Acta mater.

49 (2001) 2481–2489
www.elsevier.com/locate/actamat

INTERFACE REACTION BETWEEN COPPER AND MOLTEN

TIN–LEAD SOLDERS
K. H. PRAKASH and T. SRITHARAN†
School of Materials Engineering, Nanyang Technological University, Nanyang Avenue, Singapore 639798

( Received 23 November 2000; received in revised form 26 March 2001; accepted 27 March 2001 )

Abstract—The formation and growth of Cu–Sn intermetallic film at the interface between molten Sn–Pb
solders and Cu were studied at different temperature and exposure times. The η-phase (Cu6Sn5) was observed
to form at all conditions except at the lowest Sn level of 27 wt% and at the two highest temperatures of 290
and 310°C. The ε-phase (Cu3Sn) was then obtained. At high Sn contents and short times, a cellular film with
a rugged interface was obtained which evolved into a compact film with a scalloped interface as the Sn
content decreased and exposure time increased. The ε-phase film always formed with a relatively planar
interface. The intermetallic grains showed preferred crystallographic orientations. Thickness measurements
showed that the net growth rate depends not only on diffusion through the film but also on the film dissolution.
Precipitation of the η-phase whiskers was detected during cooling, particularly when the temperature was
significantly high.  2001 Acta Materialia Inc. Published by Elsevier Science Ltd. All rights reserved.

Keywords: Intermetallic; Interface; X-ray diffraction (XRD); Crystal orientation

1. INTRODUCTION state of the solder. Studies in the molten state are less
common. Even the growth of IMC film in the solid
Solders play an indispensable role in joining materials
state is thought to depend, to some extent, on the
and components in electronic assemblies and inter-
initial film that formed during the reflow process [7].
connections. They provide electrical conduction and
The move towards Pb-free solders is well on its
mechanical stability to the assemblies. Consequently,
way. The deleterious effects of Pb on the environment
the joint quality and reliability are crucial to success-
and human health are well publicised but its techno-
ful function of the assembly. The Sn–Pb alloys, parti-
logical benefits in solders are not. Although the inten-
cularly those near to the eutectic composition, are
tion here is not to counter the arguments for eliminat-
used as solders while the main substrate or leads are
ing Pb, it is worth recapitulating its technological
made of Cu. Sn in the solder readily reacts with Cu
benefits to remind ourselves of the problems that may
to form intermetallic compounds (IMC) as a film at
be encountered when Pb is removed from solder
the interface during the solder reflow process. Forma-
alloys. Some well known beneficial effects of Pb
tion of an IMC film is imperative for good wetting
[8] are:
and bonding, but an excessively thick film is harmful
because of its brittleness which makes it prone to
mechanical failure even under low loads. Further, the 앫 It reduces the surface tension of Sn to facilitate
film could continue to grow during service if the joint wetting.
attains high temperatures due to internal heat gener- 앫 It prevents transformation of white β tin to grey
ated by the chip or due to heat derived from the exter- α tin when cooled to less than 13°C.
nal environment. The interface between the IMC and 앫 It acts as a cheap solvent for Sn.
solder then becomes a site for easy nucleation of
cracks. Such concerns have compelled researchers to The benefit of Pb being a cheap, non-reactive solvent
study the growth of IMC films in solder joints and is particularly relevant to this study since its absence
their effects on joint embrittlement [1–6]. In most may accelerate the interface reaction with the conse-
cases, studies have focussed on growth in the solid quence of deleterious effects on the joint. The most
promising Pb-free compositions appear to be those
near the ternary eutectic Sn–4.3%Ag–1.7%Cu [9, 10]
† To whom all correspondence should be addressed. Fax: which contain a very high level of Sn compared to
+65-790-9081. the Sn–Pb eutectic solder. Furthermore, its liquidus
E-mail address: assritharan@ntu.edu.sg (T. Sritharan) temperature is about 220°C requiring a high reflow

1359-6454/01/$20.00  2001 Acta Materialia Inc. Published by Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 9 - 6 4 5 4 ( 0 1 ) 0 0 1 4 6 - X
2482 PRAKASH and SRITHARAN: COPPER AND MOLTEN TIN–LEAD SOLDERS

temperature. This combination of a high process tem- are used in the analysis. The IMC layer thickness at
perature and high Sn content could be a potential each measurement position was taken as the linear
source of problems for Pb-free solder users. Hence, distance between the Cu substrate and the top of the
the main aim of the present study is to assess the IMC at that locality, regardless of the density of the
effects of Sn content on the characteristics of the IMC film. Pole figure measurements were made on selec-
film formed in reaction with the molten Sn–Pb sold- ted samples by the Schulz reflection method in a
ers. Rigaku X-ray diffractometer fitted with an Ultima+
wide-angle goniometer. The measurement mode was
a coaxial circle scan mode with α (the angle between
2. EXPERIMENTAL PROCEDURE
the diffractometer axis and the normal to the sample
Rectangular specimens of size 20×10×5 mm3 were surface) ranging from 15 to 90° in steps of 5°. Pole
cut from a commercial grade copper bar, mirror-pol- figures were constructed on (101) projection for η-
ished on both faces and etched with 50% nitric acid. phase and (002) for ε-phase.
Then, they were fluxed with commercial rosin mildly
activated (RMA) flux. About 1 kg of solder alloy was
3. RESULTS
melted in a stainless-steel crucible and the copper
specimens were immersed in the molten bath after 3.1. Identification of the IMC
preheating to evaporate the flux. Solder alloys of four
Diffractograms of all samples except those for
compositions, shown in Table 1, were investigated at
27Sn–73Pb solder gave clear peaks corresponding to
different temperatures as listed in the table in order
(101) and (102) planes of η-phase (Cu6Sn5). Even the
to have similar superheats (⌬T) above their liquidus
27Sn–73Pb solder gave these same peaks at 260 and
temperatures. The temperatures were measured using
270°C, but at 290 and 310°C the ε-phase (Cu3Sn) was
thermocouples to the accuracy of ±2°C. The copper
detected. Examples of diffractograms obtained after
plates were removed from the solder bath after reac-
120 and 330 min are shown in Figs 1 and 2 for a
tion durations of 1, 15, 60, 120, 210 and 330 min and
melt superheat (⌬T) of 10°C. Peaks of Cu are also
were immediately quenched in water. In total,
evident but these were probably coming from the sub-
96 samples (4 alloy compositions×4 temperatures×6
reaction durations) were prepared and investigated.
After the reaction, the samples were sectioned
using a diamond cut-off wheel. One piece was pre-
pared in the standard manner for metallographic
characterization of cross sections. The reagent, four
parts nitric acid, four parts glacial acetic acid and 16
parts water, was used for metallographic etching. The
other piece of the sample was immersed in 13%
HNO3 in water to dissolve away the excess solder to
expose the IMC film. The IMC grain morphology was
studied in plan view and then the sample was used
for X-ray diffraction identification.
A JEOL 5410 SEM was used for morphology and
interface characterization while a SHIMADZU 6000
X-ray diffractometer with thin film attachment was
used for identification of the IMC layer. The layer
thickness measurements were made using an image
analyzing software on the digitized SEM micro-
graphs. For each sample measurements were made in
Fig. 1. X-ray diffractogram of the IMC film at ⌬T = 10°C for
two regions that are far apart. In each of these 120 min for all solder alloys. Note that the two η-phase peaks
regions, 30 measurements were made at regular inter- corresponding to the planes (101) and (102) are clear in all
vals of 5 µm. The averages of the 60 measurements samples. Some peaks from the Cu substrate are also evident.

Table 1. Composition of the solder alloys used for this study

Solder alloy composition (wt%) Solidus temperature Liquidus Solder bath temperatures (°C)
(°C) temperature (°C)

⌬T = 10°C 20 40 60

Pure Sn 232 232 242 252 272 292

82Sn–18Pb 183 200 210 220 240 260
63Sn–37Pb 183 183 193 203 223 243
27Sn–73Pb 183 250 260 270 290 310
 PRAKASH and SRITHARAN: COPPER AND MOLTEN TIN–LEAD SOLDERS 2483

Fig. 4. X-ray diffractogram of the IMC film at ⌬T = 60°C for

Fig. 2. X-ray diffractogram of the IMC film at ⌬T = 10°C for 330 min. As in Fig. 3, but more ε-phase peaks have appeared
330 min for all solder alloys. As the film has thickened, four for the 27Sn–73Pb solder.
η-phase peaks are evident in all samples.

is difficult at lower temperatures [12]. Thus, the over-

strate. Figures 3 and 4 show results for ⌬T = 60°C
all kinetics of ε-phase formation may be too slow at
which are similar to those at ⌬T = 10°C, except for
the lower temperatures and then η-phase is favoured
the 27Sn–73Pb solder which shows peaks corre-
even in the low-Sn solder. It has been reported that
sponding to the ε-phase.
η is the first phase to form in high-Sn solders [12,
By virtue of the high Cu:Sn ratio of 3:1 the ε-phase
13], while ε is preferred in low-Sn solders [14, 15].
may be expected when Sn atoms are in short supply.
In our results, ε-phase was obtained at 310°C in the
From thermodynamic considerations, it has been
low-Sn solder 27Sn–73Pb, while the η-phase was
shown [11] that the driving force for η-phase forma-
obtained at 260°C. In a high-Sn solder, 82Sn–18Pb,
tion is higher than that for ε-phase formation. As the
the η-phase was obtained at 260°C. Hence, while
temperature increases, the difference between the two
confirming the general dependence on Sn content as
driving forces decreases, making it thermodyn-
reported in the literature, our results further indicate
amically favourable to form ε-phase. It is also
that temperature also has an influence. The actual
reported that the nucleation of ε-phase at the interface
temperature at which this η to ε transition occurs will
depend on the solder composition. It can be expected
to decrease as the Sn content is lowered. For example,
in pure Sn, the η-phase was obtained at 292°C while
in 27Sn–73Pb solder the ε-phase was obtained at
290°C. At very low Sn levels (⬍27%) one might
reach the situation where a superheat may not be
required to form the ε-phase and then this phase will
be obtained at any reflow temperature.
In solid-state growth of IMC where the tempera-
tures are lower than those investigated here [1–3, 6]
the ε-phase was observed between Cu and η-phase
even in high Sn solders, resulting in two distinct
films. This may be due to the low diffusion kinetics
at these temperatures where the supply of Sn atoms
to the Cu–η interface may be inadequate, compelling
the formation of ε. In the present study, such double
films were not detected up to 330 min of exposure.
Tu and Thompson [12] and Clevenger et al. [13]
detected the high-temperature phases, Cu4Sn and
Cu10Sn3, when bilayer films of Cu–Sn were heated to
a temperature greater than 400°C. The fast cooling
Fig. 3. X-ray diffractogram of the IMC film at ⌬T = 60°C for
120 min for all solder alloys. Four η-phase peaks are evident inherent in their experimental procedure may have
in all except in the 27Sn–73Pb solder. The latter shows evi- stabilized these phases at room temperature. In the
dence of ε-phase only. present study, the lower temperatures involved may
2484 PRAKASH and SRITHARAN: COPPER AND MOLTEN TIN–LEAD SOLDERS

Fig. 5. (101) pole figures of η-phase films with different morphologies: (a) cellular, in pure Sn at 242°C, and
(b) scalloped, in 27Sn–73Pb at 260°C.

not have favoured such stabilization of the high-tem- the growth directions in this film were identified to
perature phases. be <102> and <031>.
3.2. Preferred orientation of IMC grains 3.3. Morphology of the IMC film
The (101) pole figures obtained in the specimens Microstructures of the interface IMC films were
containing the η-phase films showed a strong texture. examined in cross-section and in plan view for all
Typical examples are given in Fig. 5. Using the lattice samples. Micrographs of cross-sections are widely
parameters given in the JCPDS file [16] its corre- published in the literature but their plan views are not
sponding stereographic projection was obtained using commonly reported. It is necessary to observe both
a software developed by Steffen Weber [17]. The to understand the morphology of the IMC grains.
high-density poles were identified as <102> and There are two aspects to the morphology: one is the
<101> directions by superposing the pole figures on compactness of the film and the other is the nature
their stereographic projections as described by Cullity of its interface with the solder. Examples of different
[18]. It must be noted that same poles were obtained morphologies observed are shown in Fig. 7. They
even though the morphology of the η-phase film could be classified into the three categories
changed with growth conditions. described below:
An example of the (002) pole figure obtained in a
specimen containing the ε-phase film is shown in Fig. (i) Cellular film with a rugged interface: this
6. Using the lattice parameters in the JCPDS file [19], appears like columnar grains in plan view but
cross-section shows branches with substantial
intergranular space. Hence, the film is not dense.
Although the microstructure is reminiscent of a
dendritic structure, it probably is not because dif-
ferent branches may not belong to one crystal.
The interface with the solder is rugged.
(ii) Dense film with scalloped interface: this appears
similar to the cellular grains in plan view but the
film is dense beneath the surface. The interface
with the solder appears like scallops.
(iii) Dense film with planar interface: this was seen
only in the case of 27Sn–73Pb solder at the two
highest temperatures, 290 and 310°C, employed
in this study where the ε-phase formed. The inter-
face with the solder is relatively planar.

In all cases, small rounded crystals nucleated on

the Cu substrate initially and grew into the molten
solder normal to the substrate. Thereafter, growth
continued in both longitudinal and transverse direc-
Fig. 6. (002) pole figure of ε-phase film with planar mor- tions. The longitudinal growth increases the apparent
phology in 27Sn–73Pb at 310°C. thickness of the film while the transverse growth
 PRAKASH and SRITHARAN: COPPER AND MOLTEN TIN–LEAD SOLDERS 2485

Fig. 7. Microstructures of different morphologies of IMC film in cross-section and plan views. (a) Cellular
morphology of η-phase in 82Sn–18Pb at 210°C after 120 min; (b) scallop morphology of η-phase in 63Sn–
37Pb at 193°C after 210 min; and (c) planar morphology of ε-phase in 27Sn–73Pb at 310°C after 330 min.

reduces the inter-granular spacing and increases the solid lines indicate the approximate phase boundaries
film density. After 1 min, the crystals were small and and the eutectic tie line.
almost spherical with a scalloped appearance in sec- High Sn levels, low temperatures and short times
tional view. As reaction time increased, different mor- appear to favour the cellular morphology. These are
phologies evolved depending on the growth con- conditions that promote or prolong the initial transient
ditions. The transition between these morphologies in the growth mechanism before a steady condition
was gradual. Figure 8 is a schematic summary of the is established (this will be discussed in Section 4.1).
morphologies observed for the different temperatures Thus, the cellular morphology seems to be a transient
and Sn contents, but it must be noted that the mor- phenomenon. When steady conditions were reached,
phology changes with time. In this figure, the mor- the growth rate decreased to lower levels. Such low
phology obtained at increasing reaction times is indi- growth rates were achieved at conditions of low Sn
cated by symbols along the horizontal axis within a levels, long times and high temperatures. The
rectangular box. Results in each rectangular box resulting film was dense and exhibited a scalloped
belong to one particular solder composition. The box interface. The planar interface appears to form at very
is positioned on the horizontal axis at locations to low growth rates commensurate with the ε-phase
approximately match the solder alloy composition. reaction kinetics.
The three different morphologies of the IMC films
3.4. Thickness of the IMC film
appear to fall in three identifiable regions with respect
to the reaction temperature and solder composition. The average thickness of the IMC film generally
These regions are delineated by dotted lines while the increased with reaction time for all the alloys, but, as
2486 PRAKASH and SRITHARAN: COPPER AND MOLTEN TIN–LEAD SOLDERS

Fig. 10. Thickness vs time plots for IMC film at four tempera-
Fig. 8. Summary of IMC film morphologies obtained in differ- tures (in °C) for 82Sn–18Pb. The error bars show the 95%
ent solders and temperatures. The data shown within a box is confidence interval of each data point.
in the increasing order of time from 60 to 330 min. The data
in one box belongs to one alloy composition.

Fig. 11. Thickness vs time plots for IMC film at four tempera-
Fig. 9. Thickness vs time plots for IMC film at four tempera- tures (in °C) for 63Sn–37Pb. The error bars show the 95%
tures (in °C) for pure Sn. The error bars show the 95% confi- confidence interval of each data point.
dence interval of each data point.

occurring in the IMC grains. Such fluctuation was

evident from the thickness vs time curves in Figs 9– also observed by Hayashi et al. [20], but they did
12, fluctuations were observed, particularly in sub- not provide any explanations. Samples with scalloped
stantially cellular films. This could be attributed to films exhibited the least scatter.
the measurement technique employed here which Figure 12 shows the plot for the 27Sn–73Pb solder.
actually measures the distance to which the IMC has It is evident that the film thickness in these specimens
grown from the substrate even if the film is not dense. is smaller, although the temperatures are high com-
For example, at 272°C in pure Sn, the thickness at pared to the specimens in Fig. 11. This could be attri-
120 min is apparently greater than that at 210 min, buted to the low Sn content in the solder. Further-
but the film was denser at the longer time. Hence, more, thicknesses obtained at the two highest
more interfacial reaction may have occurred at 210 temperatures, 290 and 310°C, are less than those
min but it is not reflected in the thickness measure- obtained at the lower temperatures of 260 and 270°C.
ments. These specimens are still in the transient This could be because of the transition from the η-
growth stage where the transverse growth is still phase to the ε-phase for reaction times of 60 min and
 PRAKASH and SRITHARAN: COPPER AND MOLTEN TIN–LEAD SOLDERS 2487

D = Do exp 冉 冊
⫺Q
RT
(2)

For example, at 242°C, D is 4.2×10⫺11 cm2/s. Using

this in equation (1) gives x = 5.5, 7.3 and 9.1 µm for
t = 120, 210 and 330 min, respectively. The exper-
imentally measured thickness at 242°C in pure Sn at
these reaction times are 12.5, 7.4 and 11.2 µm,
respectively. Thus, the experimental values are con-
siderably higher than those calculated by equation (1).
This difference could be attributed to an apparently
fast diffusion coefficient arising from the inter-granu-
lar space in the film which short-circuit the diffusion
path. Instead of diffusing through the IMC grains, Cu
and Sn could diffuse through the inter-granular liquid
at a much faster rate. Equation (1) may be applicable
to a solid film where solid-state diffusion through the
Fig. 12. Thickness vs time plots for IMC film at four tempera- IMC film is truly rate controlling. This anomaly will
tures (in °C) for 27Sn–73Pb. The error bars show the 95% however reduce for large reaction times, as the film
confidence interval of each data point. becomes denser.
Equation (1) is not thought to be the best analytical
higher. At 1 and 15 min, the η-phase was obtained equation for use in the present study. Tunca and
with a scalloped interface. This transition also appears Smith [23] have used an alternative equation, which
to have given rise to some scatter of the data. considers both the diffusion-controlled growth and
The apparent film thickness for any reaction time dissolution. It is written in the form
increased with increases in Sn content, but the film
may not be dense at high Sn levels. The 27Sn–73Pb dx K
solder gave the smallest film thickness but the film = ⫺n exp⫺mt (3)
dt x
was particularly dense after long reaction times. This
agrees with the observations reported by other work-
ers [20, 21].
where dx/dt is the growth rate, t is time and K, n, and
m are constants. The first component on the right-
4. DISCUSSION hand side is actually the diffusion controlled growth
given by equation (1), while the second is the dissol-
4.1. Comparison with film growth theories
ution kinetics of the intermetallic in the molten solder.
IMC film thickness, x, is usually related to time, t, The actual increase in thickness is the difference
by the classic equation: between the growth rate and the dissolution rate. Inte-
gration of equation (3) results in a curve of x vs t that
x = √Dt (1) exhibits three stages: an initial transient, a steady state
and then the final stage of purely diffusion-controlled
growth when the melt has saturated with the dissolv-
where D is the diffusion coefficient of the appropriate ing species. Tunca and Smith found this equation to
species in the film. This equation assumes that the fit the results obtained in IMC growth at the interface
rate controlling process is diffusion through the solid in Mo and Nb with molten Al. Our results (Figs 9–
film. Considerable literature is available in IMC 12) show similar initial transient and steady state in
growth in the solid state of solder where equation (1) most tests, particularly in the 63Sn–37Pb and the
has been shown to fit experimental results well, i.e., 27Sn–73Pb solders. The third stage does not appear
a time exponent of 0.5 is obtained from regression to have been reached because the large mass of solder
analysis. However, reports of studies in the molten utilized may not have saturated with Cu in 330 min.
solder are not available. Below we make a compari- It is worth noting that Hayashi et al. [20] found pure
son between the experimental results obtained in the diffusion-controlled growth of IMC film when Cu
present work and that calculated using equation (1). was dipped in a molten bath of Sn saturated with Cu.
Inter-diffusion coefficient values for η-phase of Saturation of the solder by dissolving species will
2.7×10⫺12 and 2.2×10⫺11 cm2/s are reported in the eliminate the transient and steady states leaving only
literature [22] at 180 and 210°C, respectively. Using purely diffusion controlled growth.
these data, the D value could be computed for any Dissolution implies that a solute rich boundary
temperature by the Arrhenius relationship for the dif- layer will be established ahead of the IMC film. The
fusion coefficient concentration of Cu will decrease from the equilib-
2488 PRAKASH and SRITHARAN: COPPER AND MOLTEN TIN–LEAD SOLDERS

rium value (maximum solubility limit of Cu in molten

solder Cmax) at the interface to zero at large distances.
Any disturbance of this boundary layer, for example
by convection currents, will affect the growth rate.
This could be an additional reason for scatter in the
experimental results of Figs 9–12.
The 27Sn–73Pb solder has the highest liquidus
temperature and thus high exposure temperatures are
required. The cellular film was not seen in any of the
specimens. The film thickness appears to stabilize at
about 5–6 µm at the lower temperatures when η-
phase is formed. At the two higher temperatures, ε-
phase is formed with a planar interface and again the
film thickness appears to stabilize at a slightly lower
level of about 3–4 µm. Thus at steady state, the
Fig. 13. Scanning electron micrograph showing whiskers of η
growth rate is negligibly small as a result of high dis- (labelled) nucleated from the ε-phase film (labelled) in 27Sn–
solution rate. A substantial increase will not result at 73Pb solder at 290°C after 210 min of reaction.
these conditions until the melt saturates with Cu.
4.2. The effect of temperature on film growth
Figures 9–12 do not exhibit any clear correlation
between the bath temperature and growth rate of IMC phases could be easily distinguished by EDAX analy-
film. As mentioned earlier, the growth of IMC film sis. The whiskers contain approximately 53 at.% Cu
did not reach the third stage where dissolution stops. and 47 at.% Sn, which closely agrees with the for-
Therefore, the variation of the thickness of IMC film mula Cu6Sn5, while X-ray diffraction detects only the
with temperature depends on the effects of tempera- ε-phase in the film. This implies the η-phase is too
ture on both dissolution and growth rates. These small in mass fraction compared to the ε-phase to be
effects are not well characterized and, hence, it is very detected by diffraction.
difficult to predict the IMC film growth behaviour. Figure 13 seems to indicate that the η-phase whis-
Consequently, it may not always be true that increas- kers nucleated on the existing ε grains of the IMC
ing the temperature will increase the IMC film thick- film. The solute distribution in the boundary layer,
ness in the liquid state, as it is reported to happen in where the concentration of Cu is highest near the ε
the solid state [12] and in the molten Cu saturated film, would have favoured nucleation on the ε grains.
Sn [20]. Ghosh, in a recent study [25] on Ni/Pd metallized Cu
substrate, has detected rectangular precipitates of
4.3. Precipitation during cooling
PdSn4 forming on Ni3Sn4 IMC film. Although his
As evident from the Cu–Sn binary alloy phase dia- microstructures look similar to ours, his PdSn4 for-
gram [24], the solid solubility of Cu in Sn is negligi- med during isothermal reaction. Yet, his study is a
bly small, while in the liquid state it is significant. good example of one phase nucleating on another
Hence, the Cu dissolved in the molten solder, parti- with very different chemistry and crystallography.
cularly in the solute boundary layer, must precipitate Thus, our proposal of η-phase whiskers nucleating on
during cooling. This precipitation usually takes place the ε grains is an acceptable mechanism even though
on the existing grains in the IMC film. The preferred the crystal structures and chemical compositions of
precipitate is the η-phase in accordance with the equi- the two phases are different.
librium diagram. If the IMC film itself is η, then it The faceted nature of the whiskers indicates that
may not be easy to distinguish between the crystals they grew rapidly during cooling. In other solder
that form during the isothermal reaction and those alloys where the film itself constituted the η-phase,
that precipitate on cooling. Furthermore, at the lower such faceting was not observed. Even in the cases
temperatures used for the eutectic 63Sn–37Pb solder, where the IMC film was cellular with a rugged inter-
the solute boundary layer would be small and then face, the plan view clearly shows that the grains were
precipitation may be insignificant. The experiments fairly rounded even at low reaction times. Thus, when
on the 27Sn–73Pb solder where high temperatures are grown isothermally, the sharp edges inherent in a
employed would be the most susceptible to precipi- faceted crystal are avoided to minimise the curvature
tation of the η-phase during cooling. At these con- effect. Hence, we can distinguish the isothermally
ditions, the IMC film consists of ε-phase; it is dense grown grains from those that precipitated during coo-
and planar. Hence, any precipitation of η during coo- ling by the faceting. The precipitated grains have a
ling could be easily detected and distinguished from characteristic whisker shape with a hexagonal cross
the IMC film that grew isothermally. Figure 13 shows section. Furthermore, the precipitated phase is always
an example where the η-phase in the form of whis- the η-phase even if ε-phase is promoted under iso-
kers has precipitated on the ε-phase film. The two thermal growth conditions.
 PRAKASH and SRITHARAN: COPPER AND MOLTEN TIN–LEAD SOLDERS 2489

4.4. Implications for Pb-free solders 5. Generally, the IMC film growth rate exhibited an
Many Pb-free solder compositions have a high con- initial transient followed by a steady state of
tent of Sn [9, 10], thus they are likely to respond to growth. The purely diffusion controlled growth
reflow soldering in a manner similar to the pure Sn rate was never reached in our studies.
in this study. Hence, IMC films containing the η- 6. On cooling after exposure, the η-phase precipitates
phase could be the most common phenomenon when as whiskers in the solute boundary layer. They
Pb-free soldering becomes commercial. The films are nucleate on the existing grains in the IMC film.
likely to be cellular with a rugged interface with the
solder under short exposures typically used in reflow
REFERENCES
programs. The morphology of the IMC film has a pro-
found effect on the mechanical properties of the 1. Morris, J. W. Jr, Freer Goldstein, J. L. and Mei, Z., in
solder joint. A study by Yao and Shang [26] on Mechanics of Solder Alloy Interconnects, ed. D. R. Frear,
fatigue testing of the eutectic Sn–Pb/Cu joints shows S. N. Burchett, H. S. Morgan and J. H. Lau. Van Nostrand
Rheinhold, New York, 1994, p. 7.
that the crack growth rate increases with increases in 2. Marshall, J. L., Foster, A. L. and Sees, J. A., in Mechanics
the roughness of the interface, which they defined as of Solder Alloy Interconnects, ed. D. R. Frear, S. N. Burch-
2h/s, where h is the height and s is the width of the ett, H. S. Morgan and J. H. Lau. Van Nostrand Rheinhold,
IMC crystallites. Therefore, it is likely that the joint New York, 1994, p. 42.
fails easily if high Sn solders are used, which may 3. Unsworth, D. and McKay, C. A., Trans. Inst. Metal Fin-
ishing, 1965, 51, 85.
result in rugged interfaces. To form a smoother inter- 4. Beiling, R. E. and Weaver, H., Z. Metallkd., 1968, 59, 222.
face with solders may require unnecessarily high 5. Kay, P. J. and McKay, C. J., Trans. Inst. Metal Finishing,
exposure times and temperatures. If the solder is satu- 1976, 54, 68.
rated with Cu, then the IMC film morphology could 6. Dirnfeld, S. F. and Ramon, J. J., Welding Res. Suppl.,
1990, 69, 373s.
be different because dissolution is suppressed and a 7. Mei, Z., Sunwoo, A. J. and Morris, J. W. Jr., Metall. Trans.
purely diffusion-controlled growth becomes the oper- A, 1992, 23A, 857.
ative mechanism. Thus, having approximately 1.7% 8. Abtew, M. and Selvaduray, G., Mater. Sci. Engng R, 2000,
Cu in the Pb-free solders may be beneficial to produce 27, 95.
a smoother interface between the solder and the Cu 9. Miller, C. M., Anderson, I. E. and Smith, J. F., J. Electron.
Mater., 1994, 23, 595.
conductor. 10. US Patent No. 5527628, June 18, 1996.
11. Lee, B. J., Hwang, N. M. and Lee, H. M., Acta mater.,
5. CONCLUSIONS 1997, 45, 1867.
12. Tu, K. N. and Thompson, R. D., Acta metall., 1982, 30,
The following conclusions could be drawn from 947.
this study: 13. Clevenger, L. A., Arcot, B., Ziegler, W., Colgan, E. G.,
Hong, Q. Z., d’Heurle, F. M., Cabral, C. Jr., Gallo, T. A.
and Harper, J. M. E., J. Appl. Phys., 1998, 83, 90.
1. The η-phase (Cu6Sn5) is the most common com- 14. Marcotte, V. C. and Schroder, K., Mater. Res. Symp. Proc.,
pound to form at the interface between Cu and the 1983, 19, 403.
solder. For the solder with the lowest Sn content, 15. Grivas, D., Frear, D., Quan, L. and Morris, J. W. Jr., J.
27Sn–73Pb, the ε-phase (Cu3Sn) was observed Electron. Mater., 1986, 15, 355.
instead of η at 290 and 310°C. 16. Joint Committee on Powder Diffraction Standards, 02-
0713, 1997.
2. The η films showed a strong grain orientation in 17. Weber, S., Crystallographic Java Software—Stereographic
the directions <102> and <101>. The ε film Projections, http://members.nbci.com/ xoom/steffenweber
showed the orientations <102> and <031>. /index.html, 1998.
3. The morphology of the IMC film varied depending 18. Cullity, B. D., Elements of X-ray Diffraction, 2nd edn.
Addison-Wesley Pub. Co., 1978.
on the alloy composition, temperature and time of 19. Joint Committee on Powder Diffraction Standards, 01-
reaction. They could be classified into three types 1240, 1997.
as (i) cellular film with rugged interface, (ii) dense 20. Hayashi, A., Kao, C. R. and Chang, Y. A., Scripta mater.,
film with scalloped interface, and (iii) dense film 1997, 37, 393.
with planar interface. The ε-phase formed in type 21. Kim, H. K., Liou, H. K. and Tu, K. N., Appl. Phys. Lett.,
1995, 66, 2337.
(iii) only. The η-phase formed in type (i) at short 22. Starke, E. and Wever, H., Z. Metallkd., 1964, 55, 108.
reaction times and then gradually evolved into 23. Tunca, N. and Smith, R. W., Metall. Trans. A, 1989,
type (ii) as exposure time increased. 20A, 825.
4. The thickness increased with exposure time and 24. Saunders, N. and Miodownik, A. P., Bull. Alloy Phase
Deagrams, 1990, 11, 278.
Sn content of the solder alloy. Scatter in thickness 25. Ghosh, G., Acta mater., 2000, 48, 3719.
data was higher when the type (i) morphology 26. Yao, D. and Shang, J. K., IEEE Trans. CPMT-Part B,
existed compared to the other cases. 1996, 19, 154.