Professional Documents
Culture Documents
I. INTRODUCTION
Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:05 UTC from IE Xplore. Restricon aply.
TABLE I1
OF MAINDIFFERENCES
COMPARISON BETWEEN
GOLDAND COPPER
Be,Ca Ti,Yt, B
Ball shear (a) (b) (c)
Mode -ball shear Ball shear Ball shear Fig. 3. (a) Die cratering from copper wire bonding together with the picture
of a typical crater. (b) Before metal decoration for clarity. (c) After metal
Mode ull Neck break Neck break decoration for clarity.
Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:05 UTC from IE Xplore. Restricon aply.
NGUYEN et al.: OPTIMIZATION OF COPPER WIRE BONDING
7 TABLE I11
DEPENDENCY
OF PAD METALLIZATION
ON
MACHMETYPESAND DEPOSITION
CONDITIONS
2
E
* 3
d
M. Multiple
U
much Cu is piled-up against the AYSi interface, the bonding
.I
impact of the ball may force the Cu against the oxide, crack the
Fig. 4. Auger scans of Al-Cu (3%) pad metallization before and after underlying silicon, and lead to pad cratering. The latter may
annealing. occur at the onset of bonding or may be sufficiently subtle to
show up only after thermal cycling or thermal shock testing.
v. EFFECTOF METALDEPOSITION
SYSTEM
The nature of the pad metal also depends on the type of
sputterer system used (i.e., multiple wafer systems versus
single wafer systems). Deposition conditions such as rate,
temperature, time, pressure, gas composition, cool down time
and wafer load, all affect the grain structure obtained and
thus the pad metal hardness. Multiple wafer systems typically
have slow deposition rates, low deposition temperatures, and
long deposition times. In order for single wafer systems to
be comparable in throughput with multiple wafer systems,
i i2 is zi 2+
SPUrrERTIME ( m i d the reverse deposition conditions are true. These conditions
-Al-SiCu Si- can vary between different sputtering systems for similar
Fig. Auger scans of AI-Cu pad metallization at different Cu concentra- settings. Table I11 lists pad hardness values obtained for three
tions (2, 3, and systems (A, B, and C) compared with the control unit (D).
System A (hot) refers to system A run at a hotter deposition
Cu. A defocused electron beam is used to average the analysis temperature. Three die runs were used for data collection.
area to avoid variations due to the Cu grain distribution. Hardness values are normalized to the hardness of the samples
Before annealing, the pad shows little or no Cu in the first from the control System D ;each average value represents
half of the metallization. Most of the Cu is concentrated in three die runs. For each wafer, one die each was taken from
a broad peak before the AYSi interface. The annealed sample the top, bottom, left, right, and center. Five indentations were
exhibits a very different Cu profile. The Cu level becomes done on each of the four bond pads of each die. A response
more evenly distributed throughout the thickness of the pad surface methodology approach is needed to obtain the required
metal, with significantly less pile-up near the AYSi interface. metallization properties and the best deposition parameters.
Aside from annealing, the concentration of the Cu also Fig. 6 shows the enlarged views of the sectioned aluminum
dictates its solubility in the metallization. Fig. depicts such layer of samples retrieved from System A (normalized hard-
dependency for samples with three Cu concentrations: 2, 3, ness 0.74) and System C (normalized hardness Fig. 7
and 4 wt.%. shows the Cu distribution of these two samples as analyzed
At both 3% and 4%, the profiles for all the samples tested from a WDX machine. A high distribution of Cu near the
indicate a small concentration of Cu at the surface, which after surface of the metallization, as is the case for the sample of
approximately A drops down to a low level throughout (normalized) hardness translates into high pad hardness.
most of the A1 metallization. The Cu level increases toward Precaution should be taken in determining pad hardness
the AYSi interface, reaching a maximum just before or at the as measured by conventional microhardness testing. The test
interface boundary. The sample with 2% Cu differs slightly involves placing a diamond indenter of a specific geometry
in profile. The initial concentration of Cu near the surface onto the surface at a chosen load for a chosen duration. The
decreases and becomes more or less constant throughout the size of the indentation is then measured and translated into
A1 metallization without any pile-up at the AUSi interface. a hardness value. The load should be chosen such that the
The Cu distribution in the pad metallization is critical to depth of the indentation does not exceed the depth of the
the bonding quality. A high concentration of Cu at the surface metallization and thus enter the underlying oxide and silicon.
would raise the hardness of the pad and provide a damping Also, at lighter loads, the values registered may reflect whether
barrier to the copper ball. However, it was believed that if too the tip of the indenter is located directly on some Cu nodules
Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:05 UTC from IE Xplore. Restricon aply.
426 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY-PART A, VOL. 18, NO. 2, JUNE
(b)
Fig 6. Enlarged views of the aluminum layers of two representativebond pad samples of hardness 0.74 (a) and 1.13 (b) normalized
with respect to the control samples generated by System D.
Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:05 UTC from IE Xplore. Restricon aply.
NGUYEN et al.: OPTIMIZATION OF COPPER WIRE BONDING 421
TABLE IV
EVALUATION
MATRIX OF THE WIRE BONDING
PARAMETERS
FOR Al-Cu PAD METALLIZATION
i i i i i ;
Fig. 8. Hardness of film as a function of Cu concentration in AI film. Values
normalized to the hardness of the control film made with System D of Table
TABLE V
MAINEFFECTSAND INTERACTIONS RECORDED
FOR THE EVALUATIONOF Al<U (2%)
c’4
NORMAL POWER ENVELOPE Z-STAGE ENVELOPE COEFFICIENT IMPORTANCE
US.power stage 1 (US.1) Significant
US.power stage 2 2) Significant
US.1 pulse duration Not Significant
Total time Not Significant
US.1 power US.2 power Not Significant
US.1 power Total bond time Not Significant
BoNDlTME BONDTIME U.S.1 power U.S.1pulse duration Significant
US.2 power Total bond time Not Significant
(a) (b) US.2 power US1 pulse duration Significant
Fig. 9. (a) Power envelope used in normal bonding conditions. (b) Power Total bond time US1 pulse durati Not Significant
envelope used with the two-stage power supply.
Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:05 UTC from IE Xplore. Restricon aply.
428 IEEE TRANSACTIONS ON COMPONENTS. PACKAGING, AND MANUFACTURING TECHNOLOGY-PART A, VOL. 18, NO. 2, JUNE 1995
Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:05 UTC from IE Xplore. Restricon aply.
NGUYEN al.: OPTIMIZATION OF COPPER WIRE BONDING 429
Luu T. Nguyen, for photograph and biography, please see p. 22 of the March Anselm R. Danker (M’89) received the B.S. and
issue of this TRANSACTIONS. M.S. degrees in physics from the University of
Malaysia, in 1982 and 1985, respectively.
He joined National Semiconductor in 1987 as a
process engineer, and has worked on TO-220 copper
wire development since 1990.
Mr. Danker is a member of the Malaysian Insti-
tute of Physics.
Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:05 UTC from IE Xplore. Restricon aply.