lable at ScienceDirect

Journal of Alloys and Compounds 740 (2018) 500e509

Contents lists avai

Journal of Alloys and Compounds

journal homepage: http: / /www.elsevier .com/locate/ ja lcom

The interfacial reaction between In-48Sn solder and polycrystalline Cusubstrate during solid state aging

Feifei Tian a, b, Cai-Fu Li b, c, Ming Zhou a, Zhi-Quan Liu b, c, *

a Nanjing Electronic Devices Institute, Nanjing 210016, Chinab Institute of Metal Research, Chinese Academy of Sciences, Shenyang 110016, Chinac The Institute of Scientific and Industrial Research, Osaka University, Osaka 567-0047, Japan

a r t i c l e i n f o

Article history:Received 15 March 2017Received in revised form27 December 2017Accepted 30 December 2017Available online 31 December 2017

Keywords:In-48Sn solderIMCsTransformationCu2(In,Sn)Cu(In,Sn)2

* Corresponding author. Institute of Metal Researences, Shenyang 110016, China.

E-mail address: zqliu@imr.ac.cn (Z.-Q. Liu).

https://doi.org/10.1016/j.jallcom.2017.12.3550925-8388/© 2017 Elsevier B.V. All rights reserved.

a b s t r a c t

Intermetallic compounds (IMCs) with three different morphologies were formed through an interfacialreaction between In-48Sn solder and polycrystalline Cu substrate, which are chunk-type Cu(In,Sn)2 andduplex structural Cu2(In,Sn). Two types of phase transformations between Cu(In,Sn)2 and Cu2(In,Sn)IMCs occurred during low temperature aging (20 �C and 40 �C) and high temperature aging (60 �C, 80 �Cand 100 �C), respectively. At lower aging temperature, the duplex structural coarse-grain and fine-grainCu2(In,Sn) transformed into Cu(In,Sn)2, and a characteristic evolution model was provided. At a higheraging temperature, Cu(In,Sn)2 transformed completely into Cu2(In,Sn) within a very short time period.Based on the designed Cr-marker experiments, the growth kinetics of coarse-grain and fine-grainCu2(In,Sn) sublayers at higher aging temperatures were investigated. It was concluded that the growthmechanism of Cu2(In,Sn) is temperature dependent, which is volume-diffusion-controlled at 60 �Ce80 �C, whereas grain boundary diffusion becomes rate-controlling at 100 �C. Through an elaboratedesign of aging test at 80 �C and 20 �C alternately, the reversible transformation between Cu2(In,Sn) andCu(In,Sn)2 was observed and characterised.

© 2017 Elsevier B.V. All rights reserved.

1. Introduction

It is known that the intermetallic compound (IMC) layer acts asan important metallurgical bonding layer between soldering alloyand substrate metal, which has obvious effects on the mechanicalproperties of solder joints and their serving reliability [1e4]. Manyresearchers have systematically and widely investigated the inter-facial reactions of Sn-based solders, such as SnBi [5], SnAgCu [6,7],Sn-Cu [8], mainly focusing on the nucleation and morphology ofIMCs during soldering as well as the growth kinetics of IMC layersduring the solid-state aging process. However, due to thecomplexity of In-Sn-Cu ternary phase diagram and the high cost ofthe In element, only few researchers and manufacturers have paidattention to In-48Sn/Cu system. The binary eutectic In-48Sn solderhas the advantage of a low melting point and better mechanicalproperties [9e11], which can be utilized in many low-temperaturefields, such as thermal sensitive components, multiple reflowing

ch, Chinese Academy of Sci-

and low-temperature packaging. Hence, it is essential to study theinterfacial reaction between In-48Sn solder and polycrystalline Cusubstrate to clarify IMC formation and phase transformationtherein.

Owing to the participation of the In element in interfacial re-actions between solder and substrates, the In-48Sn solder jointexhibits specific characteristics such as the morphologies andphase species of IMC, which influence the mechanical property andservice reliability of the interconnects in application. Moreover, thechemical reaction of IMCs often takes place at the interface invarious systems during solid-state aging [12e15], which also playsan important role in a packaging structure's life. In microelectronicapplications, the solder joints suffer from different aging temper-atures. Thus, the main diffusing species coming from the solderalloy or substratemetal may change as a function of temperature, topromote different chemical reactions at the interface. Hence,reversible chemical reactions between IMCs can take place, a sub-ject which is seldom investigated.

In this study, the interfacial reaction between In-48Sn andpolycrystalline Cu was studied systematically, mainly focusing onthe IMC morphologies after reflowing, phase transformations

F. Tian et al. / Journal of Alloys and Compounds 740 (2018) 500e509 501

between Cu(In,Sn)2 and Cu2(In,Sn) at different solid-state agingtemperatures, as well as the growth kinetics of coarse-grain andfine-grain Cu2(In,Sn) sublayers. Considering the elemental diffusionduring the interfacial reaction, a schematic diagram was proposedto elucidate the characteristic phase transforming process fromduplex structural Cu2(In,Sn) into Cu(In,Sn)2.

2. Experimental procedures

2.1. Cu substrate with Cr-marker preparation

The high purity, oxygen-free polycrystalline copper plates(>99.99%) were cut by an electro-discharging machine into sheetsof size 40 � 4 � 2 mm3. Then, the Cu surfaces were ground withgrade SiC papers successively and polished carefully with 1 and0.5 mm diamond paste. Thereafter, the Cu surfaces were rinsed inacetone and distilled water and then dried; half of these were laterelectroplated by a thin Cr layer to study the growthmechanism andgrowth kinetics of coarse- and fine-grain Cu2(In,Sn) sublayers afteraging at 60 �C, 80 �C and 100 �C.A precise Cr-marker experiment[16] was designed to distinguish the duplex structural Cu2(In,Sn)sublayers in In-48Sn/Cu system, as shown in Fig. 1. First, half of thepolished Cu surface was masked by an electroplating tape (Fig. 1a).Then, a thin Cr layer was electroplated on the other polished Cusurface (Fig. 1b). Thereafter, the electroplating tape was removed,and the polished Cu surface was cleaned with distilled water anddried.

2.2. In-48Sn solder foil preparation

The In-48Sn solder alloy was prepared using high purity In andSn (>99.99%) according to weight ratio in a resistance vacuumfurnace with vacuum degree 10�3 torr. The In-48Sn solder stripewas rolled into a thickness of about 200 mmby amanual rolling milland cut into the size of 40 mm � 4 mm with scissors. Afterwards,the In-48Sn thin foil was cleaned ultrasonically in acetone andalcohol solvents respectively for 3e5 min to dissolve the remainingand then dried in air.

2.3. Wetting sample preparation

Wetting samples were prepared by putting the In-48Sn thin foilon polished Cu substrates, whose half surface was electroplated by

Fig. 1. Schematic diagram to elucidate the e

a thin Cr layer as a marker to distinguish the original liquid/solidinterface of In-48Sn solder and Cu. The samples were then heatedup to 160 �C for 5s on a heating plate; thereafter, the formed solderjoint (Fig. 1c) was cooled in air up to an ambient temperature of20 �C. Afterwards, the samples were cleaned ultrasonically inacetone for 5 min in order to dissolve the remaining rosin mildlyactivated (RMA) on the sample surfaces. To observe the wettingfront between the electroplated Cr layer and the solder joint, it isessential to carry out a second electroplating Cr layer as shown inFig. 1d. Finally, the solder joint samples were cut into 4 � 4 mm2

pieces and then aged in thermal ovens at a temperature range of20 �C~100 �C for different periods.

2.4. Microstructural observation

The cross-sectional samples for the scanning electron micro-scope (SEM) study were ground with SiC papers successively andthen polished carefully with 0.05 mm Al2O3 powder suspension.Top-view SEM samples first mechanically ground away the excessIn-48Sn solder to about 100 mm thickness, following which theremaining solder alloy was carefully etched with 20% H2O2 þ 80%CH3COOH (vol%), which has the advantage of having a low etchingrate and less effect on Cu substrate. The cross-sectional and top-view SEM samples were rinsed with distilled water in ultrasonicbath for 10 min to remove the residual etchant solution and Al2O3powder suspension, and then dried in air. All clean samples werestudied using LEO super35 and Quanta 600 SEMs with energy-dispersive X-ray spectroscopy (EDS) system for compositionalanalyses.

2.5. IMC thickness measurement

The thickness of IMC layers was measured using image analysessoftware as described earlier [17].

3. Results and discussion

3.1. Morphology of interfacial IMCs after reflowing and their EDSanalysis

After reflowing at 160 �C for 5s, the cross-sectional SEM image ofthe interface is shown in Fig. 2a, inwhich two distinct IMC layerseacontinuous Cu(In,Sn)2 layer next to In-48Sn solder and a Cu2(In,Sn)

xperimental procedures of Cr-marker.

Fig. 2. Cross-sectional SEM image of the interface after soldering (a), as well as the top-view SEM images of reflowed chunk-type Cu(In,Sn)2 (b), coarse-grain Cu2(In,Sn) (c), and fine-grain Cu2(In,Sn) (d).

F. Tian et al. / Journal of Alloys and Compounds 740 (2018) 500e509502

layer adjacent to Cu substrate e were observed. The formed inter-facial IMC species between In-48Sn solder and single crystalline Cuwere investigated thoroughly utilizing electron microscopy afterreflowing [18,19]; thesewere identified as Cu(In,Sn)2 and Cu2(In,Sn)IMCs. These formed IMC species are also formed in case of thepolycrystalline Cu substrate under the same conditions accordingto microstructural characterization. However, the difference be-tween the single crystalline Cu and polycrystalline Cu is the pref-erential orientation of the formed coarse-grain Cu2(In,Sn) grains.There are orientation relationships during the nucleation andgrowth of coarse-grain Cu2(In,Sn) compound on single crystallineCu [18], but the formed coarse-grain Cu2(In,Sn) grains displaychaotic rod-type morphology and have no preferential orientationon polycrystalline Cu substrate. The Cu(In,Sn)2 layer has a lighterSEM contrast than the Cu2(In,Sn) layer, whose interface was indi-cated by arrows as shown in Fig. 2a. The average thickness of bothof them is about 1 mm. By etching the solder joint from top tobottom, layer by layer, the exact morphologies of the formed IMCscould be revealed as shown in Fig. 2bed. It was found that layer IIcontains Cu(In,Sn)2 IMC (Fig. 2b) and has a chunk-typemorphology.However, top-view SEM images show that layer I of Cu2(In,Sn) IMCcontains coarse-grain (Fig. 2c) and fine-grain (Fig. 2d) sublayerswith rod-type and granular-type morphologies respectively,referred to as “duplex structure” in this study.

As SEM morphology observations could not provide sufficientinformation to determine the interfacial IMC species, compositionanalyses were further carried out using EDS, and the results areshown in Fig. 3. Elemental analyses using EDS revealed that the

chunk-type, rod-type and granular-type IMCs have a compositionclose to 37Cu-46In-17Sn (at.%), 70Cu-12In-18Sn (at.%) and 67Cu-12In-21Sn (at.%) respectively, among which the rod-type andgranular-type IMCs have similar compositions. Moreover, selectedarea electron diffraction (SAED) patterns using transmission elec-tron microscopy (TEM) showed that chunk-type IMC had thetetragonal CuIn2 type of crystal structure with unit cell dimensionsof a ¼ 0.6645 nm, c ¼ 0.5376 nm. The rod-type and granular-typeIMCs had hexagonal Cu2In type of structure with unit cell di-mensions of a¼ 0.4292 nm, c¼ 0.5232 nm [19]. Hence, two types ofIMCs with three kinds of morphology were formed on poly-crystalline Cu substrate after reflowing, which are Cu(In,Sn)2,coarse-grain Cu2(In,Sn) and fine-grain Cu2(In,Sn).

3.2. Solid-state aging at higher temperatures of 60 �C, 80 �C and100 �C

3.2.1. Phase transformation from Cu(In,Sn)2 into Cu2(In,Sn) athigher aging temperature

When comparing Fig. 2a (after reflow) with Fig. 4, it can beclearly seen that the previous continuous Cu(In,Sn)2 layer dis-appeared and duplex structural Cu2(In,Sn) sublayers became theonly dominantly growing IMC at the interface after solid-state ag-ing at 60 �C, 80 �C and 100 �C for just 1 day. This is also the case for ahigher aging temperatures and a longer aging time. Furthermore,the duplex structural Cu2(In,Sn) sublayer grew thicker as the agingtemperature increased. Thus, we conclude that the higher the agingtemperature is, the thicker the formed IMC layer will be under the

Fig. 3. EDS analyses of interfacial IMCs after soldering Cu(In,Sn)2 (a), coarse-grain Cu2(In,Sn) (b), and fine-grain Cu2(In,Sn) (c).

F. Tian et al. / Journal of Alloys and Compounds 740 (2018) 500e509 503

same aging time.Based on our earlier study [20] on phase transformation be-

tween Cu(In,Sn)2 and Cu2(In,Sn) compounds formed on singlecrystalline Cu substrate during the solid-state aging, chemical re-actions took place between these compounds. It was found thatboth Cu(In,Sn)2 and Cu2(In,Sn) IMCs grew simultaneously andcoexisted on (100)Cu and (111)Cu after aging at 60 �C [20]. How-ever, on the polycrystalline Cu substrate, the Cu(In,Sn)2 IMCstransformed into Cu2(In,Sn); accordingly, the following partialchemical reaction should take place during the aging at 60 �C.

3Cudiffuse þ CuIn2 ¼ 2Cu2In (1)

According to the theoretical model of chemical reactions at theinterface [14,15], it can be derived that dx/dt > 0 and dy/dt < 0 at60 �C; thus, the following two formulae can be obtained [19]:

83k1In1 þ k1Cu2

k1In2>xy

(2)

23

k1Cu2k1In2 þ k1Cu3

>xy

(3)

Where k1In1, k1Cu2, k1In2 and k1Cu3 are the physical (diffusion)constants (m2/s) [14,15], indicating the IMC layer controlled bydiffusion, x and y represent the thickness of Cu2(In,Sn) andCu(In,Sn)2, respectively. It is known that the thickness of bothCu2(In,Sn) and Cu(In,Sn)2 is about 1 mm after reflowing. Duringsolid-state aging at 60 �C, 80 �C and 100 �C for some days,Cu(In,Sn)2 consumption and Cu2(In,Sn) growth will eventually leadto x > y or even x[ y. Putting x¼ y into the inequality (3), it can bederived as:

k1Cu2 >3=2ðk1In2 þ k1Cu3Þ (4)

Owing to k1Cu3 � 0, it can be further described as:

k1Cu2 > 3/2 k1In2 (5)

Therefore, the stronger diffusion ability of Cu than In and Snatoms helps the chemical reaction (1) proceed, resulting inCu(In,Sn)2 consumption as well as Cu2(In,Sn) growth. When

Fig. 4. Cross-sectional SEM images taken at the In-48Sn/polycrystalline Cu interfacesafter aging at 60 �C (a), 80 �C (b) and 100 �C (c)for 1 day.

F. Tian et al. / Journal of Alloys and Compounds 740 (2018) 500e509504

Cu(In,Sn)2 transforms into Cu2(In,Sn) completely, the following twopartial chemical reactions take place at the interface to makeduplex structural Cu2(In,Sn) sublayer grow continuously on poly-crystalline Cu substrate during aging at 60 �C, 80 �C and 100 �C.

2Cudiffuse þ Insurface ¼ Cu2In (6)

Indiffuse þ 2Cusurface ¼ Cu2In (7)

Since the consumption rate of Cu(In,Sn)2 is larger than itsgrowth rate due to the stronger diffusion ability of Cu atom than In/Sn atoms, only Cu2(In,Sn), not Cu(In,Sn)2, was formed at the inter-face. Hence, duplex structural Cu2(In,Sn) became the only domi-nant IMC under this condition, and the growth mechanism andgrowth kinetics of coarse- and fine-grain Cu2(In,Sn) sublayers wereinvestigated individually.

3.2.2. The growth kinetics of coarse- and fine-grain Cu2(In,Sn) athigher-aging temperatures

The this Cr-marker experiment clearly indicates the originalreflowing interface [16] located between the coarse- and fine-Cu2(In,Sn) sublayers, especially after the etching, as shown inFig. 5a and b; the growth kinetics of coarse- and fine-Cu2(In,Sn)IMCs can be investigated precisely. Solid-state aging of In-48Sn/Cusolder joints was carried out at 60 �C, 80 �C and 100 �C, and thethickness change of coarse- and fine-grain Cu2(In,Sn) IMC as afunction of aging time was shown in Fig. 5c and d, respectively. It iswell known that the growth of an IMC layer can be generallydescribed by a power law:

Dx ¼ ktnorlogDx ¼ Aþ n log t (8)

Where, Dx is the IMC thickness, k is the growth rate constant, t isthe reaction time, and n is the growth rate exponent, whose valuerepresents the different rate-limiting mechanisms for the IMCgrowth (n ¼ 0.5 indicates diffusion controlled growth, while n ¼ 1represents reaction-controlled growth). According to Fig. 5c and d,both coarse- and fine-grain Cu2(In,Sn) sublayers are thick duringsolid-state aging. The higher the aging temperature is, the thickerboth layers grow at the same aging time. The value of the growthrate exponent n could be obtained from the slope of logDx versuslog t, which is listed in Table 1 for different aging temperatures. It isclearly seen that the rate-limiting growth mechanism of Cu2(In,Sn)sublayers is different at various temperatures.

At aging temperatures of 60 �C and 80 �C, the growth of bothcoarse- and fine-grained Cu2(In,Sn) sublayers is controlled by vol-ume diffusion, since the values of n are around 0.5 ranging from0.45 to 0.54. However, the rate-limiting diffusing elements aretotally different for the coarse- and fine-grain Cu2(In,Sn) IMCs. Asfor the coarse-grain Cu2(In,Sn), these are Cu atoms that diffusethrough both coarse- and fine-grain Cu2(In,Sn) sublayers to theCu2(In,Sn)/In-48Sn interface for the reaction with In-48Sn solder(see formula (6)), in order to control the growth of coarse-grainCu2(In,Sn) sublayer at the solder side. As for the fine-grainCu2(In,Sn), it is In and Sn atoms that diffuse to Cu2(In,Sn)/Cuinterface and react with Cu (see formula (7)), controlling thegrowth of fine-grained Cu2(In,Sn) at the Cu side.

At the aging temperature of 100 �C, a dramatic decrease in thevalues of n (0.31e0.33) takes place, see Table 1. The growthmechanism deviates appreciably from the parabolic growth law,which implies a different rate-limiting mechanism from that at60 �C and 80 �C. According to the studies on solid-liquid reactions inCu-Sn and Ni-Sn [21] as well as Cu-SnPb [22] systems, it was foundthat grain boundary diffusion, whose growth rate exponent (n) islower than that of volume diffusion-controlled mechanism, can bethe predominant transport mechanism under certain conditions.This theory should also be adaptable to the solid-state aging of In-48Sn/Cu solder joint at higher temperatures, considering that100 �C is close to the melting temperature of In-48Sn solder (about120 �C). As shown in Fig. 5a and b, after 5 and 10 days aging at100 �C, some coarse Cu2(In,Sn) grains extrude into the solder andthe interface between Cu2(In,Sn) and In-48Sn solder becomes moreuneven, which is a manifestation of grain boundary diffusion forheterogeneous growth of coarse-grain Cu2(In,Sn). The effect ofgrain boundary diffusion is also strengthened by the ceaselessnucleation of nano-scale fine Cu2(In,Sn) grains during aging, sincesubstantial concentration of grain boundaries can supply fastdiffusion of reacting elements needed for the interfacial reaction.Thus, it can be concluded that grain boundary diffusion is respon-sible for the rate-limiting growth mechanism of coarse- and fine-grain Cu2(In,Sn) sublayers at 100 �C. According to the above re-sults and discussion, it is evident that the growth mechanisms ofIMC at In-48Sn/polycrystalline Cu interfaces are temperature-dependent at higher aging temperatures.

3.3. Solid-state aging at lower temperatures of 20 �C and 40 �C

3.3.1. Phase transformation fromCu2(In,Sn) into Cu(In,Sn)2 atlower-aging temperature

Fig. 6 shows the cross-sectional images of Cu(In,Sn)2 andCu2(In,Sn) layers on polycrystalline Cu substrate after aging at 40 �Cfor up to 90 days. In Fig. 6a, the interface between Cu(In,Sn)2 layeradjacent to the solder and Cu2(In,Sn) layer next to Cu substratesafter 5 days' aging is indicated by horizontal arrows, at which some

Fig. 5. Cross-sectional SEM image showing the interface between coarse and fine-grain Cu2(In,Sn) sublayers aged at 100 �C for 5days (a) and 10 days (b), and the plots of thethickness of coarse (c) and fine-grain Cu2(In,Sn) (d) formed on polycrystalline Cu substrates as a function of aging time during solid-state aging at 60, 80and 100 �C.

Table 1Growth rate exponent (n) of duplex structural Cu2(In,Sn) at different agingtemperatures.

Temperatures (�C) coarse-grain Cu2(In,Sn) fine-grain Cu2(In,Sn)

60 0.45 0.5480 0.46 0.48100 0.33 0.31

F. Tian et al. / Journal of Alloys and Compounds 740 (2018) 500e509 505

coarse-grain Cu2(In,Sn) grains extrude into the Cu(In,Sn)2 layer. Asshown in Fig. 6b, after 10 days' aging, it was seen that the original,continuous Cu2(In,Sn) layer became discontinuous in region A.Furthermore, the thickness of the Cu(In,Sn)2 layer increased andthat of the Cu2(In,Sn) layer decreased with increasing aging time.Fig. 6c shows the chemical reaction of IMC layers taking place afteraging at 40 �C for 60 days. It can be clearly seen that with theincreasing appearance of the discontinuous regions of theCu2(In,Sn) layer, its thickness decreased dramatically due to itstransformation into Cu(In,Sn)2. After 90 days' aging at 40 �C, asshown in Fig. 6d, only very few Cu2(In,Sn) islands remained, whilethe rest of them had completely transformed into Cu(In,Sn)2eventually. It was concluded that with increasing aging time,originally reflowing Cu2(In,Sn) eventually transformed intoCu(In,Sn)2, when aged at 40 �C. Moreover, it was also observed thatthe interface between Cu2(In,Sn) and Cu substrate changed fromrelatively even to uneven, and the interface concaved downwardsto Cu substrate as shown in region B of Fig. 6d. According to theelemental analyses, new Cu(In,Sn)2 was formed in these regions. Itmeant that the original Cu2(In,Sn) layer transformed into Cu(In,Sn)2completely, and the chemical reaction was still going on with Cusubstrates due to the fast diffusion of In and Sn atoms. Suchchemical reaction occurring on polycrystalline Cu substrate wasalso observed on (100)Cu and (111)Cu substrates at a long-termaging at 40 �C and on polycrystalline Cu after the aging at 20 �C.

According to our previous study on (100)Cu and (111)Cu sub-strates [20], the following partial chemical reaction takes placewhen Cu2(In,Sn) transforms into Cu(In,Sn)2 during aging at 40 �C:

3Indiffuse þ Cu2In ¼ 2CuIn2 (9)

According to the theoretical model of chemical reactions at theinterface, it can be derived that dx/dt < 0 and dy/dt > 0; thus, thefollowing two formulae can be obtained [19]:

32k1In2 þ k1Cu3

k1Cu2>yx

(10)

38

k1In2k1In1 þ k1Cu2

>yx

(11)

Where k1In1, k1Cu2, k1In2 and k1Cu3 are the physical (diffusion)constants (m2/s), indicating that the IMC layer is controlled bydiffusion; x and y represent the thickness of Cu2(In,Sn) andCu(In,Sn)2, respectively. Putting y ¼ x into the inequality (11), it canbe derived as:

k1In2 >8=3ðk1In1 þ k1Cu2Þ (12)

Owing to k1In1 � 0, it can be safely concluded that:

k1In2 >8=3k1Cu2 (13)

It means that the diffusion ability of the In atom is at leastseveral times stronger than that of Cu, making Cu2(In,Sn) totransform into Cu(In,Sn)2 with chemical reaction (9) taking theplacewhen aged at 20 �C and 40 �C. After Cu2(In,Sn) transforms intoCu(In,Sn)2 completely, the following two partial chemical reactionstake place at the interface to make Cu(In,Sn)2 grow continuouslyduring aging at 20 �C and 40 �C on polycrystalline Cu substrate.

Fig. 6. Cross-sectional SEM micrographs taken at the In-48Sn/polycrystalline Cu interfaces after aging at 40 �C for 5 days (a), 10days (b), 60 days (c) and 90days (d).

F. Tian et al. / Journal of Alloys and Compounds 740 (2018) 500e509506

2Indiffuse þ Cusurface ¼ CuIn2 (14)

Cudiffuse þ 2Insurface ¼ CuIn2 (15)

3.3.2. Evolution model of phase transformation from duplexstructure Cu2(In,Sn) into Cu(In,Sn)2

As mentioned above, the transformation process of the duplexstructure Cu2(In,Sn) into Cu(In,Sn)2during aging at 20 �C and 40 �Cshows its own characteristics, which have a close relationship withthe special coarse- and fine-grain Cu2(In,Sn) microstructures. Basedon the experimental observations, an evolution model was elabo-rated and presented in Fig. 7. At the initial aging stage, as shown inFig. 7a, some In atoms diffuse vertically from the solder into the topof the coarse-grain Cu2(In,Sn) sublayer; subsequently, a partialchemical reaction (9) takes place and Cu(In,Sn)2 IMC is formed.Here, Indiffuse is used to represent both In and Sn species consid-ering the following two factors. On one hand, the atomic percent-age of the In element in Cu(In,Sn)2 is about 50% (more than that ofSn) and is nearly equivalent to the content of Sn element inCu2(In,Sn). On the other hand, Cu(In,Sn)2 and Cu2(In,Sn) are CuIn2and Cu2In based crystal structures, respectively.

It is known that grain boundaries allow amuch greater diffusion

Fig. 7. Schematic diagram to illustrate the evolution process of chemical transformation

rate compared to the grain itself. As the grooves between the grainboundaries and also the grain boundaries of coarse Cu2(In,Sn)grains can act as the straight and quick path for the diffusion of Inand Sn atoms, chemical reaction (9) is promoted faster at grainboundaries than in coarse-grain Cu2(In,Sn) itself, as shown in Fig. 6.Based on our previous TEM [19] and Cr-marker observations [16],there exists a distinct interface between the coarse- and fine-grainCu2(In,Sn) sublayers. When coarse-grain Cu2(In,Sn) completelytransformed into Cu(In,Sn)2in some regions, vertically diffusingalong grain boundaries In atoms of coarse-grain Cu2(In,Sn) reachthe top of fine-grain Cu2(In,Sn) sublayer, and the chemical reaction(9) between In atoms and fine-grain Cu2(In,Sn) takes place. The Inatoms diffusing through the fine-grain Cu2(In,Sn) sublayer woulddiverge into two directions: one diffusing direction is vertical to Cusubstrate and the other one is parallel to Cu substrate as shown byarrows in Fig. 7b. As the fine-grain Cu2(In,Sn) sublayer has moregrain boundaries, In atoms can diffuse towards the two directionsmentioned above quickly through the grain boundaries and reactwith the fine Cu2(In,Sn) grains to form new Cu(In,Sn)2 IMC. How-ever, the grain size of coarse-grain Cu2(In,Sn) is larger and thequantity of grain boundaries in the layer is small. Thus, the trans-formation of coarse-grain Cu2(In,Sn) sublayer is slow compared tothat in fine-grain Cu2(In,Sn) sublayer, as revealed in Fig. 7b.

When the fine-grain Cu2(In,Sn) sublayer transforms into

from duplex structure Cu2(In,Sn) into Cu(In,Sn)2 during aging at 20 �C and 40 �C.

F. Tian et al. / Journal of Alloys and Compounds 740 (2018) 500e509 507

Cu(In,Sn)2 completely, the regions are transformed into theCu(In,Sn)2 IMC. Then, the diffusing In atoms migrate through theCu(In,Sn)2 layer to reach the top of the Cu substrate and initiate anew partial chemical reaction (14), forming the Cu(In,Sn)2 IMC andmaking the Cu surface concave go down, as shown in Figs. 6d and7c (region B). With the proceeding chemical reaction (9), thecontinuous duplex structure Cu2(In,Sn) sublayers were divided intocertain discontinuous regions, and in some part of the micro-structure, the duplex structure Cu2(In,Sn) becomes sandwichedbetween two Cu(In,Sn)2 layers to form a special layered structure ofCu(In,Sn)2/Cu2(In,Sn)/Cu(In,Sn)2. Moreover, some coarse Cu2(In,Sn)grains appear to be insulating islands surrounded by newly formedCu(In,Sn)2 after a complete chemical transformation from fine-grain Cu2(In,Sn) into Cu(In,Sn)2, as clearly shown in Fig. 7c. Afterthe duplex structure Cu2(In,Sn) sublayers were consumed, only theCu(In,Sn)2 layer remained, as seen in Fig. 7d, but chemical reactions(14) and (15) still continued at the interfaces.

Due to the existence of coarse-grain and fine-grain duplexstructures, the transformation process fromCu2(In,Sn) intoCu(In,Sn)2 is specific and different, considering the diffusion of Inatoms. In other reaction diffusion systems, the consumption andgrowth of compound layers at the interfaces is uniform and nopenetrating phenomenon is observed. For example, whenCu(In,Sn)2 IMC transforms into duplex structure Cu2(In,Sn) sub-layers completely, Cu2(In,Sn) becomes the only dominant IMCwithout any IMC islands. As illustrated in Fig. 7, the fine-grainCu2(In,Sn)sublayer with more diffusion paths has a great effect onthe diffusion rate of In atoms, affecting the chemical reactionstherein.

3.4. Reversible transformation between Cu2(In,Sn) and Cu(In,Sn)2

3.4.1. Phenomenon of reversible transformation during agingSince the IMC transformation is different at higher and lower

temperatures, as described in Sections 3.2 and 3.3, special agingexperiments were designed to monitor the reversible trans-formation at different temperatures. The reflowing samplessuffering from the first aging process at 80 �C for 2 and 5 days areshown in Fig. 8a and b respectively; Cu(In,Sn)2 transformedcompletely into duplex structure Cu2(In,Sn), becoming the onlydominant IMC at the interface. Then, these corresponding sampleswere aged at 20 �C for 2 years, and Cu(In,Sn)2 appeared again be-tween the duplex structure Cu2(In,Sn) and In-48Sn solder as shownin Fig. 8c and d. As this transformation proceeded, Cu2(In,Sn) layerchanged from being continuous to discontinuous, and some exist-ing duplex structure Cu2(In,Sn) was sandwiched between theneighbouring two Cu(In,Sn)2 layers. For example, a thin unreactedfine-grain Cu2(In,Sn) sublayer lies vertically down in region A ofFig. 8c. Moreover, some other Cu2(In,Sn) grains were enveloped byCu(In,Sn)2, similar to isolated islands, whose roots are adjacent tothe Cu substrate, as indicated by horizontal arrowheads in Fig. 8c.The above transformation phenomena also appear in Fig. 8d. Be-sides, the interface between Cu2(In,Sn) and Cu substrate changedfrom relatively even to uneven and concaved downwards to Cusubstrate, as shown in region B of Fig. 8d. In this concaved region B,Cu(In,Sn)2penetrated into Cu substrate after a complete trans-formation from Cu2(In,Sn), and the interfacial reaction between Cusubstrate and In-48Sn solder proceeded with the formation of anew Cu(In,Sn)2. Comparing Fig. 6 with Fig. 8aed, we conclude thatthe same chemical reaction occurs between Cu2(In,Sn) andCu(In,Sn)2 IMCs, and the evolution process from duplex structuralCu2(In,Sn) sublayers to Cu(In,Sn)2 is also similar when aged at 20 �Cand 40 �C. From the above observation, the Cu(In,Sn)2 layer domi-nated over the whole IMC layer after long-term aging at 20 �C and40 �C. Third, the corresponding samples after 2 years' aging at 20 �C

were further aged at 80 �C for just 1 day, and it was observed thatthe transformed Cu(In,Sn)2 in Fig. 8c and d almost completelyconverted into Cu2(In,Sn) again as revealed in Fig. 8e and f. FromFig. 8, it can be clearly seen that a reversible transformation be-tween Cu2(In,Sn) and Cu(In,Sn)2 took place when samples weresubjected to a series of aging treatments at different temperatures.Moreover, it was found that the thickness of the interfacial IMClayers in Fig. 8c and d was larger than that in Fig. 8a and b and didnot change when compared to that in Fig. 8e and f.

As discussed in the above section, transformation fromCu2(In,Sn) into Cu(In,Sn)2 or from Cu(In,Sn)2 into Cu2(In,Sn) has adirect and close relationship with the main diffusing species of Cuor In/Sn atoms at different aging temperatures. When the diffusionability of In atoms is stronger than that of Cu, Cu2(In,Sn) transformsinto Cu(In,Sn)2; however, when the diffusion ability of Cu atoms isstronger than that of In, Cu(In,Sn)2 transforms into Cu2(In,Sn),leading to a reversible transformation between Cu2(In,Sn) andCu(In,Sn)2.

3.4.2. Verification of reversible transformation by theoreticalcalculations

The above chemical reactions were quantitatively describedconsidering IMC growth. Table 2 shows the thickness of IMCs onpolycrystalline Cu as a function of aging time during a series ofsolid-state aging process, representing the evolution process ofreversible transformations at the interface. According to thechemical reaction (9), the following equation can be derived:

xCuIn2¼ 2:58xCu2In (16)

It means when 1 mmof duplex structure Cu2(In,Sn) is consumed,2.58 mm of Cu(In,Sn)2 grows. According to the image analysessoftware, the thickness of the transformed Cu(In,Sn)2 layer from theduplex structure Cu2(In,Sn) is about 4.3 mm. Putting the value of4.3 mm of Cu(In,Sn)2 into equation (16), it can be derived that about1.67 mm of Cu2(In,Sn) was consumed. Thus, it is concluded that

ttotal-second ¼ tdup-first�1.67 þ 4.3, this is ttotal-second ¼ tdup-firstþ2.63 (17)

Among which, tdup-first represents the thickness of duplexstructure Cu2(In,Sn) sublayers after the first aging at 80 �C fordifferent time and ttotal-second indicates the thickness of all the IMClayers of the considered above samples (consisting of unreactedduplex structure Cu2(In,Sn) and transformed Cu(In,Sn)2 layer) afteraging at 20 �C for 2 years. The values of tdup-first after the first agingat 80 �C and ttotal-second after the second aging at 20 �C for 2 years ofthe corresponding samples are listed in Table 2. Putting the abovevalues into (17), it can be found that the results are in a goodagreement. It means that the duplex structure Cu2(In,Sn) layer afterthe first aging at 80 �C indeed transformed into Cu(In,Sn)2 after thesecond aging at 20 �C, due to the Cu atom's faster diffusion rate ascompared to In.

On the other hand, according to the chemical reaction (1), thefollowing equations can be derived:

xCu2In ¼ 1:54xCuIn2(18)

It means when 1 mm of Cu(In,Sn)2is consumed, 1.54 mm ofCu2(In,Sn) will grow. According to equation (16), the thickness ofthe transformed Cu(In,Sn)2 layer after the second aging at 20 �C for2 days is about 4.3 mm, which should make 6.6 mm of Cu2(In,Sn)grow after the third aging at 80 �C. The measured thickness ofCu2(In,Sn) from the phase transformation of Cu(In,Sn)2 after thethird aging at 80 �C and the unreacted Cu2(In,Sn) after second agingat 20 �C for 2 days is smaller than the calculated value, according to

Fig. 8. Cross-sectional SEM micrographs taken at the In-48Sn/polycrystalline Cu interfaces after first aging at 80 �C for 2 and 5days (a,b), second aging at 20 �C for 2 years (c,d), andthird aging at 80 �C for only 1 day (e,f).

Table 2The thickness of IMCs at variable aging conditions.

First aging time (days)/IMC thickness (mm) 2 4 5 6

Cu2(In,Sn) first aging at 80 �C 3.34 4.29 4.56 4.99Cu(In,Sn)2 s aging at 20 �C 4.31 4.36 4.38 4.45Total IMC second aging at 20 �C 6.05 7.04 7.49 7.66Cu2(In,Sn) second aging at 20 �C 1.74 2.68 3.11 3.21Total IMC third aging at 80 �C 6.23 7.06 7.33 7.54

F. Tian et al. / Journal of Alloys and Compounds 740 (2018) 500e509508

the equation (18) after the third time aging at 80 �C. This meansthatmost of Cu(In,Sn)2 has transformed into Cu2(In,Sn) leaving onlya few grains of un-reacted Cu(In,Sn)2, just as shown by the dottedregions in Fig. 8e and f, thus making Cu2(In,Sn) to grow continu-ously and become the dominant IMC after aging at 80 �C for alonger time.

From the above observations and analysis, we can conclude thatthe chemical reactions dependent on the aging temperature indeedoccurred at the interfaces between In-48Sn solder and poly-crystalline Cu substrate, as summarized in Table 3. As presented inTable 3, all chemical reactions can be classified into two types: theCu(In,Sn)2 layer grows while the Cu2(In,Sn) layer is consumed, andwhen the Cu(In,Sn)2 layer is consumed, the Cu2(In,Sn) layer grows.

Table 3Summary of phase transformations between Cu(In,Sn)2 and Cu2(In,Sn) compoundson polycrystalline Cu substrate at different temperatures.

Substrate Temperature (�C)

20 and 40 60, 80 and 100

polycrystalline Cu Cu2In/CuIn2 CuIn2/Cu2In

The occurrence of the type of phase transformation has a closerelationship with aging temperatures and eventually depends onthe main diffusing species of Cu and In/Sn atoms. As the SnIn/Cusolder joint was less researched, the chemical reactions providedabove were never characterised earlier. In difference from thesimultaneous growth of both Cu(In,Sn)2 and Cu2(In,Sn) on (100)Cuand (111)Cu during aging at 60 �C, Cu(In,Sn)2 transforms intoCu2(In,Sn) on polycrystalline Cu substrate after aging at 60 �C.Though the phenomenon of simultaneous growth of Cu(In,Sn)2 andCu2(In,Sn) on polycrystalline Cu substrate was not observed duringthe solid-state aging in this study, we believe that both Cu(In,Sn)2and Cu2(In,Sn) can coexist and grow at a certain aging temperatureranging from 40 �C to 60 �C, at which the diffusion ability of In/Snand Cu atoms is close to each other.

4. Conclusions

In summary, the interfacial reaction between In-48Sn solder andpolycrystalline Cu substrate was studied systematically in this pa-per and the following conclusions were drawn, focusing on themorphologies of IMCs after reflowing, the growth kinetics ofcoarse-grain and fine-grain Cu2(In,Sn) sublayers, as well as thephase transformations between Cu(In,Sn)2 and Cu2(In,Sn) atdifferent aging temperatures.

1) After reflowing at 160 �C for 5s, two kinds of IMCs were formedin three sublayers from solder to substrate side; these are asfollows: a Cu(In,Sn)2 layer with tetragonal crystal structure,coarse-grain Cu2(In,Sn) sublayer and fine-grain Cu2(In,Sn) sub-layer with hexagonal crystal structure. The morphology of theCu(In,Sn)2 grains is a chunk-type with the largest grain size,

F. Tian et al. / Journal of Alloys and Compounds 740 (2018) 500e509 509

coarse-grain Cu2(In,Sn) is chaotic and rod-like, whose grain sizeis medium, while fine-grain Cu2(In,Sn) shows granule-typemorphology with the smallest grain size.

2) Kinetics analyses reveal that the growth of both coarse- andfine-grain Cu2(In,Sn) is controlled by a bulk diffusion at lowertemperatures of 60 �C and 80 �C, while at the high temperatureof 100 �C, the grain boundary diffusion dominates over theirgrowth mechanisms.

3) When solid alloys are aged at 20 �C and 40 �C, Cu2(In,Sn)gradually transforms into the Cu(In,Sn)2 compound. At thistemperature, the diffusion rate of In and Sn atoms is higher thanthat of Cu atoms, k1In2 > 8/3k1Cu2, making the Cu(In,Sn)2 layergrow and Cu2(In,Sn) layer to be continuously consumed.Considering the duplex structure of coarse- and fine-grainCu2(In,Sn), a characteristic evolution model was proposed todescribe the transformation of Cu2(In,Sn) into Cu(In,Sn)2 at20 �C and 40 �C.

4) At the aging temperatures of 60 �C, 80 �C and 100 �C, Cu(In,Sn)2can transform into Cu2(In,Sn) quickly within just 1 day. Underthese conditions, the diffusion ability of the Cu atom is muchstronger than that of the In atom, k1Cu2 > 3/2 k1In2, causing theCu2(In,Sn) layer to grow and the Cu(In,Sn)2 later to be continu-ously consumed.

5) The reversible transformation between Cu2(In,Sn) andCu(In,Sn)2 IMCs was first discovered through the special designsof three-step aging experiments performed at different tem-peratures. Its occurrence is closely related to the diffusion abilityof the main reaction species (In/Sn or Cu atoms) at different

aging temperatures, which was verified by quantitative analysesof IMC growth.

Acknowledgements

This work was financially supported by the National Key R&DProgram of China (Grant No. 2017YFB0305700), the InnovationFund Project of IMR, CAS (Grant No. 2015-ZD02) as well as theOsaka University Visiting Scholar Program (Grant No. J135104902).

References

[1] C.M.L. Wu, D.Q. Yu, C.M.T. Law, L. Wang, Mater. Sci. Eng. R 44 (2004) 1.[2] G. Ghosh, J. Electron. Mater. 27 (1998) 1154.[3] M.D. Cheng, S.S. Wang, T.H. Chuang, J. Electron. Mater. 31 (2002) 171.[4] J.W. Ronnie Teo, Y.F. Sun, Acta Mater. 56 (2008) 242.[5] P.J. Shang, Z.Q. Liu,a, D.X. Lia, J.K. Shang, Scripta Mater. 59 (2008) 317.[6] F.J. Wang, Z.S. Yu, K. Qi, J. Alloy. Comp. 438 (2007) 110.[7] J.W. Yoon, S.W. Kim, S.B. Jung, J. Alloy. Comp. 392 (2005) 247.[8] O.M. Abdelhadi, L. Ladani, J. Alloy. Comp. 537 (2012) 87.[9] D.G. Kim, S.B. Jung, J. Alloy. Comp. 386 (2005) 151.

[10] T. Chuang, C. Yu, S. Chang, S. Wang, J. Electron. Mater. 31 (2002) 640.[11] S. Sommadossi, W. Gust, E.J. Mittemeijer, Mater. Chem. Phys. 77 (2002) 924.[12] T.H. Chuang, Y.T. Huang, L.C. Tsao, J. Electron. Mater. 30 (2001) 945.[13] G. Ghosh, J. Electron. Mater. 27 (1998) 1154.[14] V.I. Dybkov, J. Mater. Sci. 21 (1986) 3078.[15] V.I. Dybkov, J. Mater. Sci. 21 (1986) 3085.[16] F.F. Tian, P.J. Shang, Z.Q. Liu, Mater. Lett. 121C (2014) 185.[17] P.J. Shang, Z.Q. Liu, D.X. Li, J.K. Shang, Acta Mater. 57 (2009) 4697.[18] F.F. Tian, Z.Q. Liu, J. Alloy. Comp. 588 (2014) 662.[19] F.F. Tian, P.J. Shang, Z.Q. Liu, J.D. Guo, J. Alloy. Comp. 591 (2014) 351.[20] F.F. Tian, Z.Q. Liu, J.D. Guo, J. Appl. Phys. 115 (2014), 043520.[21] S. Bader, W. Gust, H. Hieber, Acta Metall. Mater. 43 (1995) 327.[22] M. Schaefer, R.A. Fournelle, J. Liang, J. Electron. Mater. 27 (1998) 1167.