1
Published in J. Electron. Mater. 29 (2000) 1122-1236
Experimental and Thermodynamic Assessment
of Sn-Ag-Cu Solder Alloys
K.-W. Moon, W. J. Boettinger, U. R. Kattner, F. S. Biancaniello, and C. A. Handwerker
Metallurgy Division
Materials Science and Engineering Laboratory
NIST
Gaithersburg, MD 20899
USA
Abstract
Sn-rich alloys in the Sn-Ag-Cu system are being studied for their potential as Pb-free solders. Thus, the location
of the ternary eutectic involving L, (Sn), Ag
3
Sn and Cu
6
Sn
5
phases is of critical interest. Phase diagram data in the
Sn-rich corner of the Sn-Ag-Cu system are measured. The ternary eutectic is confirmed to be at a composition of 3.5
wt % Ag, 0.9 wt % Cu at a temperature of 217.2
±
0.2
°
C (2
σ
). A thermodynamic calculation of the Sn-rich part of
the diagram from the three constituent binary systems and the available ternary data using the CALPHAD method is
conducted. The best fit to the experimental data is 3.66 wt % Ag and 0.91 wt % Cu at a temperature of 216.3
°
C.
Using the thermodynamic description to obtain the enthalpy- temperature relation, the DTA signal is simulated and
used to explain the difficulty of liquidus measurements in these alloys.
Key Words:
Pb-free solder, Sn-Ag-Cu solder, phase diagram, ternary eutectic, and thermal analysis.
Introduction
Ternary alloys based on Sn-rich Sn-Cu and Sn-Ag
binary eutectics have attracted considerable attention
as potential Pb-free solders. The National Center for
Manufacturing Science report
[1]
on Pb-free alloys
showed that these binaries as well as their
combinations have favorable solderability and
wetting properties. It is important to have a rather
precise knowledge of the phase diagram in order to
optimize solder compositions for industrial trials
because the levels of Cu and Ag in these solders are
quite small (typically 3.5 wt % Ag and 1 wt % Cu).
In particular, a Pb-free task group of the National
Electronic Manufacturing Initiative
[2]
has focused on
these alloys for manufacturing and reliability testing.
In 1959, Gebhardt and Petzow
[3]
presented a
liquidus surface for the entire ternary. Based on very
little data, they proposed a transition reaction, L +
Cu
6
Sn
5
→
(Sn)
∗
+ Ag
3
Sn at 225
°
C with a liquid
composition of 4.0 wt % Ag, 0.5 wt % Cu. In 1960,
Fedorov et al.
[4]
presented three isopleths, where a
ternary eutectic reaction at 218
°
C is evident in the
Sn-rich corner. In 1994, Miller et al.
[5]
, using DTA,
found a ternary eutectic at 217
°
C and placed its
composition at 4.7 wt % Ag, 1.7 wt % Cu. A patent
was issued based on this work
[6]
. Most recently,
Loomans and Fine
[7]
place the ternary eutectic
composition at 3.5 wt % Ag and 0.9 wt % Cu using
thermal analysis of the signal from the monovariant
binary eutectics, L
→
(Sn) + Cu
6
Sn
5
and L
→
(Sn) +
Ag
3
Sn.
∗ The symbol, (Sn), will be used to designate the
Sn phase in contrast to the component Sn.
2
Using thermal analysis, the present work has
examined alloys along two isopleths in the vicinity of
the reported ternary eutectic compositions. These
results and other selected data are used to develop a
thermodynamic model for the Sn-rich portion of the
ternary phase diagram. The difficulty of liquidus
measurement for the intermetallics in this system is
discussed using simulated DTA curves. These DTA
curves are based on the calculated enthalpy-
temperature predictions of the thermodynamic model.
Experimental Procedures
Preliminary thermodynamic calculations
performed by one of the authors (URK), and
reported
*
by Miller et al.
[5]
predicted a ternary
eutectic and indicated that the Cu
6
Sn
5
and Ag
3
Sn
liquidus surfaces were quite steep compared to the
(Sn) liquidus. This can also be easily seen from the
two binary diagrams. Thus, it was important to
perform thermal analysis at sufficiently high
temperatures to access the entire melting interval. In
addition, simple lever law calculations indicated that
the amount of primary intermetallic in the
composition range of interest is quite small (
≈
2 wt
%). Thus the liquidus signal during thermal analysis
was likely to be weak; therefore, special attention
was paid to the sensitivity of the thermal analysis
technique.
Small alloy ingots were prepared by melting
99.99% purity metals in sealed and evacuated quartz
ampoules at 1100
°
C followed by agitation and water
quenching. The chosen compositions lie along two
sections, A and B, as shown in Fig.1. Section A was
chosen to study the liquidus surfaces of Cu
6
Sn
5
and
Ag
3
Sn. Section B was chosen to include the ternary
eutectic composition reported by Loomans and Fine
[7]
. Elemental weights are considered accurate to
approximately 0.1 mg producing very small
composition errors.
For thermal analysis, 2 g samples were cut
longitudinally from each ingot to minimize any
macrosegregation effects and were re-melted in a test
*
An error was made in the conversion from atomic to
weight % conversion by Miller et al. The
composition obtained from the initial estimate was
Sn - 3.25 wt % Ag - 0.69 wt % Cu.
tube in air at approx. 250
°
C. A fine 250
µ
m Inconel
∗
sheathed chromel-alumel thermocouple was inserted
in the center of the melt. The interior of the test tube
and the thermocouple were coated with boron nitride.
This coating was found to reduce, but not eliminate,
the tendency for the liquid to supercool with respect
to the (Sn) phase. The thermocouple was held in
place by a glass test tube stopper. The test
tube/thermocouple assembly was then inserted into a
hollow graphite cylinder resting inside of a furnace.
A reference thermocouple was inserted into a vertical
hole in the graphite with its tip at the same height as
the sample thermocouple. Heating and cooling was
performed with the furnace programmed at constant
cooling and heating rates of 0.5 K/min and 5 K/min.
Data were acquired with a commercial thermocouple
logging software.
The thermocouples were calibrated using the
melting of pure Sn. Data are reported in DTA type
format; i.e., the difference between the sample
temperature and the reference temperature is plotted
versus the sample temperature. To obtain a flat base
line, a test was performed at each heating/cooling rate
with an empty sample test tube. The DTA signal
from this dummy test was subtracted from that
obtained with the alloy samples. In standard DTA or
DSC, the sample thermocouple is located just below
the sample container. In the present experiments the
thermocouple probe is extremely thin, is in direct
contact with the alloy and is thus more sensitive than
standard methods. Fig. 2 shows the melting and
freezing signal for pure Sn at two heating/cooling
rates. In contrast to the standard DTA/DSC, the drop
of the DTA signal during melting at 231.8
°
C is
vertical and is not sensitive to heating rate. During
cooling (Sn) nucleation occurs approx. 30
°
C below
the melting point. The recalescence from this
temperature is indicated by the positive slope of the
DTA signal. The maximum temperature reached is
the Sn melting point at 232.0
°
C along the short
vertical segment. Thus the cooling and heating
difference is 0.2
°
C.
Standard metallographic examination was
performed on selected alloys after cooling at the two
rates. Phase identity was confirmed by energy
dispersive x-ray analysis in the SEM using elemental
standards.
∗
Trade names are used in this paper for completeness
only and their use does not constitute an endorsement
by NIST.
3
Results and Discussion
Thermal Analysis
The general solidification behavior of alloys in a
simple ternary eutectic system is well known.
Solidification consists of three stages (primary,
secondary and tertiary) and involves a liquid and
three solid phases. If the three solid phases are
denoted
α, β
and
γ
, the primary stage would be L
→
α
, the secondary would be L
→
α+β,
and the tertiary
would be L
→
α+β+γ.
Depending on the alloy
composition, the identity of the three phases is
permuted. The first two stages occur over a range of
temperatures and the third occurs at a fixed
temperature. We will call the reaction L
→
α+β
a
monovariant binary eutectic reaction. It is
monovariant
because it has one degree of freedom
and
binary
in the sense that only two solid phases
form from the liquid. Assuming no nucleation or
growth difficulties the following should be possible.
During cooling, thermal analysis should be able to
detect three temperatures corresponding to the
beginning of each stage; i.e., at temperatures at which
each new solid phase appears. During heating,
thermal analysis should be able to detect three
temperatures at which each solid phase finally
disappears. In the above example, a signal should be
present when all of the
γ
phase is finally gone, a
temperature where all of the
β
phase is gone, and
finally a temperature where all of the
α
phase is gone.
Temperatures obtained on heating and cooling should
bracket the true thermodynamic temperature.
The preferred method of thermal analysis is
heating. Imperfections in solid structures generally
provide ample nucleation sites for phase changes
(grain and interphase boundaries). On cooling of a
liquid, however, liquid supercooling is often
observed. One of the peculiarities of these alloys is
the difficulty of observing the liquidus temperature
during melting. Thus, data is presented from both
heating and cooling. Signals obtained on cooling
remain significant because they establish a lower
bound on the possible reaction temperature.
Section A – Fig. 3 shows DTA curves for both
heating (lower curve) and cooling (upper curve) for
alloys from a portion of Section A; viz., from 1.5 wt
% Ag-2.7 wt % Cu to 5.0 wt % Ag-1.4 wt % Cu. For
brevity, these alloys will be specified by their Ag
content only. On heating, the onset of melting is
found near 217
°
C (see Table 1 for a summary of all
results for Section A). The invariant melting process
at the ternary eutectic temperature causes a vertical
drop in the DTA curve. For the first two alloys in this
series (1.5 wt % Ag and 2.5 wt % Ag), a second peak
is observed on heating slightly above the ternary
eutectic temperature that corresponds to the cessation
of one of the monovariant binary eutectic reactions.
Upon heating to 300 ºC, it is difficult to distinguish
any other signal above the noise. For the other four
alloys, only the peak for ternary eutectic melting near
217 ºC is detected.
On cooling, a small peak is seen in each scan well
above the ternary eutectic temperature. We will
establish that the onset of this peak corresponds to the
beginning of primary intermetallic solidification (in
this case Cu
6
Sn
5
). The signal is small because of the
very small amount of the phase that was expected in
these alloys. The onset of these small peaks are listed
in Table 1 as T
L
(Cooling). Note in Table 1 that the
onset temperatures of these small peaks at the highest
temperatures in the scans follow a decreasing trend as
the Ag content increases along Section A until 5.0 wt
% Ag. Thereafter, it increases as the Ag content
increases along Section A as would be expected for
the liquidus temperature in a section cutting across
the L
→
Cu
6
Sn
5
+ Ag
3
Sn monovariant binary eutectic
line.
The DTA scans in Fig. 3 show that during further
cooling three of the alloys (4.1, 4.7 and 5.0 wt % Ag)
exhibit a second small peak whose onset temperature
corresponds to the beginning of the secondary stage
of solidification; viz., the monovariant binary
eutectic. We will establish that the eutectic is L
→
Ag
3
Sn + Cu
6
Sn
5
. The size of these peaks are also
small because only a small amount of solid forms by
this eutectic. These onset temperatures are recorded
in Table 1 as T
B
(Cooling); i.e., the monovariant
binary eutectic onset temperature during cooling.
Further cooling leads to supercoolings of the
order of 20
°
C below the ternary eutectic
temperature. This is due the difficulty of (Sn)
nucleation in these alloys. After nucleation of the
(Sn) phase, the maximum temperature reached after
recalescence is noted. For the two alloys (1.5 and 2.5
wt % Ag) that had a second peak, this temperature is
above the ternary eutectic temperature. It is recorded
in Table 1 as T
B
(Cooling). Also seen in these two
cases is a short vertical segment near the ternary
eutectic temperature. For the other alloys (4.1, 4.7,
and 5.0 wt % Ag) the maximum recalescence
temperature is the ternary eutectic temperature. This
is because a monovariant binary eutectic has formed
earlier in these three alloys. One also notes the
absence of T
B
signals for several alloys. Estimated
uncertainty (2
σ
) based on interpretation of the DTA
signals is 0.2 K.
4
Table 1 – Summary of thermal analysis data for section A
Heating Cooling
Sn wt % Cu wt % Ag wt %
T
L
T
B
T
E
T
L
T
B
T
E
91.10 0.00 8.90
295.0 221.0
91.70 0.30 8.00 287* 220.2 217.6 280.1 219.6 217.1
92.20 0.60 7.20 271* 218.9 217.3 258.2 217.7 217.0
92.50 0.80 6.70 267* 217.9 217.3 262.9 217.9 217.1
92.70 0.90 6.40 264* 217.6 217.3 262.9 213.9
+
217.1
92.95 1.05 6.00 n.d. n.d. 217.2 253.9 223.9 216.9
93.40 1.30 5.30 245* n.d. 217.2 243.2 237.7 217.0
93.60 1.40 5.00 245* 240.0 217.2 241.7 238.2 216.9
93.80 1.50 4.70 n.d. 236.2 217.2 251.7 233.9 217.1
94.15 1.75 4.10 271* n.d. 217.2 269.8 224.3 217.1
94.50 2.00 3.50 n.d. 217.9 217.3 278.9 n.d. 217.1
95.20 2.30 2.50 293* 219.8 217.2 290.5 218.9 217.0
95.80 2.70 1.50 309* 222.6 217.3 307.3 221.6 217.0
96.73 3.27 0.00
325.0 227.0
Italics
– not measured, from binaries n.d. – not detected
* – measured with cycling experiments
+
– large supercooling of (Sn)
Table 2 – Summary of thermal measurements for section B
Heating Cooling
Sn wt % Cu wt % Ag wt %
T
L
T
B
T
E
T
L
T
B
T
E
89.51 0.00 10.49
305.0 221.0
91.70 0.30 8.00 287* 220.2 217.6 280.1 219.6 217.1
93.40 0.60 6.00 258* 219.2 217.2 257.0 217.9 217.1
94.28 0.72 5.00 244* 218.5 217.3 240.2 n.d. 217.1
94.98 0.82 4.20 226* 218.2 217.4 224.9 n.d. 217.1
95.51 0.90 3.59 n.d. 217.6 217.3 213.6
+
217.5 217.0
95.85 0.95 3.20 n.d. 217.8 217.3 203.3
+
217.7 217.1
95.96 0.97 3.07 219* 217.8 217.3 220.3 217.9 217.0
96.64 1.07 2.29 232* 221.3 217.4 228.6 218.8 217.0
97.32 1.18 1.50 248* 223.0 217.2 235.3 221.9 216.9
98.64 1.36 0.00
255.0 227.0
Italics
– not measured, from binaries n.d. – not detected
* – measured with cycling experiments
+
– metastable Ag
3
Sn liquidus
5
Cycling Experiments - Also found in Table 1 are
the T
L
values for heating for three of the alloys.
These data were obtained by thermal cycling
experiment following the method described by Wu
and Perepezko
[8]
. In such experiments, a sample was
heated to a temperature slightly above the
T
L
(Cooling) value, held for 3 hours and cooled. If no
signal occurrs on cooling, then it is evident that the
hold temperature is below the true liquidus; i.e., some
solid phase present. This process is repeated at
increasing hold temperatures until a signal does
occur. By this procedure, the true liquidus could be
determined. An example of a cycling experiment is
shown in Fig. 4. For the 7.2 wt % Ag alloy in Section
A, the true liquidus was 13 K higher than the peak
onset observed during cooling. For the other alloys
subjected to cycling experiments, T
L
(Cooling) and
T
L
(Heating) are much closer. Cycling experiments
were not performed on the other alloys.
The various temperatures are plotted in Fig. 5 for
the Section A isopleth. The topology of this diagram
was guided by known topology of isopleths through
ternary eutectic systems. The absence of T
B
signals
on heating between 3.5 wt % Ag and 6.4 wt % Ag is
now evident. These signals involve the melting of
monovariant binary eutectic Cu
6
Sn
5
+ Ag
3
Sn
→
L
whose amount is very small.
In Fig. 5, the corners of the three-phase field of
L+Cu
6
Sn
5
+Ag
3
Sn at 217.2
°
C (6.45 wt % Ag, 0.88
wt % Cu and 3.43 wt % Ag, 1.99 wt % Cu) are very
important. These were located precisely using curve
fitting of the (L+ Ag
3
Sn)/ (L+ (Sn)+ Ag
3
Sn) and (L+
Cu
6
Sn
5
)/ (L+ (Sn)+ Cu
6
Sn
5
) boundary data obtained
on heating. Lines drawn on the ternary composition
plot from each of the two intermetallic compositions
through these points respectively will intersect at the
ternary eutectic composition. Using this construction,
we obtain 3.5 wt % Ag and 0.9 wt % Cu. This
concentration agrees with that of Loomans and Fine
[7]
and lies exactly on Section B.
Section B - Table 2 and Fig. 6 summarize the
results for Section B. The results are clear except for
some uncertainty around 3.5 wt % Ag. Extrapolation
of the liquidus curves for the Ag
3
Sn and Cu
6
Sn
5
phases would intersect below 217 ºC.
∗
This low
intersection temperature suggests that Section B cuts
through a small piece of the (Sn) phase liquidus
∗
Signals for T
L
(Cooling) were obtained below
217 ºC for 3.2 wt % Ag and 3.95 wt % Ag. These
temperatures lie on the extrapolated Ag
3
Sn liquidus
and indicate metastable solidification of Ag
3
Sn in the
absence of (Sn) and Cu
6
Sn
5
surface and that the ternary eutectic composition is
slightly to the Cu- and/or Ag-rich side of the section.
However, because most of these temperatures were
determined on cooling, the two intermetallic liquidus
curves could be higher in temperature thereby
reducing the concentration range of the (Sn) liquidus
in Section B. In fact, the results from Section A
concluded that the ternary eutectic composition lies
on Section B and would imply that no (Sn) liquidus
should appear in Section B. We, therefore, conclude
that the ternary eutectic composition may lie at most
(0.2 wt %) to the Cu- and/or Ag-rich side of 3.5 wt %
Ag, 0.9 wt % Cu.
Metallography
Energy dispersive microprobe x-ray composition
analysis was performed on large intermetallics of the
Cu
6
Sn
5
and Ag
3
Sn and a (Sn) dendrite arm to
determine the composition of the various phases in as
cast samples. The results are shown in Table 3. It can
be noted that the Ag solubility in Cu
6
Sn
5
and the Cu
solubility in Ag
3
Sn are quite small. Likewise for the
Cu and Ag solubility in (Sn). Because solidification
and melting in this system involves phases with
negligible solubility ranges, there is little possibility
for microsegregation within a phase. Thus predictions
of equilibrium (lever) phase diagram calculations are
quite valid during melting and solidification. This
greatly simplifies the interpretation of DTA signals.
Fig. 7 and Fig. 8 show the microstructure of six of
the alloys, labeled A1, A2, A3 in Fig. 5 (Section A)
and B1, B2, and B3 in Fig. 6 (Section B). All contain
widely separated large intermetallic needles. For
alloys A1 and B1, the intermetallic is Ag
3
Sn. In A3
and B3, the large intermetallic is Cu
6
Sn
5
. These
observations are consistent with the liquidus curves
for the intermetallics identified in Figs. 5 and 6. In
alloys A2 and B2, both types of large intermetallics
are present. These alloys lie near the L
→
Cu
6
Sn
5
+
Ag
3
Sn monovariant binary eutectic line and the two
intermetallics are expected. They grow independently
because eutectic reactions between two facetted
intermetallic phases do not exhibit coupled growth;
i.e., an interposed mixture of the two phases.
The microstructure between the needles compromises
approx. 98% of the sample volume and is composed
of a dendritic pattern of the (Sn) phase. The dendritic
pattern occurs because of the supercooling prior to
the formation of (Sn). This supercooling occurs even
though large intermetallic particles are present and
indicates that the intermetallics are ineffective as
heterogeneous nucleation substrates for (Sn). Prior to
(Sn) nucleation, but during intermetallic growth, the
![Figure 13. Calculated isopleth between Sn-Ag eutectic in the binary Sn-Ag system and pure Cu with the data of Chada et al. [17]. This plot shows the predicted maximum solubility of Cu in liquid Sn- 3.5 wt % Ag solder as a function of temperature.](/figures/figure-13-calculated-isopleth-between-sn-ag-eutectic-in-the-17rffby3.webp)
