Recrystallization of Sn Grains due to Thermal Strain in Sn-1.2Ag-0.5Cu-0.05Ni Solder

Transcription

1 Materials Transactions, Vol. 45, No. 4 (2004) pp to 1390 #2004 The Japan Institute of Metals EXPRESS REGULAR ARTICLE Recrystallization of Sn Grains due to Thermal Strain in Sn-1.2Ag-0.5Cu-0.05Ni Solder Shinichi Terashima 1, Keiko Takahama 2, Masako Nozaki 2 and Masamoto Tanaka 1 1 Advanced Technology Research Laboratories, Nippon Steel Corp., Futtsu, , Japan 2 Nippon Steel Techno Research Corp., Futtsu, , Japan The formation of fine Sn grains in a Sn-1.2 mass%ag-0.5 mass%cu-0.05 mass%ni solder due to thermal strain was investigated from the viewpoint of recrystallization. After thermal fatigue, small general grains recrystallized at the strain concentrated location in Sn-1.2Ag-0.5Cu- 0.05Ni. Through isothermal annealing, however, grains, which had near h110i orientation at a chip-substrate direction before isothermal annealing, coarsened preferentially. Hence, not isothermal annealing but thermal strain was a driving force for recrystallization. Both grain growth after recrystallization and coarsening of recrystallized grains in Sn-1.2Ag-0.5Cu-0.05Ni were slower than those in Sn-1.2 mass%ag- 0.5 mass%cu, which suppressed crack initiation and increased fatigue life of Sn-1.2Ag-0.5Cu-0.05Ni. (Received January 13, 2004; Accepted March 15, 2004) Keywords: soldering, lead-free, tin, silver, nickel, flip chip, thermal fatigue, microstructure, recrystallization, electron backscattering diffraction pattern (EBSP), microjoining 1. Introduction Instead of conventional eutectic Sn-Pb solders, the use of lead-free solders increases for electronic devices as an environmental friendly technology. 1) Many Sn-based alloys have been proposed as candidates of lead-free solders, and Sn-xAg-Cu solders (x: from 3 to 4 in mass%) are becoming major lead-free solder compositions because they have good reliability and solderability. 2 4) The melting points of SnxAg-Cu solders (x: from 3 to 4 in mass%) are still higher than that of eutectic Sn-Pb solder but lower than those for Sn-xAg- Cu solders (x: under 3 or over 4 in mass%) because they have near eutectic compositions. A flip chip interconnect, which consists of a Si chip and a substrate joined by solders, is used widely as one of the major high density packaging techniques. However, when electronic devices containing flip chips are in operation, temperature cycles generate thermomechanical fatigue reacting with a substrate and a chip because they have different coefficients of thermal expansion, and the fatigue degrades long term reliability of the interconnect. Therefore, the thermal fatigue issue is important for flip chip technologies. In order to discuss the thermal fatigue issue, it is important to clarify the relationship between crack propagation through the interconnects and the microstructural changes of interconnecting materials. In the literature, many findings have reported about microstructural changes in Sn-based alloys. For instance, Mei and Morris 5) and Seyyedi 6) have reported that a eutectic Sn-Bi alloy recrystallizes after a fatigue test, and Seyyedi 6) and Miao and Duh 7) have pointed out that grains in a eutectic Sn-Bi alloy coarsen by annealing. A eutectic Sn-Pb alloy shows similar microstructural change as the eutectic Sn-Bi alloy. Namely, Raman and Reiley have observed dynamic recrystallization behaviour for a Pb- 2 mass%sn alloy due to reverse bending low cycle fatigue test. 8) Furthermore, Kim et al. have reported the discontinuous precipitation (DP) behaviour in a water quenched 95 mass%pb-5 mass%sn alloy, and have pointed out that recrystallization which is resulted from strain energy occurs during DP reactions at the location of original grain boundaries. 9) On the other hand, a eutectic Sn-In alloy is different from these two alloys. Freer and Morris 10) and Seyyedi 6) have pointed out that the eutectic Sn-In alloy does not recrystallize nor coarsen after creep and themomechanical fatigue tests, respectively. Hence, microstructural changing behavior is different even in the Sn-based alloy system. There are few reports concerning microstructural change due to fatigue test in Sn and Sn-Ag alloy system. McLean and Farmer 11) and Kanchanomai et al. 12) have observed subgrain formation in pure Sn after a creep test and in a eutectic Sn-Ag alloy after an isothermal fatigue test, respectively. Kanchanomai et al. have argued that the possible reasons of the subgrain formation are recrystallization and recovery of dislocation structure. 12) They have inferred that observed subgrains are induced by polligonization due to recovery only because recrystallization is generally favored in materials with low stacking fault energy and this energy of Sn is high. 12) Moreover, Marshall and Walter have insisted that a Sn-5 mass%ag alloy does not recrystallize nor coarsen through creep test. 13) Hence, it has been believed that no recrystallization nor coarsening occurs in Sn-Ag alloys. However, it has been reported recently that the thermal fatigue properties of Sn-xAg-0.5Cu (x ¼ 1, 2, 3 and 4 in mass%) flip chip interconnects are strongly correlated to coarsening of Sn grains, 14) and solder joints with higher silver content (x ¼ 3 and 4) have better fatigue resistance than those with lower one (x ¼ 1 and 2) due to suppression of Sn grain coarsening. 14) Moreover, it has been reported that Sn- 1.2 mass%ag-0.5 mass%cu-0.05 mass%ni joints have longer thermal fatigue life than Sn-1.2 mass%ag-0.5 mass%cu alloy even though its silver content is low such as 1.2 mass% because of the suppression of Sn grain coarsening. 15,16) These microstuructural changes 14 16) are different from those observed by Kanchanomai et al., 12) and Marshall and Walter. 13) In their papers, general grain formation has not been observed, 12) an effect of strain energy on subgrain formation has not been discussed, 12) or grain size of Sn has been determined only from a scanning electron microscope

2 1384 S. Terashima, K. Takahama, M. Nozaki and M. Tanaka Table 1 The chemical compositions of Sn-1.2Ag-0.5Cu-xNi (x ¼ 0 and 0.05) solders. Unit: mass% Ag Cu Ni Pb Sb Bi Zn Fe Al As Cd Sn < <0.001 <0.001 <0.001 <0.001 < <0.001 Bal <0.001 <0.001 <0.001 <0.001 < <0.001 Bal. (SEM) observation. 13) However, grain sizes or recrystallization behaviour should be carefully discussed with crystallographic orientation of alloys together with SEM analysis. In this study, the formation of fine Sn grains in the Sn- 1.2Ag-0.5Cu-0.05Ni solder due to thermal strain was investigated from the viewpoint of recrystallization behaviour. 2. Experimental 2.1 Preparation of flip chips Sn-1.2Ag-0.5Cu-0.05Ni and Sn-1.2Ag-0.5Cu solders were utilized. The chemical compositions of these solders are shown in Table 1. The initial diameter of these balls was 300 mm. Flip chips consisted of a Si chip (8 mm by 8 mm and 0.4 mm thick) and a FR-4 substrate (50 mm by 50 mm and 0.7 mm thick) using the solder balls mentioned above. In a flip chip packaging process, a rosin mildly activated (RMA) type flux was utilized. Ball arrangement of the flip chip interconnects was 240-pin with 450 mm pitch as shown in Fig. 1. Electroless Ni-P (thickness: 3 mm) and electro Au (0.1 mm) were plated on the Cu electrodes of the substrates (Cu/Ni/Au). Surface of Al electrode (1 mm thick) on the Si chip was finished by 40 nm thick Cr, 500 nm thick Ni and 200 nm thick Au (Al/Cr/Ni/Au). All flip chip interconnects were prepared by a reflow treatment in a N 2 atmosphere with a peak temperature of 518 K. A temperature profile of the reflow treatment is shown in Fig Thermal fatigue test The specimens were thermally cycled in an air atmosphere between 233 K and 398 K at a frequency of 2780 s per cycle up to 600 cycles. A detailed profile of the thermal cycles is shown in Fig. 3. For comparison, some of the specimens were annealed Solder balls (0.45mm pitch) Temperature, T / K Temperature, T / K 600 Peak Temp.: 518K Time, t / s Fig. 2 Temperature profile of the reflow treatment s per cycle K Time, t / s Fig. 3 Detailed profile of the thermal cycles. 8 Si chip Substrate Unit: mm Fig. 1 Schematic illustrations of the ball arrangement of flip chip interconnects employed in thermal fatigue test. isothermally instead of thermal fatigue test in order to heat specimens without thermal strain. As seen from Fig. 3, thermally fatigued specimens were kept for 1.11 ks between 388 K and the peak temperature (398 K) per cycle. After 600 thermal cycling, therefore, the specimens were annealed for 666 ks at between 388 K and 398 K. Hence, isothermal annealing was performed for 666 ks at 398 K in the present study for comparison with the specimens after 600 thermal cycling. 2.3 Microstructural observation Solder joints were polished by three kinds of SiC papers

3 Recrystallization of Sn Grains due to Thermal Strain in Sn-1.2Ag-0.5Cu-0.05Ni Solder 1385 points were different more than 15 degrees and under 5 degrees, respectively. 3. Results and Discussion Fig. 4 Shape of a Sn-1.2Ag-0.5Cu-0.05Ni solder bump just after the reflow treatment. (number: 120, 600 and 1200) and by three grades of diamonds pastes (6, 3 and 1 mm). Etching was carried out in hydrochloric vapor at room temperature for several seconds. Microstructures of the specimens were observed by an optical-microscope and SEM. Figure 4 shows a shape of a Sn-1.2Ag-0.5Cu-0.05Ni solder bump just after the reflow treatment observed by a optical microscope. The crystallographic orientation of alloys was identified by the electron backscattering pattern (EBSP) with an electron beam of about 0.2 mm in diameter. A measuring step was 1.0 mm. It has been reported that microstructure of the Sn- 1.2Ag-0.5Cu-0.05Ni solder joint consists of -Sn matrix together with a number of Ag 3 Sn and (Cu,Ni) 6 Sn 5 intermetallic compounds dispersoids. 15,16) Additionally, sizes of these dispersoids are from several of 10 nm to several mm. 15,16) Considering the beam diameter and the measuring step, the dispersoids were too small for the present measuring conditions. Therefore, orientation image mapping was carried out only for Sn in the present measurement. Figure 5 shows a relationship among a sample position, a normal (ND), a transverse (TD) and a reference direction (RD) in the present EBSP measurement. As shown in Fig. 5, TD was parallel to a chip-substrate direction in the specimen. In the present paper, general and sub grains were determined when orientations of neighbouring measured Fig. 5 Relationship between a sample position and directions. 3.1 Recrystallization behaviour due to thermal strain Figures 6, 7 and 8 show (a, b) inverse pole figures at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu-0.05Ni solder joint before, after 150, and 600 thermal cycling, respectively. As seen from Fig. 6(a), the joint had two large and several small general grains before thermal cycling. These general grains consisted of several subgrains over 50 mm diameter as shown in Figs. 6(d) and (e). As seen from Figs. 6(a) and (b), orientations of Sn grains before thermal cycling formed a band from h001i to h100i directions. After 150 thermal cycling, shown in Fig. 7(c), cracks initiated from the edge of the solder near the solder/chip bonded interface. A number of subgrains under 30 mm in diameter can be seen near this bonded interface (Figs. 7(d) and (e)) together with large original general grains (Fig. 7(c)). Namely, a lot of fine subgrains were observed after 150 thermal cycling, which suggests that thermal strain induced these subgrains. Additionally, several general grains about 10 mm in diameter can be also seen where cracks initiated. From inverse pole figure shown in Fig. 7(a), these general grains had different orientations compared with large original general grains. As shown in Fig. 8, cracks perfectly propagated through the solder after 600 thermal cycling. A number of general grains covered the solder (Fig. 8(c)) and these grains did not have preferred orientations (Figs. 8(a) and (b)). This indicates that a lot of general grains were formed and grew between 150 and 600 thermal cycling. Thermal strain energy can cause recrystallization. It has been reported that thermal strain is concentrated near the solder/chip bonded interface for Sn-1.2Ag-0.5Cu-xNi (x: 0, 0.02, 0.05, 0.10 and 0.20) solder joints. 15,16) This strain concentration behaviour can be also seen in the present results; namely, cracks initiated and propagated near this bonded interface (Figs. 7 and 8). In order to clarify the effect of thermal strain, further investigation is carried out in the following. Figure 9 shows (a, b) an inverse pole figure at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu-0.05Ni solder joint after isothermal annealing for 666 ks at 398 K. As mentioned in 2.2, this isothermal annealing condition was determined from a time length for specimens to be heated between 388 K and 398 K in 600 thermal cycling. As shown in Fig. 9, four general grains can be seen after isothermal annealing, while a lot of general grains covered the solder after thermal fatigue (Fig. 8). Namely, grain size was different between isothermally annealed and thermally fatigued specimens. Inverse pole figure after isothermal annealing shows a shift to near h110i orientation (Figs. 9(a) and (b)), while no preferred orientation can be seen after 600 thermal cycling (Figs. 8(a) and (b)). Lee et al. have pointed out that the preferred growth direction for Sn dendrites is h110i. 17) Hence, it is considered that a heat flux in isothermal annealing was parallel to a chipsubstrate direction (namely, TD), and that a few grains, which had near h110i orientation at TD before isothermal

4 1386 S. Terashima, K. Takahama, M. Nozaki and M. Tanaka Fig. 6 (a, b) Inverse pole figure at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu- 0.05Ni solder joint before thermal cycling. Fig. 7 (a, b) Inverse pole figure at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu- 0.05Ni solder joint after 150 thermal cycling.

5 Recrystallization of Sn Grains due to Thermal Strain in Sn-1.2Ag-0.5Cu-0.05Ni Solder 1387 Fig. 8 (a, b) Inverse pole figure at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu- 0.05Ni solder joint after 600 thermal cycling. Fig. 9 (a, b) Inverse pole figure at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu- 0.05Ni solder joint after isothermal annealing for 666 ks at 398 K. annealing, coarsened preferentially during the annealing. From these results, it is indicated that not isothermal annealing but thermal strain was a driving force for the formation of small subgrains and small general grains at the strain concentrated location shown in Figs. 7 and 8. Namely, observed grain formation due to thermal strain may well be occurred by recrystallization. Recrystallization behaviour in Sn-1.2Ag-0.5Cu-0.05Ni

6 1388 S. Terashima, K. Takahama, M. Nozaki and M. Tanaka thermally fatigued, and a number of subgrains form at the strain concentrated location (Fig. 10(b)). This subgrain formation is the second step for recrystallization. Even though thermal strain energy is used for recrystallization, this energy may still exist and cause crack initiation. After further thermal cycling, the third step occurs, namely, general grains form and grow especially at the strain concentrated location, and cracks propagate owing to thermal strain (Fig. 10(c)). Finally, as the fourth step, cracks perfectly propagate through the solder due to thermal strain, and a number of general grains grow and cover the solder (Fig. 10(d)). Namely, recrystallization behaviour in Sn-1.2Ag-0.5Cu-0.05Ni solder was governed by thermal strain. Fig. 10 Schematic illustrations to explain recrystallization behaviour shown in Sn-1.2Ag-0.5Cu-0.05Ni solder. solder can be explained by four steps shown in schematic illustrations (Fig. 10). The first step is that a solder has a few general grains just after solidification (Fig. 10(a)). When isothermal annealing is carried out for this solder, grains, which have near h110i orientation at TD before isothermal annealing, coarsen preferentially through the annealing (Fig. 10(e)). However, recrystallization occurs if specimens are 3.2 Formation of fine Sn grains due to nickel addition into Sn-1.2Ag-0.5Cu Sn-1.2Ag-0.5Cu-0.05Ni joints have longer thermal fatigue life than Sn-1.2Ag-0.5Cu alloy because crack propagation through solders is suppressed due to fine Sn matrix. 15,16) In the following, an effect of nickel addition into Sn-1.2Ag- 0.5Cu on formation of fine Sn grains has been investigated from the viewpoint of recrystallization. Figures 11, 12 and 13 show (a, b) inverse pole figures at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu solder joint before, after 150, and 600 thermal cycling, respectively. As shown in Fig. 11(c), two large and several small general grains can be seen before thermal cycling, and these general grains consisted of several subgrains over 50 mm in length (Figs. 11(d) and (e)). Figures 11(a) and (b) reveal that Sn grains before thermal cycling formed a band from h001i to h100i Fig. 11 (a, b) Inverse pole figure at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu solder joint before thermal cycling.

7 Recrystallization of Sn Grains due to Thermal Strain in Sn-1.2Ag-0.5Cu-0.05Ni Solder 1389 Fig. 12 (a, b) Inverse pole figure at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu solder joint after 150 thermal cycling. orientations. After 150 thermal cycling, cracks occurred from the edge of the solder near the solder/chip bonded interface (Fig. 12(c)). General grains about 30 mm in length formed at the crack initiation location (Fig. 12(c)), and some grains shifted from the initial orientations (Figs. 12(a) and (b)). Although it is unclear in Fig. 13, it is confirmed by SEM analysis that cracks propagated through the solder after 600 thermal cycling. One large general grain almost covered the solder and several small general grains from 10 to 50 mm in length formed near strain concentrated location as seen from Fig. 13(c). Sn grains had a slight preference of h110i orientation (Figs. 13(a) and (b)). From above results, the relationship between grain refinement of Sn and an effect of nickel addition into Sn-1.2Ag- 0.5Cu solder can be explained from the comparison of recrystallization behaviour in Sn-1.2Ag-0.5Cu-0.05Ni and Sn-1.2Ag-0.5Cu solder joints. Just after solidification, both Sn-1.2Ag-0.5Cu-0.05Ni and Sn-1.2Ag-0.5Cu solder joints had several general grains. However, after 150 thermal cycling, Sn-1.2Ag-0.5Cu-0.05Ni had smaller general grains about 10 mm in length at the crack initiation location (Fig. 7(c)), while these grains in Sn-1.2Ag-0.5Cu were larger such as 30 mm in length (Fig. 12(c)). Additionally, Sn-1.2Ag- 0.5Cu-0.05Ni had smaller general grains after 600 thermal cycling (Fig. 8(c)), whereas one large general grain almost covered the solder in Sn-1.2Ag-0.5Cu (Fig. 13(c)). Namely, not only grain growth after recrystallization but also coarsening of recrystallized grains was faster in Sn-1.2Ag- 0.5Cu than in Sn-1.2Ag-0.5Cu-0.05Ni. Because -Sn has BCT structure, only six different families of slip systems have been reported in a -Sn single crystal. 18) The number of different families of slip systems is smaller in -Sn than in other majour crystal structures such as BCC or FCC, which suggests that polycrystalline -Sn with coarsened grains are more difficult for plastic deformation than that with fine grains. In the present alloy system, fine Sn grains can be resistance for transgranular fracture in Sn-based alloys 14 16) because grain boundaries can be resistance for transgranular crack propagation and because fine Sn grains assist plastic deformation. From above discussion, it is considered that because grain growth after recrystallization was faster in Sn-1.2Ag-0.5Cu than in Sn-1.2Ag-0.5Cu- 0.05Ni, crack initiation was also faster in Sn-1.2Ag-0.5Cu, and therefore, cracks propagated easily and thermal fatigue life became short. However, when nickel of a small amount was added into the Sn-1.2Ag-0.5Cu solder, grain growth of Sn after recrystallization was suppressed and cracks propagated slowly, which resulted in longer fatigue life. Namely, increase in fatigue life due to addition of nickel into the Sn- 1.2Ag-0.5Cu solder was resulted from suppression of Sn grain growth after recrystallization due to thermal strain.

8 1390 S. Terashima, K. Takahama, M. Nozaki and M. Tanaka Fig. 13 (a, b) Inverse pole figure at TD and mapping of grains with boundaries over (c) 15, (d) 5 and (e) 2 degrees for Sn-1.2Ag-0.5Cu solder joint after 600 thermal cycling. 4. Conclusions In this study, the formation of fine Sn grains in the Sn- 1.2Ag-0.5Cu-0.05Ni solder was investigated from the viewpoint of recrystallization behaviour induced by thermal strain. (1) The Sn-1.2Ag-0.5Cu-0.05Ni solder had several general grains just after solidification. (2) Several small general grains formed at the strain concentrated location after thermal fatigue due to recrystallization for Sn-1.2Ag-0.5Cu-0.05Ni. Additionally, cracks initiated owing to thermal strain. (3) Finally, general grains grew and covered the Sn-1.2Ag- 0.5Cu-0.05Ni solder, and cracks perfectly propagated the solder due to thermal strain. (4) Grains, which had near h110i orientation at TD before isothermal annealing, coarsened preferentially through the annealing of Sn-1.2Ag-0.5Cu-0.05Ni. (5) Not only grain growth after recrystallization but also coarsening of recrystallized grains was faster in Sn- 1.2Ag-0.5Cu than in Sn-1.2Ag-0.5Cu-0.05Ni. Acknowledgements Thanks are due to Dr. Yoshiharu Kariya (National Institute for Materials Science, Japan) for valuable discussions. Authors also wish to thank Mr. Chikara Abe (Nippon Steel Techno-Research Corporation, Japan) for sample preparation in EBSP measurement. REFERENCES 1) D. Napp: SAMPE Journal 32 (1996) ) C. M. Miller, I. E. Anderson and J. F. Smith: J. Electron. Mater. 23 (1994) ) D. R. Frear: JOM 48 (1996) ) H. Mavoori, J. Chin, S. Vaynman, B. Moran, L. Keer and M. Fine: J. Electron. Mater. 26 (1997) ) Z. Mei and J. W. Morris: J. Electron. Mater. 21 (1992) ) J. Seyyedi: J. Electron. Packag. 115 (1993) ) H. W. Miao and J. D. Duh: Mater. Chem. Phys. 71 (2001) ) V. Raman and T. C. Reiley: Scri. Metall. 20 (1986) ) J. S. Kim, M. S. Kang and S. B. Lee: Scri. Mater. 38 (1998) ) J. L. Freer and J. W. Morris: J. Electron. Mater. 21 (1992) ) D. McLean and M. H. Farmer: J. Inst. Metals 85 ( ) ) C. Kanchanomai, Y. Miyashita, Y. Mutoh and S. L. Mannan: Mater. Sci. Eng. A 345 (2003) ) J. L. Marshall and S. R. Walter: J. Hybrid Microelectron. 10 (1987) ) S. Terashima, Y. Kariya, T. Hosoi and M. Tanaka: J. Electron. Mater. 32 (2003) ) S. Terashima, Y. Kariya and M. Tanaka: Mater. Trans. 45 (2004) ) S. Terashima and M. Tanaka: Mater. Trans. 45 (2004) ) D. V. Lee, K. H. Kim, Y. G. Lee and C. H. Choi: Mater. Chem. Phys. 47 (1997) ) S. N. G. Chu and J. C. M. Li: Mater. Sci. Eng. 39 (1979) 1 10.