Reliability of Ag Nanoporous Bonding Joint for High Temperature Die Attach under Temperature Cycling

Min-Su Kim1,2,*, Kaori Matsunaga2, Yong-Ho Ko3, Chang-Woo Lee3 and Hiroshi Nishikawa1

1Joining and Welding Research Institute, Osaka University, Ibaraki 567–0047, Japan2Graduate School of Engineering, Osaka University, Suita 565–0871, Japan3Micro-Joining Center, Korea Institute of Industrial Technology, 156 Gaetbeol-ro, Incheon 21999, Republic of Korea

The long-term reliability of a Ag nanoporous bonding joint for high temperature die attach under temperature cycling from −55°C to 150°C was investigated. A Ag nanoporous sheet was adopted as a bonding layer for the die attach of a Si chip on an active metal brazed Cu Si3N4 substrate. The initial joint strength was similar to that with Pb-5Sn die attach. There was no signi�cant change in the joint strength after tem-perature cycling up to 1500 cycles. It was possible to con�rm that the shear strength of the Ag nanoporous bonding joint had good stability under temperature cycling.　[doi:10.2320/matertrans.MD201518]

(Received April 4, 2016; Accepted April 19, 2016; Published June 25, 2016)

Keywords:　 silver nanoporous, die attach, high temperature, temperature cycling, fracture

1.　 Introduction

The eutectic Pb-Sn solder used for general electronic de-vices has been replaced with Sn-based solders, especially eu-tectic Sn-Ag-Cu1), in response to the environmental regula-tion on “the restriction of the use of certain hazardous sub-stances in electrical and electronic equipment”, referred to as RoHS2). However, solder alloys have not yet been established as substitutes for high Pb-containing solders such as Pb-5Sn and Pb-10Sn. High temperature interconnection methods have recently received attention with the increasing demand for automotive electronics or power electronics. At present, the high Pb-containing solders have been designated as ex-emptions from recasting version of the RoHS. The validity period for high Pb-containing solders (Pb content > 85 mass%) is 5 years from July 21, 2011. Thus, the exemp-tion of high Pb-containing solder will expire or renew in 20163). Therefore, the development of substitution joining techniques for high temperature applications will be neces-sary in the near future.

As an effort to eliminate the Pb containing solders in high temperature electronics, numerous alternative Pb-free bond-ing techniques that could be used for high temperature die attach have been reported, including high melting tempera-ture solders, such as Au-4), Bi-5), and Zn-based6) solders; tran-sient liquid phase bonding7), and Ag sintering methods using a nanoparticle8–10) or Ag nanoporous sheet11,12). Among these, in our previous research, we successfully demonstrated a novel die attach method using a Ag nanoporous sheet, called Ag nanoporous bonding (NPB), without the aid of the any organic substance. The Ag NPB joint layer exhibited a porous structure, which is expected to relieve the thermo-mechanical stress under a temperature swing. It is a potential candidate for a high temperature die attach technique. However, its long-term reliability under severe environments was not esti-mated11,12). Temperature cycling tests are generally per-formed to examine the durability of electronic device joints under the thermal stress induced by a coef�cient of thermal

expansion (CTE) mismatch. Thus, it was necessary to inves-tigate the long-term reliability of the previously suggested Ag NPB joint under temperature cycling.

Extensive temperature cycling tests on Sn-based solder joints have been performed under various temperature swing conditions, in order to determine life time of the solder joint and its failure mechanism13,14). In the case of a Ag nanoparti-cle sintered joint, Bai et al. reported the reliability of Ag sin-tered joints with SiC on two direct bond copper (DBC) sub-strates (with Al2O3 and AlN) under temperature cycling from 50°C to 250°C (temperature variation, ΔT  =  200°C). They also performed a comparative study on the effect of the sur-face �nish (Ag and Au). This report showed that the AlN DBC had good thermal stability compared with the Al2O3 DBC, because the AlN exhibited a lower CTE than the Al2O3. Speci�cally, the shear strength was maintained up to 2000 cycles and gradually decreased with an increase in the num-ber of cycles. They concluded that Ag sintered die attach with Ag- and Au-coated DBC substrates could withstand up to 4000 and 6000 cycles, respectively, under the assumption that the failure criterion was a 50% drop in the shear strength8).

In this study, a reliability test was performed on a Ag NPB joint for a bare Si chip on an active metal brazed (AMB) Cu Si3N4 substrate under temperature cycling from −55°C to 150°C up to 1500 cycles. The long-term reliability of the Ag NPB joint was estimated in terms of the shear strength change after temperature cycling. The fracture mode change of the Ag NPB joint was also observed.

2.　 Experimental Procedure

2.1　 Ag nanoporous bonding processA Ag nanoporous sheet was fabricated using a dealloying

method in a 2M hydrochloric acid solution from an Al-20Ag precursor. The dealloying treatment was conducted for 3 h for 75°C. Detailed information on the microstructure and phase constitution of the Al-Ag precursor and dealloyed Ag nanoporous sheet was given in the previous report. The thick-ness and ligament size of the Ag nanoporous sheet (Fig. 1(a)) were approximately 90–100 μm and 110 nm, respectively11).* Corresponding author, E-mail: mskim927@gmail.com

Materials Transactions, Vol. 57, No. 7 (2016) pp. 1192 to 1196 ©2016 The Japan Institute of Metals and Materials

A schematic illustration of the Ag NPB process with the Si chip and AMB Si3N4 substrate is shown in Fig. 1(b)–(d). A bare Si chip (3 mm ×  3 mm ×  0.4 mm) with the backside met-allization of Ti/Ni/Au, and a Si3N4 substrate (10 mm  ×  10 mm ×  0.635 mm) with double-side AMB Cu electrodes (8 mm ×  8 mm ×  0.3 mm) purchased from Denka were used for the temperature cycling test. The Cu surface of the AMB substrate was �nished using electroless Ni/immersion Au (ENIG). The thicknesses of the Ti, Ni(P), and Au in the back-side metallization of the Si chip were 100 nm, 2 µm, and 50 nm, respectively, and the thicknesses of the Au and Ni(P) in the surface �nish of the AMB Cu Si3N4 substrate were 50 nm and 2 µm, respectively. Ti, Ni(P), and Au were adopted as an adhesion layer, a diffusion barrier, and an oxidation pre-vention layer, respectively. Before the bonding process, the Si chip and Si3N4 substrate were rinsed with acetone in order to eliminate any surface contamination.

The Ag nanoporous sheet was inserted between the Si chip and AMB Si3N4 substrate. A graphite rubber sheet was placed on the top of the joint to prevent any damage to the Si chip from the pressure applied during the bonding process. The Ag NPB process was performed by a thermo-compression bond-ing system with double-side graphite heating plates at 300°C for 30 min. The applied pressure was 20 MPa, which was controlled using an air compressor.

2.2　 Temperature cycling testThe temperature cycle test (up to 1500 cycles) was carried

out in a temperature test chamber (VCS 7027–15, Vötsch) in accordance with the joint electron device engineering council (JEDEC) standard for temperature cycling (JESD22-A104E)15). The nominal minimum soaking temperature (Tmin) and maximum soaking temperature (Tmax) were −55°C and 150°C (ΔT =  205°C), respectively. The nominal soaking times at Tmax and Tmin were 10 min each, and the heating and cooling rates were 15°C/min. The temperature pro�le (nomi-nal and measured temperatures) during the temperature cy-cling is shown in Fig. 2. The nominal and measured tempera-ture cycling conditions are given in Table 1. The measured temperature variation was well within the tolerance tempera-tures (+10°C for Tmax and −10°C for Tmin).

2.3　 Characterization methodsIn order to examine the occurrence of some defects (voids

or cracks) at the Ag NPB joint as a result of thermal stress, a nondestructive inspection of the die attach was carried out using a scanning acoustic tomograph (SAT, FS300III, Hita-chi) with a high-frequency transducer (frequency: 200 MHz, focal distance: 6.9 mm) before and after the temperature cy-cling test. The Si surface re�ection image in the SAT inspec-tion could provide information on any damage or crack of in the Si and tiny surface contaminations on the Si surface.

The shear strength of the Ag NPB joint was measured us-ing a die shear tester (STR-1000, Rhesca) with a shear rate of 1 mm/min and a shear height of 80 μm from the surface. The fracture surface was examined using an electron probe micro-analyzer (EPMA, JXA-8530F, JEOL), a scanning electron microscope (SEM, SU-70, Hitachi) and a focused ion beam (FIB, JIB-4500, JEOL).

3.　 Results and Discussion

3.1　 Non-destructive inspection of Ag NPB jointThere are two possible causes of crack propagation in a Si

chip. One is the excessive pressure used during the bonding process. A non-uniform or an excessive compression stress can cause a crack in the Si because of the fragile nature of a Si wafer. Another cause of crack propagation in a Si chip is the thermal stress experienced during a temperature cycling test16). During the temperature swing, a large stress is gener-ated at the joint, as a result of the CTE mismatch between the Si wafer and bonding material/substrate. There was no evi-dence of cracking in the Si chip, even after temperature cy-

Fig. 2　A temperature pro�le of temperature cycling test from −55°C (Tmin) to 150°C (Tmax) according to JESD-A104E.

Table 1　Nominal and measured temperature cycling conditions.

Parameters Nominal values Measured values

Tmax 150°C 154.2°CTmin −55°C −57.5°C

Heating rate 15°C/min 14.28 ±  0.11°C/min

Cooling rate −15°C/min −15.36 ±  0.19°C/min

Soak time at Tmax 10 min 8 min 26 s ±  14 s

Soak time at Tmin 10 min 8 min 2 s ±  10 s

Fig. 1　(a) Microstructure of Ag nanoporous sheet11) (reprinted with permis-sion, copyright 2014, Elsevier), (b) Ag nanoporous bonding process using thermo-compression bonding system, and (c) schematic diagram and (d) appearance of Si die attach using Ag nanoporous sheet on Si3N4 active metal brazed (AMB) Cu substrate.

1193Reliability of Ag Nanoporous Bonding Joint for High Temperature Die Attach under Temperature Cycling

cling up to 1500 cycles.Meanwhile, the bonding layer re�ection image from the

SAT inspection could provide information on the void or crack propagation after temperature cycling, as shown in Fig. 3. The bonding layer re�ections showed somewhat dif-ferent contrasts, regardless of the number of temperature cy-cles. In the case of the as-bonded condition, the white area observed seemed to be due to the porous part of the Ag layer, as previously reported11). Before 1000 cycles, there was no clear evidence of Ag layer delamination. After 1000 cycles, the delamination of the bonding layer, which was indicated as white arrows in Fig. 3(e) and Fig. 3(f), was partially observed at corner side of the joint. Mei et al.9) used a �nite element method and reported that the maximum thermal stress of a Ag sintered joint induced by temperature cycling was mainly generated at the corner of the joint. Navarro et al.16) also showed that the maximum normal stress was generated at the corner of the bonding interface; and it sharply decreased to-ward the center of the joint at 275°C. It seems reasonable that the fracture at the Ag porous bonding layer is initiated at the corner of joint. Nevertheless, the delamination of Ag layer was not signi�cant up to 1500 cycles.

3.2　 Shear strength of Ag NPB and fracture mode after temperature cycling

Figure 4 shows the change in the shear strength of the Ag NPB joint after the temperature cycling test. The shear strength of the as-bonded Ag NPB joint is approximately

22 MPa. This is almost the same as the shear strength of Ag NPB with an ENIG �nished Cu disk11), and that of a Pb-5Sn die attach5). There was no signi�cant change in the average value the shear test results up to 1500 cycles compared with the as-bonded results. Furthermore, there was no serious frac-ture propagation in the Ag layer. The combination of the SAT inspection (Fig. 3) and shear test results (Fig. 4) shows that the Ag NPB joint can withstand up to 1500 cycles under tem-perature cycling from −55°C to 150°C.

The individual shear test results are also presented using dots and classi�ed according to the fracture mode in Fig. 4. The fracture mode could be classi�ed into �ve types, includ-ing Ag layer dominant (indicated as open squares), metalliza-tion dominant (open circles), Ag/Si mixed (open triangles), Si/metallization mixed (open diamonds), Ag/Si/metalliza-tion mixed (open stars), as shown in Fig. 4. The fracture modes of the as-bonded Ag NPB joint exhibit almost Ag layer dominant fracture. Typical fracture surface (substrate side) images of Ag NPB joint after temperature cycling are shown in Fig. 5, where the samples in the top row show a high shear strength and the samples in bottom row show a low shear strength. The fracture mode was determined by the area frac-tion of each fracture location. The area fraction criterion used for classi�cation was 5%. For example, the area fractions of the Ag layer fracture, Si fracture, and metallization fracture of sample 1 were 5.71%, 21.02% and 73.27%, respectively. Thus, it can be classi�ed as a Ag/Si/metallization mixed fracture. In the case of sample 2, the area fractions of the Si

Fig. 3　Non-destructive inspection results for Ag nanoporous bonding joint: (a) as-bonded, and after temperature cycling for (b) 100 cycles, (c) 250 cycles, (d) 500 cycles, (e) 1000 cycles, and (f) 1500 cycles. The white dotted arrows in the inset indicate the delamination at the Ag bonding layer.

Fig. 4　Variation in the shear strength after temperature cycling test. The number marks in the inset correspond with those in the fracture surface image in Fig. 5.

Fig. 5　Scanning electron microscope (SEM) secondary electron (SE) imag-es showing morphological changes in fracture surface (substrate side) af-ter temperature cycling for (a) 100 cycles, (b) 250 cycles, (c) 500 cycles, (d) 1000 cycles, and (e) 1500 cycles. The regions marked with A, S, and M indicate the Ag layer fracture, Si fracture, and metallization delamina-tion, respectively.

1194 M.-S. Kim, K. Matsunaga, Y.-H. Ko, C.-W. Lee and H. Nishikawa

fracture and metallization fracture were 2.85% and 97.15%, respectively. Thus, it can be classi�ed as a metallization dom-inant fracture.

The samples that demonstrate a Ag layer fracture, Ag/Si mixed fracture, or Ag/Si/metallization mixed fracture tended to show a higher shear strength compared with the average value. In addition, the samples that demonstrated a metalliza-tion layer delamination or Si/metallization mixed fracture tended to show a lower shear strength compared with the av-erage value.

3.3　 Elemental distribution in the metallization fractured region

Figures 6 and 7 show the elemental distribution (Ag, Ag, Ni, Ti, Si, and O) of the fracture surface and cross-sectional image underneath the fracture surface after temperature cy-cling for 1000 cycles (Fig. 6 is the same as sample 8 in Fig. 5) and 1500 cycles (Fig. 7 is the same as sample 10 in Fig. 5). High Ni and Ti signals and very weak Si and O signals were detected in the metallization fractured region on the substrate side from the EPMA results (Fig. 6(c) and Fig. 7(c)). In addi-tion, a high Si signal and very weak Ti signal were detected on the Si side of the metallization fractured region, as shown in Fig. 8. The EPMA results for the both fracture surfaces of the substrate side and Si side (Fig. 7 and Fig. 8) con�rmed that a metallization fracture occurred along the Si/Ti inter-face. Unfortunately, it was unclear whether the Si backside metallization was degraded during the temperature cycling. Further study will be needed on the interfacial reaction be-tween Si chip and backside metallization after temperature cycling, in order to reveal the cause of the metallization frac-ture and its effect on the shear strength of Ag NPB joint.

The stability of the Ag layer after temperature cycling could be con�rmed from the cross-sectional images of the

edge part of the Ag NPB joint, as shown in Fig. 6(e) and Fig. 7(e). The Ag layer exhibited a porous structure, as report-ed in a previous study11). There was no crack propagation in the Ag layer after 1500 cycles. The Ag NPB layer seemed to show good stability under temperature cycling, because of the ductile property of Ag and the low elastic modulus due to the porous structure11).

3.4　 Microstructure of Ag bonding layer and fracture morphology of Ag fractured region

Figure 9 shows FIB milled images on the Ag bonding lay-ers for as-bonded and after temperature cycling for 1000 and 1500 cycles. The Ag bonding layer was gradually coarsened with the increase of number of cycling. The Ag bonding layer seemed to be gradually coarsened with increase of the num-ber of temperature cycling. This coarsening of Ag layer could lead to increase of mechanical strength of Ag layer. However, a variation of shears strength of Ag nanoporous bonding joint during temperature cycling was not signi�cant. Thus, it can

Fig. 7　SEM SE images showing (a) top view and (b) tilted view of fracture surface (substrate side) after 1500 cycles. (c) X-ray elemental mapping results for Ag, Au, Ni, Ti, Si, and O; and cross-sectional FIB SE image of (d) edge part of fracture surface and (e) magni�ed image of joint layer.

Fig. 8　(a) SEM SE images of fracture surface (Si side) and (b) X-ray ele-mental mapping results. The samples of Fig. 7 and Fig. 8 are from the same joint, and they exhibit a mirror plane relationship.

Fig. 6　SEM SE images showing (a) top view and (b) tilted view of fracture surface (substrate side) after 1000 cycles. (c) X-ray elemental mapping results for Ag, Au, Ni, Ti, Si, and O; and cross-sectional focused ion beam (FIB) SE image of (d) edge part of fracture surface and (e) magni�ed image of joint layer.

1195Reliability of Ag Nanoporous Bonding Joint for High Temperature Die Attach under Temperature Cycling

be concluded that this Ag layer coarsening is not effective in our results.

The features of the fractured Ag porous layers found in the by shear test and temperature cycling after 1500 cycles were drastically different, as shown in Fig. 10. These fracture mor-phologies were observed near the corner of the Ag NPB joint (substrate side). The fracture morphology of the Ag layer found in the shear test, namely no crack propagation during the temperature cycling, showed clear ductile deformation (necking) along the shear direction. Meanwhile, the metalli-zation fracture region after 1500 cycles of temperature cy-

cling showed a cleavage fracture, which is a typical feature of a brittle fracture, as shown in Fig. 11. It seems that the brittle fracture of the metallization layer led to the lower joint strength of the Ag NPB compared with the Ag-layer-domi-nant fracture cases.

4.　 Conclusion

The long-term reliability of a Ag NPB joint under tempera-ture cycling from −55°C to 150°C was investigated. The change in the joint strength of the Ag NPB after temperature cycling was not signi�cant when examining the average val-ues. The Ag NPB layer also showed good stability with limit-ed crack propagation along the Ag layer up to 1500 cycles. The fracture occurrence at the Si backside metallization, es-pecially at the Si/Ti interface, showed a tendency to decrease the shear strength; however, it is uncertain whether a degrada-tion of the Si/Ti interface occurred as a result of the tempera-ture cycling.

Acknowledgment

This research was supported by the Japan Society for the Promotion of Science (JSPS) Grant-in-Aid for Scienti�c Re-search (Grant no. 25289241).

REFERENCES

1) K. Zeng and K.N. Tu: Mater. Sci. Eng. R. 38 (2002) 55. 2) Directive 2002/95/EC of the European Parliament and of the Council

of 27 January 2003 on the restriction of the use of certain hazardous substances in electrical and electronic equipment: Of�cial Journal of the European Union 46 (2003) L 37/19.

3) Directive 2011/95/EC of the European Parliament and of the Council of 8 June 2011 on the restriction of the use of certain hazardous sub-stances in electrical and electronic equipment (recast): Of�cial Journal of the European Union, 54 (2011) L 174/88.

4) V. Chidambaram, J. Hald and J. Hattel: J. Alloy. Compd. 490 (2010) 170.

5) Y. Shi, W. Fang, Z. Xia, Y. Lei, F. Guo and X. Li: J. Mater. Sci. Mater. Electron. 21 (2010) 875.

6) A. Haque, B.H. Lim, A.S.M.A. Haseeb and H.H. Masjuki: J. Mater. Sci. Mater. Electron. 23 (2012) 115.

7) J.-B. Lee, H.-Y. Hwang and M.-W. Rhee: J. Electron. Mater. 44 (2014) 435.

8) J. Bai and G. Lu: IEEE Trans. Device Mater. Reliab. 6 (2006) 436. 9) Y.-H. Mei, J.-Y. Lian, X. Chen, G. Chen, X. Li and G.-Q. Lu: IEEE

Trans. Device Mater. Reliab. 14 (2014) 194. 10) S. Sakamoto, S. Nagao and K. Suganuma: J. Mater. Sci. Mater. Elec-

tron. 24 (2013) 2593. 11) M.-S. Kim and H. Nishikawa: Scr. Mater. 92 (2014) 43. 12) M.-S. Kim and H. Nishikawa: Mater. Sci. Eng. A 645 (2015) 264. 13) J. Hokka, T.T. Mattila, H. Xu and M. Paulasto-Kröckel: J. Electron.

Mater. 42 (2013) 1171. 14) J. Hokka, T.T. Mattila, H. Xu and M. Paulasto-Kröckel: J. Electron.

Mater. 42 (2013) 963. 15) JEDEC, JESD22-A104E: Temperature Cycling, 2014. 16) L.A. Navarro, X. Perpiñà, M. Vellvehi, V. Banu and X. Jordà: Micro-

electron. Reliab. 52 (2012) 2314.

Fig. 9　FIB SE images of Ag bonding layer of (a) as-bonded state; and after temperature cycling for (b) 1000 cycles and (c) 1500 cycles.

Fig. 10　SEM SE images of fracture morphology on Ag layer fractured re-gion after temperature cycling for 1500 cycles. The observation position is indicated with the white arrow in the low-magni�cation inset image.

Fig. 11　SEM SE image of fracture morphology on metallization fracture region after temperature cycling for 1500 cycles. The observation position is indicated with the white arrow in the low-magni�cation inset image.

1196 M.-S. Kim, K. Matsunaga, Y.-H. Ko, C.-W. Lee and H. Nishikawa