Solid state interfacial reactions in electrodeposited Cu/Sn couples

TANG Wen-ming(汤文明)¹, HE An-qiang², LIU Qi², D. G. IVEY²

1. School of Materials Science and Engineering, Hefei University of Technology, Hefei 230009, China;

2. Department of Chemical and Materials Engineering, University of Alberta, Edmonton, Alberta, T6G 2G6, Canada

Received 24 October 2008; accepted 8 August 2009

Abstract: Cu/Sn couples, prepared by sequentially electroplating Cu and Sn layers on metallized Si wafers, were employed to study the microstructures, phases and the growth kinetics of Cu-Sn intermediate phases, when electroplated Cu/Sn couples were aged at room temperature or annealed at temperatures from 373 K to 498 K for various time. Only Cu₆Sn₅ formed in aged couples or couples annealed at temperature below 398 K. The Cu₆Sn₅ layer was continuous, but not uniform, with protrusions extending into the Sn matrix. When Cu/Sn couples were annealed at temperatures from 423 K to 498 K, two continuous and uniform Cn₆Sn₅/Cu₃Sn layers formed within the reaction region between Sn and Cu. There were many voids near the Cu₃Sn/Cu interface and within the Cu₃Sn layer. Cu₆Sn₅ and Cu₃Sn formations both follow parabolic growth kinetics with activation energies of 41.4 kJ/mol for Cu₆Sn₅ and 90.4 kJ/mol for Cu₃Sn, respectively.

Key words: Cu/Sn couple; electrodeposition; solid state reaction; microstructure; growth kinetics

1 Introduction

In electronic devices, common conducting metals, such as Cu, Ni, Ag, Au and their alloys, form intermetallic compounds (IMCs) with Sn, the major component in Pb-free solders used for packaging applications[1-6]. It was concluded that extensive formation of Cu-Sn IMCs not only degrades mechanical strength of solder joints, but also changes the electrical characteristics of the interconnects because these compounds, in general, are brittle and have high electrical resistivities[7-8]. The microstructure and phase composition of the reaction region and the reaction kinetics for the Cu-Sn binary system are essential for predicting the lifetime and the reliability of chip-to-substrate connections.

Interfacial reaction between Cu and Sn generally results in the formation of Cu₆Sn₅ at room temperature, and Cu₆Sn₅ and Cu₃Sn at higher temperatures. The onset temperature for Cu₃Sn has been reported to range from 333-363 K[9-10]. For the growth kinetics of these intermediate phases, it is widely recognized that the growth of the Cu₆Sn₅ phase may change from a reaction-controlled process to a diffusion-controlled process if the Cu₆Sn₅ layer is thick enough so that the atom flux across the layer is reduced and diffusion becomes the rate-limiting process. For example, the submicron-thick Cu₆Sn₅ layer in Cu/Sn thin film couples exhibits linear growth, while the thicker Cu₆Sn₅ layer in bulk couples usually obeys a parabolic growth law. The formation of Cu₃Sn at elevated temperatures by consumption of Cu₆Sn₅ also exhibits parabolic kinetics[11-12]. Activation energy data[11-12], for the growth of both Cu₆Sn₅ and Cu₃Sn, are quite variable, possibly due to the different Cu/Sn couple fabricating techniques and different annealing methods.

Electrodeposition is frequently used to prepare solders and in some cases metallization because it offers advantages such as simplicity, cost-effectiveness and controllable patterning and it can be applied to fabricating films with thicknesses from ＜＜ 1 ?m to   ＞100 ?m[13-15]. Electrodeposition can also be employed conveniently to prepare two-layer or multi-layer films by sequential plating. For example, TANG et al[16] have successfully prepared Au-rich (Sn-rich) Au-Sn eutectic alloy solders, and Au/Sn (Sn/Au) couples for packaging applications.

In this work, the Cu/Sn couples were prepared by electroplating a Cu layer onto a metallized Si wafer, and followed by plating of a Sn layer. Cu and Sn layers are thick enough to provide a sufficient supply of Cu and Sn for reaction during annealing. The phases and microstructures in the reaction regions of the Cu/Sn couples were investigated after aging at room temperature and annealing at higher temperatures (up to 598 K) for various time. The growth kinetics and phase formation mechanisms for the Cu₆Sn₅ and Cu₃Sn layers were also analyzed.

2 Experimental

The Sn plating aqueous solution contained tribasic ammonium citrate and SnCl₂·2H₂O with β-naphthol and gelatin as levelers, and the Cu plating solution contained tribasic ammonium citrate and CuCl₂·2H₂O. These solutions were used to sequentially plate Sn and Cu layers on Si pieces sectioned from wafer substrates with 100 mm in diameter. The Si wafers were metallized with 25 nm of TiW (90% W and 10% Ti, mass fraction) and 200 nm of Pt. TiW acted as an adhesion layer and a diffusion barrier layer, while Pt served as a seed layer for plating. The Si substrates were sectioned into about 1 cm² pieces.

A Dynatronix DuPR 10-0.1-0.3 pulse plating power supply was used for electrodeposition. The forward direction was set at 10 ms, with a forward on-time of 2 ms and a forward off-time of 8 ms. A current density of 25 mA/cm² was selected to plate the Sn at a rate of about 0.5 μm/min. A current density of 10 mA/cm² was chosen to deposit the Cu at a rate of about 0.3 μm/min. An RC2400 Bio-Rad alloying furnace was used to anneal the deposited Cu/Sn couples. Plated samples were annealed at temperatures from 373 K to 498 K for 0.5-5.0 h. The annealing process was done in a forming gas (95% N₂+5% H₂, volume fraction) atmosphere at a heating rate of 100 K/min and a cooling rate of 40 K/min.

Phases in the aged and annealed Cu/Sn couples were identified using a Rigaku Rotaflex rotating anode X-ray diffraction (XRD) system. Specific Cu/Sn couples for XRD were prepared by plating a 1 μm-thick Sn layer on the Cu layer, in order to allow the X-ray to penetrate the Sn layer and reach the reaction region at the Sn/Cu interface. The aged and annealed samples were also examined, in cross-section orientations, using a Hitachi H2700 scanning electron microscope (SEM) equipped with a PGT IMIX energy dispersive X-ray (EDX) detector. For cross-section observations, the samples were cleaved, mounted with epoxy resin, ground with SiC papers in water and then polished with a 0.05 μm Al₂O₃ particle-water suspension.

3 Results and discussion

3.1 Interfacial reaction at room temperature

Fig.1 shows XRD patterns for the same Cu/Sn couple, which consists of a thin Sn layer (1 mm) on a plated Cu layer (9 mm), aged at room temperature for various time. For comparison, XRD measurement was also conducted on a fresh “as-deposited” sample immediately after the sequential electrodeposition without aging, and the XRD pattern is shown as Fig.1(a). This pattern shows peaks corresponding to Cu and Sn, as well as the Pt seed layer, as expected. In addition, there is a weak peak at about 35? (2q), which corresponds to Cu₆Sn₅. Evidently, some reaction occurs between Cu and Sn at room temperature, either during electrodeposition or shortly afterwards. With increasing aging time (Figs.1(b)-(f)), the peaks at 35? increase in intensity, indicating an additional reaction between Cu and Sn, and the formation of more Cu₆Sn₅. Other Cu₆Sn₅ peaks show up and increase in intensity with aging (e.g., at 2q of about 63? and 85?). No Cu₃Sn is detected, even after aging for 76 d.

Fig.1 XRD patterns for Cu/Sn couple with 1 mm-thick Sn layer aged for various time: (a) 0 d; (b) 5 d; (c) 10 d; (d) 20 d; (e) 40 d; (f) 76 d

It is well established that group ⅠB noble metal atoms (e.g., Cu, Ag and Au) can diffuse rapidly into group ⅢA and group ⅣA host metals (e.g., In, Sn and Pb) via an interstitial diffusion process, even at room temperature[17]. The interstitial diffusivity of Cu in Sn at room temperature can be as high as 1.23×10^-7 m²/s with an apparent activation energy for diffusion of 0.34 eV (32.8 kJ/mol), and the activation energy for grain boundary diffusion of Sn in Cu has been measured as 0.78 eV (75.3 kJ/mol) by BANDYOPADHYAY et al[10]. As such, the transport of Cu atoms from the Cu layer into the Sn layer should predominate and control the formation of Cu₆Sn₅ at room temperature. The presence of Cu₆Sn₅ as the only reaction product in as-deposited and aged Cu/Sn couples can be rationalized in terms of the structure of Cu₆Sn₅. Cu₆Sn₅ has a pseudohexagonal structure, which can form without markedly changing the tetragonal structure of Sn. In addition, Cu₆Sn₅ formation follows the empirical rule where the first phase to form should have the composition nearest the lowest temperature eutectic[18-19].  The lowest temperature eutectic occurs at a composition of 99.3% Sn and a temperature of 405 K (Fig.2)[20]. Cu₆Sn₅ is the closest intermetallic phase to the eutectic, compared with other Cu-Sn intermetallic compounds, such as Cu₃Sn, Cu₁₀Sn₃ and Cu₄₁Sn₁₁.

Fig.2 Cu-Sn phase diagram [20]

Fig.3(a) shows a cross section backscattered electron (BSE) image of the Cu/Sn couple after aging for 76 d. Sn appears bright, while Cu is dark and the interfacial region has intermediate contrast. The reaction region is determined to be Cu₆Sn₅ by EDX analysis (50.2%Sn and 49.6%Cu in molar fraction, Fig.3(b)), which is consistent with the XRD result. The Cu₆Sn₅ layer is continuous, but not uniform. At some locations, it is very thin, while at other locations there are some relatively thick protrusions towards the Sn layer. The formation of Cu₆Sn₅ protrusions through reaction between molten Sn-based solders and Cu substrates has been reported previously by ZENG et al[21], LIU et al[22] and BADER et al[23], and is attributed to the anisotropic nucleation and growth behavior of Cu₆Sn₅. The Cu₆Sn₅ protrusions in the current study are formed through solid state reaction between Cu and Sn, which is different from the reaction conditions for the cited literature. Therefore, the formation mechanism for Cu₆Sn₅ in the current study may be different from those previously reported[21-23].

Fig.3 BSE image (a) and EDX spectrum taken from point A (b) of Cu/Sn couple aged at room temperature for 76 d

3.2 Interfacial reaction at elevated temperatures

Fig.4 shows XRD patterns for Cu/Sn couples annealed at different temperatures from 373 K to 473 K. For couples annealed at 373 K and 398 K, respectively, for 1 h, only Cu₆Sn₅ is detected as the reaction product (Fig.4(a) and 4(b)), which is the same as the results for the couples aged at room temperature. BSE image for a couple annealed at 373 K is shown in Fig.5, confirming that only Cu₆Sn₅ forms. The Cu₆Sn₅ layer is still non-uniform with protrusions at some locations. A particularly prominent protrusion extending into the Sn matrix, perhaps along Sn grain boundary, is shown by the arrow in Fig.5. The formation of Cu₆Sn₅ protrusions is likely due to a combination of grain boundary diffusion and interstitial diffusion of Cu atoms into the Sn. Similar behavior has also been reported in Ref.[15] for Au/Sn diffusion couples aged at room temperature, where AuSn₄ protrusions formed as a result of Au diffusion (both grain boundary and interstitial) into Sn.

Fig.4 XRD patterns of annealed Cu/Sn couples with 1 mm-thick Sn layer: (a) 373 K, 1 h; (b) 398 K, 1 h; (c) 423 K, 0.5 h; (d) 473 K, 0.5 h

Fig.5 BSE image of Cu/Sn couple annealed at 373 K for 4 h

When the annealing temperature was increased to 423 K and the Cu/Sn couple was annealed for a short time (0.5 h, Fig.4(c)), weak diffraction peaks for Cu₃Sn are detected. The onset temperature for Cu₃Sn formation appears to be between 398 K and 423 K for the electroplated Cu/Sn couples studied here, which is higher than the onset temperature for Cu₃Sn formation in evaporated Cu/Sn couples (333-363 K) previously reported by TU and THOMPSON[9] and BANDYOPADHYAY and SEN[10], but lower than that for evaporated Cu/Sn thin film couples (457 K) reported by LIAO and WEI[11]. It is clear from the Cu-Sn phase diagram (Fig.2) that solid Sn is in equilibrium with Cu₆Sn₅, but solid Cu is not in equilibrium with Cu₆Sn₅. Therefore, the Cu₆Sn₅/Cu interface is not thermodynamically stable so that a Cu₃Sn layer forms between the Cu₆Sn₅ layer and the Cu layer by the following reaction:

Cu₆Sn₅+9Cu→5Cu₃Sn                         (1)

Two intermetallic layers are clearly visible for the couple annealed at 423 K for 5 h (Fig.6). The Cu₃Sn layer shows slightly darker contrast relative to the Cu₆Sn₅ layer due to its lower Sn content. The morphology of the reaction region is very different from that for the Cu/Sn couples annealed at room temperature and 373 K. The main difference is the lack of Cu₆Sn₅ protrusions extending into Sn, i.e., both the Cu₃Sn and Cu₆Sn₅ layers are quite uniform. For the Cu/Sn couples annealed at 423 K, Cu₆Sn₅ forms initially (as room temperature annealed couples), forming a non-uniform layer. Subsequently, Cu atoms diffuse to the Cu₆Sn₅/Sn interface through the Cu₆Sn₅ layer to react with Sn, so Cu₆Sn₅ continues to grow towards the Sn layer. Since it takes much longer for Cu atoms to diffuse through the thicker protrusions than through the thin regions between neighbouring protrusions, Cu₆Sn₅ grows faster in the thin regions than at the protruding locations. As a result, the morphology of Cu₆Sn₅ layer evens out over time, producing a uniform Cu₆Sn₅/Sn interface. This flattening process is also driven by thermodynamics because of the higher interfacial energy between Sn and Cu₆Sn₅ for the uneven Cu₆Sn₅/Sn interface.

Fig.6 BSE image of Cu/Sn couple annealed at 423 K for 5 h

As shown in Fig.4(d), strong diffraction peaks for Cu₃Sn are detected for the Cu/Sn couple annealed at 473 K for 0.5 h. Therefore, more rapid growth of the Cu₃Sn layer is achieved by increasing the annealing temperature form 423 K to 473 K. The reaction regions are shown in Fig.7 for Cu/Sn couples annealed at 473 K for various time from 0.5 to 3 h. There are two reacted layers, i.e., Cu₆Sn₅ and Cu₃Sn, as for 423 K. Voids begin to appear for the shorter annealing time (0.5 h, Fig.7(a)), primarily at the Cu₃Sn/Cu interface. The number and size of voids increase with increasing annealing time (1 h, Fig.7(b)). As the annealing time is further increased (2 h), the density of voids increases, although the size remains approximately constant, i.e., no coalescence occurs (Fig.7(c)). In addition, as the annealing time is increased, the voids are present within the Cu₃Sn layer as well as near the Cu₃Sn/Cu interface. Moreover, the intermetallic thickness ratio (Cu₃Sn to Cu₆Sn₅) increases with annealing time, indicating faster growth of the Cu₃Sn layer relative to the Cu₆Sn₅ layer.

Fig.7 BSE images of Cu/Sn couples annealed at 473 K for various time: (a) 0.5 h; (b) 1 h; (c) 3 h

The voids in the Cu₃Sn layer and near the Cu/Cu₃Sn interface are important because void formation can result in weakening of the joint interface, possibly leading to brittle fracture, between the Sn alloy solder and the Cu substrate and loss of much of the electrical connectivity[24]. Some researchers have attributed the voids to Kirkendall voids, while others attributed void formation to impurities in the Cu layer of the Cu/Sn couple[8]. Recently, PAUL et al[25] have found voids in annealed Cu/Sn bulk couples with high-purity Cu and Sn plates. Their result implies that void formation is primarily due to the Kirkendall effect, as a result of the imbalance in migration of the diffusing species, i.e.,   Cu diffuses significantly faster than Sn. OH[26] has determined that the diffusivity of Cu in Cu₃Sn is approximately 3 times larger than that of Sn in Cu₃Sn, leading to void formation at the Cu₃Sn/Cu interface. The voids at the Cu₃Sn/Cu interface will remain as long as Cu is the only diffusing species. The voids will be displaced into the Cu₃Sn layer, if some of the Sn atoms diffuse to the Cu₃Sn/Cu interface to react with the Cu behind the voids. Based on the fact that most of the voids are near the Cu₃Sn/Cu interface and within the Cu₃Sn layer in this work, it can be concluded that diffusion of Sn through the Cu₃Sn layer has occurred, with Sn reacting with Cu at the Cu₃Sn/Cu interface to form Cu₃Sn:

3Cu+[Sn]→Cu₃Sn                            (2)

where [Sn] refers to the dissolved Sn.

Cu remains the dominant diffusing species in the Cu₃Sn phase. Of course, Cu₃Sn is also formed at the Cu₃Sn side of the Cu₆Sn₅/Cu₃Sn interface through reaction between the Cu₆Sn₅ layer and diffusing Cu atoms from the Cu layer:

Cu₆Sn₅+9[Cu]→5Cu₃Sn                        (3)

where [Cu] refers to the dissolved Cu.

Sn is the main diffusing species in the Cu₆Sn₅ phase[8] and the growth of Cu₆Sn₅ is based on the following reactions. At the Cu₆Sn₅ side of the Cu₆Sn₅/Cu₃Sn interface, reaction occurs between the Cu₃Sn layer and the Sn atoms from the Sn layer:

2Cu₃Sn+3[Sn]→Cu₆Sn₅                        (4)

At the Cu₆Sn₅/Sn interface, reaction occurs between the Sn layer and Cu atoms diffusing from the Cu layer:

5Sn+6[Cu]→Cu₆Sn₅                           (5)

In summary, interdiffusion of Sn and Cu atoms during the growth of Cu₃Sn and Cu₆Sn₅ intermetallic layers can be schematically illustrated in Fig.8. The larger arrow in this illustration denotes the faster diffusion species in the given phases, i.e., Cu in Cu₃Sn and Sn in Cu₆Sn₅.

Fig.8 Schematic illustration of interdiffusion of Cu and Sn in reaction region of electroplated Cu/Sn couple annealed at temperature above 423 K

3.3 Reaction kinetics

As mentioned above, the growth of the Cu₆Sn₅ layer dominates at lower temperature; however, with increasing temperature, the Cu₃Sn layer grows rapidly at the expense of Cu₆Sn₅ and the remaining Cu layer. In general, the growth kinetics of Cu₆Sn₅ and Cu₃Sn in the Cu-Sn binary system depends on two types of processes: 1) mass transport through the resultant layers, i.e., the Cu₆Sn₅ layer and the Cu₃Sn layer and 2) the reaction at the interfaces, i.e., the Sn/Cu₆Sn₅, Cu₆Sn₅/Cu₃Sn and Cu₃Sn/Cu interfaces.

The reaction rate constant (k) is a function of temperature (T) as described by an Arrhenius-type equation:

k=k₀ exp[-Q/(RT)]                            (6)

where k₀ is the frequency factor; Q is the apparent activation energy; and R is gas constant (8.31 J/(K·mol)). Kinetic plots for the Cu₆Sn₅ and Cu₃Sn layers are shown in Figs.9 and 10, respectively. Because the Cu₆Sn₅ layer is not uniform when the Cu/Sn couples are annealed at 398 K and below, the layer thicknesses cannot be measured accurately.

Fig.9 Average thickness of Cu₆Sn₅ layer vs square root of time for Cu/Sn couples annealed at different temperatures

Therefore, only the Cu₆Sn₅ layer thicknesses for couples annealed at temperatures from 423 K to 498 K for various time are plotted in Fig.9. The approximate linear relationships, for the Cu₆Sn₅ layer and Cu₃Sn layer thicknesses, with the square root of the annealing time indicate diffusion controlled kinetics for the two compounds. The reaction rate constants, k, for each temperature for the Cu₆Sn₅ and Cu₃Sn layers are marked in Fig.9 and Fig.10, respectively. For the temperatures studied (from 423 K to 498 K), Cu₆Sn₅ layer has a higher k value than Cu₃Sn at 423 K, and lower k values than Cu₃Sn at 473 K and 498 K. The kinetic data obtained in this work can be compared with the interdiffusion data for Cu₆Sn₅ and Cu₃Sn in Ref.[27]. As shown in Table 1, k values for Cu₆Sn₅ at temperatures from 423 K to 498 K are about one order of magnitude less than the interdiffusion coefficients for Cu₆Sn₅ in Ref.[27]. This is, at least partially, because some of the Cu₆Sn₅ is consumed through reaction (3) to form Cu₃Sn. k values for Cu₃Sn at the same temperature are approximately equal to the interdiffusion coefficients for Cu₃Sn in Ref.[24]. Reaction (4) at the Cu₆Sn₅ side of the Cu₆Sn₅/Cu₃Sn interface is not significant, perhaps even negligible. In other words, reaction (3) dominates at the Cu₆Sn₅/Cu₃Sn interface during annealing of Cu/Sn couples at 423 K and above, resulting in faster growth for the Cu₃Sn layer compared with the Cu₆Sn₅ layer.

Fig.10 Average thickness of Cu₃Sn layer vs square root of time for Cu/Sn couples annealed at different temperatures

Table 1 Growth rate constants (k) and interdiffusion coefficients (D_inter) for Cu₃Sn and Cu₆Sn₅

The logarithms of the rate constants are plotted vs. the reciprocal of the absolute temperature, for Cu₆Sn₅ and Cu₃Sn, in Fig.11 and Fig.12. From the linear relationships, the frequency factors and activation energies were determined to be 4.3×10^-12 m²/s and 41.4 kJ/mol for Cu₆Sn₅ and 3.3×10^-3 m²/s and 90.4 kJ/mol for Cu₃Sn, respectively. Cu₆Sn₅ growth has a lower activation energy than Cu₃Sn growth, which agrees with most results previously reported[11, 27]. Although kinetic data have been reported previously, the data are quite variable, e.g., 0.67-0.87 eV (64.6-83.9 kJ/mol) for Cu₆Sn₅ and 0.77- 1.07 eV (74.3-103.3 kJ/mol) for Cu₃Sn. The activation energy obtained here for the growth of Cu₆Sn₅, 41.4 kJ/mol, is lower than the reported values of 0.67-0.87 eV (64.6-83.9 kJ/mol). The reasons are not clear yet, but may be related to the initial abnormally rapid growth of Cu₆Sn₅ in the annealed Cu/Sn couples. The activation energy obtained here for Cu₃Sn, 90.4 kJ/mol, is in reasonable agreement with other data of 0.77-1.07 eV (74.3-103.3 kJ/mol).

Fig.11 Plot of k vs annealing temperature for Cu₆Sn₅ layer in annealed Cu/Sn couples

Fig.12 Plot of k vs annealing temperature for Cu₃Sn layer in annealed Cu/Sn couples

4 Conclusions

1) Cu₆Sn₅ formed in as-deposited Cu/Sn couples. The amount of Cu₆Sn₅ increased with aging time. The Cu₆Sn₅ layer was continuous, but non-uniform, with protrusions extending into the Sn matrix. The protrusions were likely due to a combination of grain boundary diffusion and interstitial diffusion of Cu atoms into the Sn layer.

2) When Cu/Sn couples were annealed at 373 K and 398 K, Cu₆Sn₅ was still the only intermetallic that formed and the layer was continuous and non-uniform.

3) When Cu/Sn couples were annealed at 423-498 K, two continuous and uniform layers, Cu₆Sn₅ and Cu₃Sn, formed. The Cu₃Sn phase formed by consuming Cu₆Sn₅ phase and the remaining Cu. Kirkendall voids formed near the Cu₃Sn/Cu interface and inside the Cu₃Sn layer due to the much faster Cu diffusion rates relative to Sn.

4) The Cu₆Sn₅ and Cu₃Sn layers both followed parabolic growth kinetics during annealing of Cu/Sn couples from 423 K to 498 K. The growth rate constants were determined as k=4.3×10^-12exp[-41.4×10³/(RT)] for Cu₆Sn₅ and k=3.3×10^-3exp[-90.4×10³/(RT)] for Cu₃Sn.

Acknowledgments

The authors wish to thank the Natural Sciences and Engineering Research Council (NSERC) of Canada and Micralyne, Inc. for providing research funding and Si substrates for electroplating (Micralyne).

References

[1] SIMIC V, MARINKOVIC Z. Room-temperature reactions in thin metal couples [J]. J Mater Sci, 1998, 33: 561-624.

[2] KIM J, LEE C C. Fluxless Sn-Ag bonding in vacuum using electroplated layers [J]. Mater Sci Eng A, 2007, 448: 345-350.

[3] CHAO B, CHAE S H, ZHANG X F, LU K H, IM J, HO P S. Investigation of diffusion and electromigration parameters for Cu-Sn intermetallic compounds in Pb-free solders using simulated annealing [J]. Acta Mater, 2007, 55: 2805-2814.

[4] CHEN C M, CHEN S W. Electromigration effect upon the Sn/Ag and Sn/Ni interfacial reactions at various temperatures [J]. Acta Mater, 2002, 50: 2461-2469.

[5] TANG W M, HE A Q, LIU Q, IVEY D G. Fabrication and microstructures of sequentially electroplated Au-rich, Au/Sn alloy solders [J]. J Mater Sci: Mater Electron, 2008, 19:1176-1183.

[6] TANG W M, HE A Q, LIU Q, IVEY D G. Fabrication and microstructures of sequentially electroplated Sn-rich, Au/Sn alloy solders[J]. J Electron Mater, 2008, 37: 837-844.

[7] LIAO C N, WEI C T. Effect of intermetallic compound formation on electrical properties of Cu/Sn interface during thermal treatment [J]. J Electron Mater, 2004, 33: 1137-1143.

[8] LAURILA T, VUORINEN V, KIVILAHTI J K. Interfacial reactions between lead-free solders and common base materials [J]. Mater Sci Eng R, 2005, 49: 1-60.

[9] TU K N, THOMPSON R D. Kinetics of interfacial reaction in bimetallic Cu-Sn thin-films [J]. Acta Metall, 1982, 30: 947-952.

[10] BANDYOPADHYAY A K, SEN S K. A study of intermetallic compond formation in a copper-tin bimetallic couple [J]. J Appl Phys, 1990, 67: 3681-3688.

[11] LIAO C N, WEI C T. An isochronal kinetic study of intermetallic compound growth in Sn/Cu thin film couples [J]. Thin Solid Films, 2006, 515: 2781-2785.

[12] TAKENAKA T, KANO S, KAJIHARA M, KUROKAWA N, SAKAMOTO K. Growth behavior of compound layers in Sn/Cu/Sn diffusion couples during annealing at 433-473 K [J]. Mater Sci Eng A, 2005, 396: 115-123.

[13] DJURFORS B, IVEY D G. Pulsed electrodeposition of the eutectic Au/Sn solder for optoelectronic packaging [J]. J Electron Mater, 2001, 30: 1249-1254.

[14] SUN W, IVEY D G. Microstructural study of co-electroplated Au/Sn alloys [J]. J Mater Sci, 2001, 36: 757-766.

[15] HE A, LIU Q, IVEY D G. Development of stable, non-cyanide solutions for electroplating Au-Sn alloy films [J]. J Mater Sci: Mater Electron, 2006, 17: 63-70.

[16] TANG W M, HE A Q, LIU Q, IVEY D G. Room temperature interfacial reactions in electrodeposited Au/Sn couples [J]. Acta Mater, 2008, 56: 5618-5827.

[17] NAKAHARA S, MCCOY R J, BUENE L, VANDENBREG J M. Room-temperature-diffusion studies of Au-Sn thin-film couples [J]. Thin Solid Films, 1981, 84: 185-196.

[18] SUH J O, TU K N, TAMURA N. Preferred orientation relationship between Cu₆Sn₅ scallop-type grains and Cu substrate in reactions between molten Sn-based solders and Cu [J]. J Appl Phys, 2007, 102: 063511.

[19] DYSON B F, ANTHONY T R, TURNBULL D. Interfacial diffusion of copper in tin [J]. J Appl Phys, 1967, 38: 3408.

[20] CHAO B, CHAE S H, ZHANG X F, LU K H, DING M, IM J, HO P S. Electromigration enhanced intermetallic growth and void formation in Pb-free solder joints [J]. J Appl Phys, 2006, 100: 084909.

[21] ZENG K J, STIERMAN R, CHIU T C, EDWARDS D, ANO K, TU K N. Kirkendall void formation in eutectic SnPb solder joints on bare Cu and its effect on joint reliability [J]. J Appl Phys, 2005, 97: 024508.

[22] LIU C Y, CHEN J T, CHUANG Y C, KE L, WANG S J. Electromigration-induced Kirkendall voids at the Cu/Cu₃Sn interface in flip-chip Cu/Sn/Cu joints [J]. Appl Phys Lett, 2007, 90: 112114.

[23] BADER S, GUST W, HIEBER H. Rapid formation of intermetallic compound by interdiffusion in the Cu-Sn and Ni-Sn systems [J]. Acta Metall Mater, 1995, 43: 329-337.

[24] LIU P L, SHANG J K. Segregant-induced cavitation of Sn/Cu reactive interface [J]. Scripta Mater, 2005, 53: 631-634.

[25] PAUL A, KODENTSOV A A, VAN LOO F J J. Intermetallic growth and Kirkendall effect manifestations in Cu/Sn and Au/Sn diffusion couples [J]. Z Metallkd, 2004, 95: 913-920.

[26] OH M. Intermetallic growth and Kirkendall effect manifestations in Cu/Sn and Au/Sn diffusion couples [D]. Pennsylvania: Lehigh University, 1994.

[27] ONISHI M, FUJIBUCHI H. Reaction-diffusion in Cu-Sn system [J]. Trans JIM, 1975, 16: 539-547.

Corresponding author: TANG Wen-ming; Tel: +86-551-2901372; Fax: +86-551-2901362; E-mail: wmtang69@126.com

DOI: 10.1016/S1003-6326(09)60102-3

(Edited by YANG Hua)