A New Mechanism for Hillock Formation over Electrodeposited Thin Tin Film
Jing Cheng
1
, Samuel Chen
2
, Paul Vianco
3
and James C.M. Li
1
1
Materials Science Program, University of Rochester, Rochester, NY 14627
2
Eastman Kodak Company, Rochester, NY 14559
3
Sandia National Laboratories, Albuquerque, NM 87185
Email:jicheng@me.rochester.edu, Phone:1-585-489-5302
Abstract
Tin film about one micron thick was electroplated over
a silicon wafer pre-coated with a layer of Cr and another
layer of Ni by evaporation. A special sample holder was
designed to apply compressive stresses in the electroplated
tin film. After incubation in a vacuum oven at 160
o
C for 7
days, huge hillocks about 10-30µm diameter and 30-150µm
length grew with a density about 7 per square mm. On the
top of the hillocks there appeared a polycrystalline layer
similar to the tin film. FIB technique was used to reveal the
inner microstructure of these huge hillocks. They are single
crystals of Sn with [001] direction perpendicular to the film.
The flow path of tin atoms appears to originate from the
stressed Sn/Ni interface moving along the interface radially
toward the root of the hillock.
Key Words: tin whiskers, tin hillocks, growth mechanism,
flow path, FIB
Introduction
In microelectronic industry, tin whiskers growing from
tin coatings has been an important problem for many years.
The literature was reviewed recently by Galyon and Palmer
[1]. The problem was mitigated by electroplating tin-lead
(Sn with 3-10% Pb by weight) alloys instead of pure tin by
Arnold [2] in1959 even though to this day we still do not
know clearly why it worked. However in recent years,
some environmental related regulations such as Restriction
of Hazardous Substances (RoHS) and Waste Electrical and
Electronic Equipment (WEEE) listed Pb as a toxic element.
In fact Europe has not allowed lead-containing solders since
July 1, 2006. So this old problem of tin whiskers surfaced
again.
Some landmark discoveries were as follows: Koonce
and Arnold in 1954 [3] showed that the growth was from
the base and not from the top by observing that the tip
morphology was unchanged while the whisker grew longer.
Fisher et al. [4] in 1954 found that a compressive stress of
52 MPa could accelerate the whisker growth to 1 micron
per sec at room temperature and the rate was too fast for
lattice diffusion. Baker [5] in 1957 showed that the base
of the whisker was usually a grain boundary or an interface
boundary inconsistent with a dislocation mechanism.
Lindborg [6] in 1975 used x-ray to determine stresses in
electroplated Zn films and showed that a minimum stress
level was required for the initiation of whisker growth.
Not much activity appeared in the 80’s and 90’s. Tu [7]
in 1994 suggested that the oxide layer surrounding the
whisker must be broken to allow the whisker to grow. Lee
and Lee [8] in 1998 measured the residual stresses in Sn
film on Cu substrates and found 11MPa tensile as deposited
which changed to -8 MPa compressive after storage about 5
days.
They also determined the thickness effect, 60 whiskers per
mm
2
for a thickness of 1 micron decreased continuously to 20
whiskers per mm
2
for a thickness of 4.5 microns. In 2001
Zhang et al [9.10] used FIB and found that the root of the
whisker was close to intermetallic particles projecting
upwards from the tin/substrate interface. Later [11,12] they
studied the effect of a Sn/Ni interface suggesting that Sn
diffuses into Ni faster than Ni into Sn. Choi et al [13,14] used
micro-focus x-ray from a synchrotron facility and found that
the stress surrounding the whisker area was more compressive
than the whisker area and the whisker grain had a different
orientation (210) than the surrounding grains (321).
Boettinger et al. [15] measured and analyzed the residual
stress development in Sn, Sn-Cu and Sn-Pb layers
electrodeposited on phosphor bronze cantilever beams and
observed whisker and hillock formation. Tu and Li [16]
introduced the idea that grain boundary fluid flow may be
faster than grain boundary diffusion. This idea comes from a
comparison between self diffusion and viscous flow in liquids
[17].
More recently Woodrow [18] did a delicate experiment
by plating one layer of isotope tin
118
on brass coupon, then
covered by another layer of isotope tin
120
. The area of tin
120
was larger than that of tin
118
, so in the middle there were two
layers surrounded by a single layer of tin
120
. The two layers
share the common columnar grains so there were no grain
boundaries between the two layers. After some time, in the
whiskers grown in the single layer area, tin
118
was also found
there. Since almost all the grain boundaries were
perpendicular to the surface, the diffusion of tin atoms must
have taken place along the interface of copper and tin unless
the atoms could move above or under the oxide layer which
appeared unlikely. Williams et al. [19] electrolytically
deposited 15 micron Sn and, in another experiment, Sn
(1.4-3.7w/oCu) Cu alloy onto a W substrate and found
hillocks on Sn layers and whiskers on Sn-Cu layers. The
purpose was to show that it is not necessary to form
intermetallic compounds with the substrate for whisker
formation. But they did not rule out the possibility of
residual stresses caused by electroplating.
Since the driving force is generally accepted as a stress
gradient around the root of a whisker, we designed a device
in which a known compressive stress can be applied
mechanically by bending a silicon wafer coated with tin. To
show whether the interface could be a path for tin atoms we
deposited only one grain layer of tin to eliminate all the
horizontal grain boundaries.
978-1-4244-2231-9/08/$25.00 ©2008 IEEE 472 2008 Electronic Components and Technology Conference