Rate of consumption of Cu in soldering accompanied by ripening
H. K. Kim and K. N. Tu
Department of Materials Science and Engineering, University of California at Los Angeles,
Los Angeles, California 90095-1595
͑
Received 22 May 1995; accepted for publication 25 July 1995͒
What is the rate of consumption of Cu in soldering reactions has been a critical question in
electronic packaging technology. The Cu films are consumed by Cu–Sn compound formation.
Because the Cu thickness is limited and the rework of a solder joint requires a layer of unreacted Cu,
the loss of Cu in soldering must be under control. At the solder interface, Cu–Sn intermetallic
compounds do not form layered structures. Rather, the Cu Sn phase grows as scalloplike grains into
6
5
the molten solder and ripening occurs between the grains. Therefore, it has been difficult to
determine the compound growth rate, and in turn the Cu consumption rate. Using cross-sectional
and top-polished samples, we have measured the total volume of Cu–Sn intermetallic compounds
formed between eutectic SnPb alloy and Cu substrate as a function of reflow time and temperature.
We have deduced that after 1 min reflow, for example, the thickness of Cu consumed was about
0.36, 0.47, and 0.69 m at 200, 220, and 240 °C, respectively. © 1995 American Institute of
Physics.
9
The advance in very large scale integrated ͑VLSI͒ tech-
loplike grains is accompanied by ripening, so it was difficult
nology has increased the demand on functionality and reli-
ability of input/output ͑I/O͒ connections. The need of a
to measure the Cu consumption rate directly.
In order to determine the rate of Cu consumption, we
assume that the total loss of Cu from the substrate is equal to
the sum of Cu in the Cu–Sn compounds and in the liquid
solder. From the mass balance of Cu, the consumed thickness
of Cu is
1
greater number of I/O connections has caused the redistribu-
tion of wiring from boards, cards, and modules to the Si chip
itself. To accommodate a large number of I/O counts, area
arrays of solder bumps are being used in the controlled col-
lapsed chip connection ͑C4͒ configuration for chip joints and
1
nV
1
–3
ball grid array ͑BGA͒ for lower level joints.
For these
⌬hϭ
LϩfCucVc
,
͑1͒
ͩ
ͪ
CuA 100
applications and, in particular, for a direct chip-to-board
electronic packaging, there is a renewed interest in solder
where A is the total interfacial area between solder and Cu, n
is the wt % of Cu in liquid solder, V is the total volume of
liquid solder, fCu is the weight fraction of Cu in the Cu–Sn
1
technology. The reaction kinetics at the solder interface be-
comes an important issue.
At the liquid solder/Cu substrate interface, the formation
and growth of intermetallic compounds occur by the disso-
lution of Cu into the molten solder. Although a number of
studies have been performed on the dissolution of metals,4
very little is known about the dissolution kinetics of Cu into
a molten solder accompanied by compound formation. In the
case of thin-film metallization for making solder joints on a
Si chip, the typical thickness of Cu is about 1 m. Since the
amount of Cu is limited and some of the Cu must remain
intact through all the reflows and subsequent reworks to
avoid dewetting of the solder, the understanding of the con-
sumption rate of Cu is very important. In this letter, we re-
port the consumption rate of Cu in the soldering reaction
with eutectic SnPb alloy.
compound, V is the total volume of the Cu–Sn compound,
c
and Cu , , and are the density of the Cu, liquid solder,
L
c
and Cu–Sn compound, respectively. After the saturation of
Cu in the molten solder, the consumption rate of Cu depends
only on the change of volume of Cu–Sn compounds, and the
rate equation is given by
–7
dh fCu c dVc
ϭ
.
͑2͒
dt
A Cu dt
Since the solubility of Cu in the liquid SnPb solder is very
low1
0–12
and the amount of the liquid solder used is small,
the entire liquid solder cap will be saturated with Cu very
quickly. Therefore, during the reflow, we can calculate the
consumption rate of Cu by the rate of change of the total
volume of Cu–Sn compounds.
The eutectic SnPb/Cu samples were prepared by reflow-
ing eutectic ͑63Sn37Pb, wt %͒ SnPb alloy on Cu plates in
the mildly activated rosin flux ͑RMA͒ at three different tem-
peratures; 200, 220, and 240 °C. The electropolished-Cu
Two kinds of eutectic SnPb/Cu samples were prepared
for measuring the total volume of Cu–Sn compounds; cross-
sectional samples and top-polished samples. To see why
these samples are needed, a selective Pb etching was per-
formed to reveal the morphology of the compounds, as
shown in Fig. 1. Figures 1͑a͒ and 1͑b͒ show the cross-
sectioned and top-polished views of the compounds, respec-
͑
99.95% purity͒ plate was immersed in the heated flux at one
of the reflow temperatures with Ϯ3 °C control, and a small
ϳ2.0 mg͒ solder ball was placed on the plate. The wetted
͑
ball spread out as a cap and it was solidified by cooling after
a given reflow time. Due to the selective Cu–Sn reaction in
9
tively. In a previous work, we have reported the growth of
8
which Pb does not react with Cu, a very rough scalloplike
the scalloplike Cu Sn compound grains through combined
6
5
compound structure was developed. The growth of the scal-
kinetic processes of ripening and interfacial reaction. Be-
2
002
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
28.189.203.83 On: Fri, 12 Dec 2014 07:42:10
Appl. Phys. Lett. 67 (14), 2 October 1995
0003-6951/95/67(14)/2002/3/$6.00
© 1995 American Institute of Physics
1
Kim, H. K.
