond. A better correlation with the bond strength is
obtained with the maximum γmax of a newly introduced parameter γ(t) = [ FT(t) - min( FT(t) ) ] / [ σ S - min( FT(t) ) ]
where σ is a measure for the yield shear stress and FT(t) is taken during friction only. The parameter varies from 0 for
no bond to 1 for a 100% complete metallic interconnection. The findings of this work support efforts towards auto-
matic process optimization on test chips and towards control methods suitable for the production floor.

Introduction contribute to successful interconnections, parameters


could be chosen suitable for the used materials and other
Wire bonding has been a time consuming process step in conditions.
microelectronic packaging. The setup of a wire bonding
process can be tedious and dependent on the skill of the In situ and real-time monitoring of temperature and
process engineer. Two examples of recent efforts in wire bonding forces enables efficient research of the physics
bonding are (1) bonding on novel low-k VLSI layer sys- during wire bonding. Such monitors usually are inte-
tems with low mechanical robustness and (b) using Cu grated sensors on a test chip [1-13]. In the following, the
wire bonds on Cu or Al pads. A common strategy to application of a previously reported xyz-force sensor test
address such bonding challenges is to carry out a design- chip for ball bond quality determination is reported.
of-experiment (DOE) to systematically scan through the
input parameter space followed by determining a Test Chip with Microsensors
response surface and the parameter window. If the wire
Sets of five test chips are die bonded on custom made
material, the chip layer system or another significant fac-
polyimide substrates [12] with Au metallization on one
tor changes, the DOE usually should be repeated.
side as shown in Fig. 1. The test chips were produced
Speeding-up process development and setup aims at using a commercial CMOS process. The design of the
reducing the efforts for DOEs. This could be achieved chip includes a total of 48 xyz-force sensors, placed
by introducing an automated method to measure the pro- along the four chip edges at two different distances to
cess responses combined with a real-time feedback to the edges, as shown in Fig. 2. The xyz-force sensor ele-
the wire bonder. Another method could be the applica- ments are the piezoresistive CMOS source/drain diffu-
tion of an efficient physical model of the ultrasonic sions, both p+- and n+-doped, placed around a standard
bonding process. By understanding the mechanisms that pad for the test bond. A multiplexing circuit is co-inte-
grated with the sensors on the chip. Power supply lines, at temperatures lower than needed for interdiffusion,
address lines, and sensor signal lines are connected via i. e., about 175°C.
wire bonds to pads at the four chip corners. The
The ultrasonic parameter is the amplitude of the freely
microsensors were calibrated using a conventional shear
vibrating capillary tip A as schematically shown in Fig. 3
tester, and a sensitivity of 96 mN/mV/V was found [11].
(a). Values used for the following investigations are var-
ied from 0.119 to 0.359 µm. The time of ultrasound is
Experimental 10 ms. A 25 µm diameter AW-14 wire is used. The free-
The substrates are loaded on an automatic wire bonder air ball diameter is 45.2 µm. The impact force is
with a modified clamping plate. A contact finger assem- 500 mN, and the bonding force FN is varied from 50 to
bly fixed to this plate connects the measurement equip- 450 mN. In particular, the chip temperature is set to
ment via the substrate metallization and the connection 35°C and the impact force is chosen relatively high in
wire bonds to the test chip. The xyz-force sensor on the order to avoid additional thermal and ultrasonic defor-
test chip proves useful for monitoring stick-slip motions mation effects. The low chip temperature does not allow
between ball and pad [8, 12]. Such a friction welding bonds with much gold remaining on the pad after shear
mechanism is used for ultrasonic bonding of Au and Al testing. This is used to readily measure the actual contact
zone diameter after ball removal. The wire and capillary
are suitable for an 80 µm pad pitch ball bond process.
The transducer frequency and impedance are 130 kHz
and 8.3 Ohms, respectively. The second bond is opti-
mized for small free-air ball diameter variations.
3 mm
Various FN-A parameter combinations are investigated
in greater detail. Balls were bonded in six experiments
with parameters along six lines in the FN-A plane. The
six experiments are denoted with letters A-F and defined
as shown in Table 1.

(a) (b)

BDC
Capillary
f = 130 kHz BH
BDI
Fig. 1 Microsensor test chip on substrate.
A

Fig. 3 Schematic of (a) freely vibrating capillary, (b) cross-


section of bonded ball with definitions of geometry
parameters.

Table 1: Definition of six experiments.

Parameter
A [µm] FN [mN]
Experiment
A 0.119-0.355 100
B 0.119-0.355 200
1 mm

C 0.119-0.355 300
D 0.178 50-450
E 0.237 50-450
Fig. 2 Microsensor test chip with wire bonds after measure-
ment. F 0.296 50-450
Process Characterization A(FN) = c µ0 FN (3)

To characterize the ball geometry, the ball height (BH), using Amonton’s Law of friction
ball diameter at capillary (BDC), and the contact zone
FT = µ0 FN (4)
(interface) diameter (BDI) are measured as schemati-
cally shown in Fig. 3 (b). The BDI values are obtained and
after shearing the balls off the pad. The shearing is done
towards the direction of ultrasound. The BDI then is
obtained by measuring the imprint diameter perpendicu- 30
(a) (b)
lar to the ultrasound direction. The experimental values F
for BDC and BDI are given in Fig. 4, those for BH, shear

Shear Force [cN]


force (SF), and shear strength (SS) are given in Figs. 5-7, 20
E
respectively. The shear strength SS is defined as C
D
A
SS = SF
------ , (1) 10 B
S

where
0
2 0.15 0.2 0.25 0.3 0.35 100 200 300 400 500
S =  BDI
---------- ⋅ π . (2) A [µm] FN [mN]
 2 
Fig. 6 Experimental values for SF for various parameter
The results are summarized in the FN-A plane as shown combinations (a) A-C and (b) D-F.
in Fig. 8. Two characteristic lines can be identified:
150
(1) Line of friction, defined by (a) (b)
Shear Strength [MPa]

70 100 F
(a) (b)
65 C E
A
F 50 D
Diameter [µm]

C
60 B
BDC B BDC
55 E
A 0
0.15 0.2 0.25 0.3 0.35 100 200 300 400 500
D A [µm] FN [mN]
50
BDI Fig. 7 Experimental values for SS for various parameter
BDI
45 combinations (a) A-C and (b) D-F.
0.15 0.2 0.25 0.3 0.35 100 200 300 400 500
A [µm] FN [mN]
Fig. 4 Experimental values for BDC and BDI for various
parameter combinations (a) A-C and (b) D-F. 0.40
Lin Excessive
e of deformation
15 De
(a) (b) for
0.30 m atio
a
A D MP n
E 100
Ball Heigth [µm]

10
A [µm]

B 0.20
C F
Ball non-sticks
5 t io n
0.10 ric
fF
eo
L in
0 0.00
0.15 0.2 0.25 0.3 0.35 100 200 300 400 500
A [µm] FN [mN] 0 100 200 300 400 500
Bonding force [mN]
Fig. 5 Experimental values for BH for various parameter
combinations (a) A-C and (b) D-F. Fig. 8 Experimental results in the FN-A plane.
A = c FT (5) Shear strengths of 100 MPa are obtained using FN-A
parameter combinations as shown on the corresponding
where c is the compliance of the bonding system. For line in Fig. 8. Dashed lines are estimations.
FN-A parameter combinations with A < c µ0 FN , only
ball non-sticks are observed. No bonding occurs in this
case as interfacial sliding is inhibited according to
Microsensor Results
Amonton’s Law of friction. For each of the parameter combinations, the average of 6
(2) Line of Deformation, defined by to 10 microsensor signals FT(t) and 3rd harmonics are
determined. Such signals have been described in detail
A(FN) = A(0) - cD FN (6) in [4, 7, 13]. Some averaged FT(t) curves of experiments
A-C and D-F are shown in Figs. 9 and 10, respectively.
where the parameter cD describes the slope of the line of
All curves are shown for the period between ultrasound
deformation. For FN-A parameter combinations with
start and stop. For most of the curves, a break-off is
A(FN) > A(0) - cD FN , a strong increase of the BDC
observed shortly after the start of ultrasound, corre-
accompanied by a decrease of the BH is observed.

A B C
300
A [µm]: 0.355
A [µm]: A [µm]:
0.355 0.355 0.296
0.296
0.296
200 0.237 0.237
FT [mN]

0.237 0.178 0.178

100 0.178 0.119 0.119

0.119

0
0 5 10(0) 5 10(0) 5 10
Time [ms]
Fig. 9 Microsensor results: FT(t) for experiments A-C.

D E F
300

FN [mN]: FN [mN]: FN [mN]: 250 350 450

200 250 350 450


150
FT [mN]

250 350 450


50

100 150
50
150
50
0
0 5 10(0) 5 10(0) 5 10
Time [ms]
Fig. 10 Microsensor results: FT(t) for experiments D-F.
sponding with a simultaneous rise of the 3rd harmonic. t US
This rise is used to numerically determine the break-off
value, i. e., the start of friction.
E = ∫ P ( t ) dt , (12)
0

where tUS is the duration of ultrasound. Results for the


Friction Power and Energy
maximum power density and the energy density are
The formula for the friction power produced at the ball/ given in Figs. 12 and 13, respectively.
pad interface is
P(t) = 4 Arel(t) f FT(t) , (7) Degree of Bond Growth
The parameter γ is introduced as a measure for the bond
where 4 Arel is the relative sliding distance during one
growth during friction. Especially, γ = 0 if there is no
ultrasonic vibration period and f = 130 kHz is the ultra-
bond at all, and γ = 1 if there is a complete metallic
sonic frequency. Laser measurements of the capillary tip
bond. The tangential force FT(t) is denoted FT, friction(t) if
amplitude during bonding with various FN showed that
only the friction period is taken. It can be expressed
the amplitude behaves like
using the approach
A(FT) = A(0) - c1 FT (8)
FT, friction(t) = [ 1 - γ(t) ] · min[ FT, friction(t) ] + γ(t) σ S (13)
until A(FT) equals the pad surface amplitude. The pad
where σ S = Fmax is the yield force of the bond and σ is a
surface behaves like
measure for the yield stress. Equation (13) can be solved
Apad(FT) = c2 FT (9) for γ(t):
Thus, the relative sliding amplitude is F
T, friction ( t ) – min [ F T, friction (t)]
γ(t) = -------------------------------------------------------------------------------
- (14)
σS – min [ F T, friction ( t ) ]
Arel(FT) =A(FT) - Apad(FT) = A(0) - c FT (10)

where c = c1 + c2 is the compliance of the bonding sys-


Max. Power density [W/mm2]

tem. The friction power becomes (a) (b)


3
P(t) = 4 f FT(t) [ A - c FT(t) ]. (11)
Using (4) and the experimental result in Fig. 11, 2 F
µ0 = 0.38 is found. The compliance is found to be
c = 1.7 µm/N by using (5) and the experimental values E
1
of µ0 and the slope of the line of friction in Fig. 8, A C D
µ0 c = 0.66 µm/N.
B
The densities of the friction power P(t) / S and the fric- 0
0 0.1 0.2 0.3 0 100 200 300 400 500
tion energy E / S can now be determined using the A [µm] FN [mN]
microsensor signals for each parameter combination and Fig. 12 Evaluated maximum friction power density for ball
bonds using various parameters.

6
(a) (b)
Energy density [mJ/mm2]

0.2
D
min[FT, friction(t)] [mN]

4
0.15

0.1
2 F
E
0.05 A
Slope: µ0 = 0.38 C
B
0
0 0 0.1 0.2 0.3 0 100 200 300 400 500
0 0.2 0.4 0.6 A [µm] FN [mN]
FN [mN]
Fig. 13 Evaluated friction energy density for ball bonds using
Fig. 11 Minimum friction force vs. normal force. various parameters.
For a quantification of the bond quality, the maximum References
value for γ(t) is a candidate. Experimental values are
[1] M. Mayer, O. Paul, and H. Baltes, “In-situ Measurement
obtained after choosing an appropriate value for σ. A
of Stress and Temperature under Bonding Pads During
lower boundary is 70 MPa. Values below this boundary Wire Bonding Using Integrated Microsensors”, Proc. 2nd
yield γ > 1 for some t. This is beyond the initial assump- Int. Conf. Emerging Microelectr. and Interconn. Technol.
tions. During shear testing the pad aluminum is yielding. EMIT’98, IMAPS, Bangalore, India, pp. 129-133, 1998.
For aluminum alloys, σ = 90 MPa is a possible shear [2] M. Mayer, O. Paul, D. Bolliger, and H. Baltes, “In-Situ
strength [14] and therefore chosen to calculate the maxi- Calibration of Wire Bonder Ultrasonic System using Inte-
mum values of γ(t), γmax , using (14). They are given in grated Microsensor”, Proc. 2nd IEEE Electr. Packaging
Technol. Conf. EPTC’98, Singapore, pp. 219-223, 1998.
Figs. 14 (a) and (b) for the experiments A-C and D-F,
[3] M. Mayer, O. Paul, D. Bolliger, and H. Baltes, “Integrated
respectively. Temperature Microsensors for Characterization and Opti-
mization of Thermosonic Ball Bonding Process”, IEEE
Conclusions Trans. Comp. Packaging Technol., vol. 23, no. 2, pp. 393-
398, 2000.
The classical experimental results describe the influence [4] M. Mayer, J. Schwizer, O. Paul, D. Bolliger, and H. Balt-
of normal force and ultrasonic amplitude on the ball es, "In-Situ Ultrasonic Stress Measurements During Ball
bond quality. The reported model approach enables to Bonding using Integrated Piozoresistive Microsensors,"
evaluate a parameter γmax from the microsensor tangen- Proc. 1999 Intersociety Electron. Pack. Conf.
(InterPACK99), Maui, Hawaii, pp. 973-978, 1999.
tial force signals FT, friction(t). This parameter is in good
[5] M. Hizukuri and T. Asano, “Measurement of Dynamic
qualitive correspondence with the ball shear strength. Strain during Ultrasonic Au Bump Formation on Si Chip”,
Experimental values of the friction power and energy are Jpn. J. Appl. Phys., vol. 39, pp. 2478-2482, 2000.
evaluated using the microsensor measurements. These [6] M. Hizukuri, Y. Wada, N. Watanabe, and T. Asano, “Real
Time Measurements of the Strain Generated on Substrate
values do not correlate with the ball shear strengths.
during Ultrasonic Flip Chip Bonding”, 6th Symp. Micro-
A more complete understanding of how the normal force joining and Assembly Technol. in Electronics 2000,
and ultrasonic amplitude influence the bond strength is Yokohama, pp. 169-174, 2000.
gained if the model approach could be extended with a [7] J. Schwizer, M. Mayer, D. Bolliger, O. Paul, and H. Balt-
es, “Thermosonic Ball Bonding: Friction Model Based on
model function for FT, friction(t). The basic assumption for Integrated Microsensor Measurements”, Proc. 24th IEEE/
such a function could be µ· ∝ P . CPMT Intl. Electronic Manufacturing Technology Sym-
posium IEMT’99, Austin, Texas, Oct. 18-19, pp. 108-114,
Acknowledgments 1999.
[8] J. Schwizer, M. Mayer, O. Brand, and H. Baltes, “Analy-
This work was partially funded by the Swiss Federal Pri- sis of Ultrasonic Wire Bonding by In-situ Piezoresistive
ority Program MINAST (Micro & Nano System Tech- Microsensors”, Proc. Transducers ‘01 / Eurosensors XV,
nology) and the Swiss Federal Commission of pp. 1426-1429, 2001.
Technology and Innovation (CTI). [9] J. Schwizer, M. Mayer, O. Brand, and H. Baltes, “In situ
Ultrasonic Stress Microsensor for Second Bond Charac-
terization”, Proc. Intl. Symp. Microelectronics IMAPS
2001, pp. 338-343, 2001.
[10] S. Suman, M. Gaitan, Y. Joshi, G. Harman, “Wire Bond
Temperature Sensor”, Proc. Intl. Symp. Microelectronics
IMAPS 2001, pp. 344-349, 2001.
1
[11] M. Mayer and J. Schwizer, "Wire Bonder Ultrasonic Sys-
(a) (b)
tem Calibration Using Integrated Stress Sensor", Proc.
0.8 SEMI Technical Symposium, Advanced Packaging Tech-
F nologies II, SEMI Singapore, pp. 169-175, 2002.
0.6 [12] J. Schwizer, Q. Füglistaller, M. Mayer, Michael Althaus,
γmax [−]

C O. Brand, and H. Baltes, "MEMS System with Multiplex-


0.4 er for In Situ and Real-time Wire Bonding Diagnosis",
A D
E Proc. SEMI Technical Symposium, Advanced Packaging
0.2 B Technologies I, SEMI Singapore, pp. 163-167, 2002.
[13] M. Mayer, Microelectronic Bonding Process Monitoring
0 by Integrated Sensors, Ph.D. thesis, No. 13685, ETH Zu-
0.15 0.2 0.25 0.3 0.35 100 200 300 400
A [µm] FN [mN] rich, Zurich, 2000; also: Hartung-Gorre, Konstanz, Ger-
many, ISBN 3-89649-620-4, 2000.
Fig. 14 Evaluated quality parameter values for experiments [14] http://www.matweb.com
A-C (a) and D-F (b).

You might also like