- Email: [email protected]
Sn–Cu based joint system with low current density
Microelectronics Reliability 55 (2015) 2554–2559
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/mr
Electromigration behavior in Cu/Ni–P/Sn–Cu based joint system with low current density Takuya Kadoguchi a,d,⁎, Keisuke Gotou b,1, Kimihiro Yamanaka c, Shijo Nagao d, Katsuaki Suganuma d a
Power Electronics Development Division, Toyota Motor Corporation, Japan School of Information Science and Engineering, Chukyo University, Japan School of Engineering, Chukyo University, Japan d The Institute of Scientific and Industrial Research, Osaka University, Japan b c
a r t i c l e
i n f o
Article history: Received 30 April 2015 Received in revised form 2 October 2015 Accepted 4 October 2015 Available online 16 October 2015 Keywords: Electromigration Solder Power module Ni–P Sn–Cu
a b s t r a c t Although electromigration in solder joints has great influence on reliability, few study has been reported on the Cu/Ni–P/Sn–Cu based joint system electromigration with realistic current density range lower than 10 kA/cm2. We investigated a Cu/Ni–P/Sn–0.7Cu/Ni–P/Cu joint with current densities of 5.0 and 7.5 kA/cm2 at 423 K. Solder joint breakdown at the cathode side was detected for both stress conditions. Ni–P plating disappeared completely at the cathode side and a Cu–Sn intermetallic compound (IMC) formed at the interface. Cu–P IMC formed on the solder breakdown interface. Ni diffusion in Ni–P plating at the cathode was accelerated and the P-rich layer grew thicker than at the anode side before breaking down under electromigration stress. The P-rich layer reached the Cu electrode resulting in cracking along the interface between solder layer and Cu. Sn was diffused from the Ni3SnP IMC to the P-rich layer cracks and formed Cu3Sn IMC with the Cu electrode. Thus, the electromigration mechanism in an electroless Ni–P plating/Sn–Cu based joint system with low current density was clarified. © 2015 Elsevier Ltd. All rights reserved.
1. Introduction Electromigration in the semiconductor field has been studied extensively in recent decades to understand and to avoid any fatal damage caused by this phenomenon in metallization. Electromigration is the movement of metal atoms in the direction of strong electron flow, which is termed “electron wind”, and results from momentum transfer between conducting electrons and diffusing metal atoms. The mean time to failure caused by electromigration decreases as the temperature or current density increases, and has been expressed well by Black's equation [1–7]. For example, the current density of a fine pitch solder bump between a chip and substrate has been one of the major concerns in the semiconductor field. It is well known that current densities of 10 kA/cm2 may cause serious electromigration. Although power modules with insulated gate bipolar transistors (IGBTs) and diodes do not require joints as small as the flip chip joints'. Power modules for hybrid electric vehicles are usually operated above 200 A. In addition, the service temperature of engine compartments is very high. Thus, there are serious concerns about electromigration failure of power modules for hybrid electric vehicles. The current density of power devices with solder in the ⁎ Corresponding author at: Power Electronics Development Division, Toyota Motor Corporation, Japan. E-mail address: [email protected] (T. Kadoguchi). 1 Present address: IHI Inspection & Instrumentation Corporation, Ltd., Japan.
http://dx.doi.org/10.1016/j.microrel.2015.10.003 0026-2714/© 2015 Elsevier Ltd. All rights reserved.
double-sided cooling power modules varies from 0.24 kA/cm2 to 0.4 kA/cm2. It is believed that the influence of solder electromigration is minimal in the current operating environment [8–12]. Environmentally friendly vehicles must achieve a good fuel efficiency and a high power density simultaneously. To achieve these demands, low energy consumption, high heat dissipation and high thermostability are required in power modules. The current density of solder joints will increase according to the demand for high thermostability SiC power devices and device miniaturization [13,14]. The miniaturization of power modules can be achieved by integrating 6-in-1 and 2-in-1 circuit compositions, which operate a three-phase alternating current motor, into one package [15,16]. Kadoguchi et al. reported that an emitter electrode in the upper arm IGBT usually jointed a collector electrode in the lower arm IGBT for 2-in-1 power modules. The joint is smaller than that of power devices, so the current density increases and solder electromigration occurs easily [16]. Therefore, it is necessary to investigate the mechanism of solder electromigration and to enhance the electromigration resistance for power modules. Several studies reported that joints with under bump metallurgy, such as Ni and Ni/Ti, could enhance the electromigration reliability of Sn–Cu, Sn–Ag and Sn–Pb solders with high current. However, few study exists on solder electromigration with current density lower than 10 kA/cm2, which is a realistic value [17–28]. The Ni–P plating is expected to have a diffusion barrier effect to suppress electromigration [18,29,30]. We therefore studied electromigration in an electroless Ni–P plating/Sn–0.7Cu based joint system with current densities lower than
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559
2555
10 kA/cm2. The diffusion behavior of each element caused by electromigration was mainly identified by comparing the solder joint interfaces at an anode and at a cathode. 2. Experimental procedure Fig. 1(a) and (b) show an overview of the sample and a schematic cross-section of a cathode side respectively. Oxygen free copper (C1020) was used for the electrodes and processed into 1-mm-long 0.4 mm × 0.4 mm cross sections. Soldering pad for the Cu electrode was finished by electroless Ni–P/Au plating. The Ni–P plating thickness varied from 7 to 14 μm. Two electrodes were jointed with a 0.5-mm-diameter solder ball. The solder composition was M725 (Sn–0.7Cu–Ni–P, wt.%) from Senjyu Metal Industry Co., Ltd. Fig. 2 shows the scanning electron microscope (SEM) image of the cross-section of a joint interface. The intermetallic compound (IMC) at the interface was ~ 2.5 μm thick. Energy dispersive X-ray spectroscopy (EDX) analysis showed that the IMC consisted of 33.7 at.% Cu–21.4 at.% Ni–44.9 at.% Sn. Therefore, the IMC composition can be estimated as (Cu,Ni)6Sn5. A P-rich layer was ~1.3 μm thick, and this layer consists of mixture of Ni3P and Ni [31]. The sample composition was confirmed in Fig. 1(b). We applied currents of 8.0 and 12 A to the solder joint to generate current densities of 5.0 and 7.5 kA/cm2 respectively. This test sample was placed on a hot plate, which was controlled so that the joint temperature was maintained at 423 K when a specified constant direct current passed through the joint. The temperature profile of the test sample was determined by thermocouple and by measuring the resistance change in the junction line between the Cu/solder/Cu joints. The change in resistance caused by electromigration was measured by monitoring voltage. The failure criterion was defined as a 20% increase in resistance from the initial value. The failed sample cross-section was polished and studied using a SEM. The electromigration failure mode was investigated by analyzing the elemental composition with EDX. To check the electromigration process phenomenon, the joint was studied after applying a stress of 423 K and 5.0 kA/cm2 for 250, 500, 750, and 2500 h. 3. Results and discussion 3.1. Electromigration stress test Fig. 3 shows the change in resistance with time for the stress condition at 423 K with current densities of 7.5 and 5.0 kA/cm2. The failure
Fig. 1. EM test sample. (a) Overview; (b) Schematic of cathode.
Fig. 2. Cross-section SEM image of the solder joint interface after reflow.
times were 1060 and 2320 h respectively. The failure samples after the electromigration stress tests are shown in Fig. 4. The solder of both failed samples was detached from the Cu electrode at the cathode side. The samples maintained at 423 K without applying current were not failed at the joint interface. Therefore, the failure mode of both samples with applied current was caused by electromigration. Fig. 5 shows a SEM image of the failed sample for a stress condition of 423 K and 7.5 kA/cm2. This sample was failed at the solder of a cathode side as shown in Fig. 5(a). (Cu,Ni)6Sn5 IMC existed entirely in the solder layer. Fig. 5(b) and (c) show enlarged images of the cracking area. The Ni–P plating and P-rich layers, which remained after reflow, disappeared. The Cu electrode reacted with the solder and formed Cu3Sn and (Cu,Ni)6Sn5 IMCs. Breakdown by electromigration occurred in the solder though continuous Cu–P layer was detected by EDX measurements on the solder surface. EDX point analysis on the layer indicated the atomic ratio of Cu and P was about 3:1, suggesting the compound may be Cu3P IMC. Jang et al. reported that electromigration makes P in the Ni–P plating at a cathode side diffused and reacted with the Cu electrode (without Ni–P plating) at an anode resulting in the formation of Cu–P IMC [18]. The position of Cu–P IMC in this sample coincided with that of the P-rich layer. Therefore, it is thought that electromigration makes Cu of the electrode diffuse to the P-rich layer to form Cu–P IMC. (Cu,Ni)6Sn5 IMC which remained after reflow was migrated by electron and failure occurred at the Cu–P IMC interface.
Fig. 3. Change of resistance ratio under electromigration stress.
2556
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559
Fig. 4. Overview of failure samples (a) 423 K, 7.5 kA/cm2, 1060 h; (b) 423 K, 5.0 kA/cm2, 2320 h.
Fig. 5(d) shows a cross-section of a joint interface at the anode side. The P-rich layer was ~1.8 μm thick and grew 1.4 times thicker than the initial one. However, the Ni–P plating remained and the influence of electromigration at the anode side was less than that at the cathode side. 3.2. P-rich layer behavior with electromigration stress To confirm the P-rich layer behavior influenced by electromigration, test sample cross sections were analyzed after applying a stress of 423 K and 5.0 kA/cm2 for 250, 500, 750, and 2500 h. No samples exceeded the current breakdown criterion of 20% resistance increase from the initial value. Fig. 6 shows cross-section SEM images after applying a stress of
423 K and 5.0 kA/cm2. The thickness of the (Cu,Ni)6Sn5 IMC at the cathode side was ~ 15 μm after 250 h. It grew approximately six times thicker than initial value. After 500 h, the (Cu,Ni)6Sn5 IMC became thinner. After 2500 h, the (Cu,Ni)6Sn5 IMC disappeared from the joint interface and Ni3SnP IMC formed over the joint instead. Thus, electromigration caused the decomposition of (Cu,Ni)6Sn5 IMC. The Prich layer at the cathode side grew with time, and all the Ni–P plating changed into the continuous P-rich layer (~8 μm in thickness) resulting in crack formation. Delamination at the cathode interface of the Cu electrode was confirmed. The Ni–P plating layer (P: 9.5 wt.%) was amorphous and compressive stress was applied. The volume shrinkage by transformation from an amorphous (Ni–P) layer to a crystal (Ni3P) one caused cracks and delamination [32,33]. The P-rich layer at the anode side was ~2.0 μm thick after 2500 h. It grew thicker after reflow but the rate at the anode side was slower than at the cathode side. Fig. 7 shows a cross-section SEM image of a joint with annealing only at 423 K for 2500 h. A Ni–P plating layer remained and the interface of the Cu electrode was undetached, although the (Cu,Ni)6Sn5 IMC and P-rich layer grew thicker than initial value. The P-rich layer was ~4.0 μm thick. Electromigration almost doubled the growth rate of the P-rich layer at the cathode side after 2500 h and halved that at the anode side after 2500 h. Fig. 8 shows the comparison in Ni–P thickness with time at 423 K with current density of 5.0 kA/cm2 and without current density. The reaction rate of Ni–P at the cathode side at 423 K with current density was faster than that at 423 K without current density until 2500 h. Therefore, it is believed that electromigration enhances diffusion of Ni from the Ni–P plating to the solder layer. The P-rich layer at the anode side increased slowly than that with only annealing. Thus, the electromigration suppressed the diffusion of Ni in the Ni–P plating at the anode toward the solder. 3.3. Formation of Cu–Sn IMC under P-rich layer As shown in Fig. 5, after current stressing, Sn diffuses into the Cu electrode passing through the Ni–P layer and formed Cu–Sn IMC. However, as shown in Fig. 6, Ni–P plating and P-rich layers still existed on the Cu electrode, and so the solder did not contact the Cu electrode directly. Fig. 9(a) shows a cross-section SEM image of a joint after 2500 h at 423 K with current density of 5.0 kA/cm2. The Ni–P plating changed into a P-rich layer and cracks existed in Fig. 9(b). These cracks opened the diffusion paths of Cu and Sn atoms, and may accelerate Cu migration
Fig. 5. (a) Cross-section SEM image of the solder joint after 1060 h with 7.5 kA/cm2 at 423 K; (b), (c) enlarged images of cathode side; (d) enlarged image of anode side.
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559
2557
Fig. 6. Cross-section SEM images of the solder joint at the cathode and anode with 5.0 kA/cm2 at 423 K.
toward the anode. Sn was detected by EDX measurements inside the cracks of the P-rich layer on Ni3SnP IMC or (Cu,Ni)6Sn5 IMC as shown in Fig. 9(c). Though Sn was transferred from each IMC, the transfer direction of Sn was opposite to the electron flow in this electromigration test. It is thought that Sn is transferred by thermal diffusion. Two types of IMC were confirmed inside cracks of the P-rich layer as shown in Fig. 9(d). One IMC in the Ni3SnP side was estimated to be (Ni,Cu)3Sn4 IMC because its composition was 34.1 at.% Ni–10.5 at.% Cu–55.5 at.% Sn from elemental EDX analysis. The other IMC was estimated to be Cu3Sn IMC because its composition was 75.2 at.% Cu–24.8 at.% Sn from elemental EDX analysis. It is believed that the Sn of the Ni3SnP IMC is transferred by thermal diffusion, reacts with the Cu electrode and forms these IMCs. 3.4. Breakdown mechanism of the cathode side with electromigration stress Fig. 10 shows an estimated breakdown mechanism of the cathode side by analyzing the electromigration failure mode and test samples with electromigration stress. (1) Electromigration diffused Ni in Ni–P plating and accelerated the formation of a P-rich layer. When the P-rich layer reached a Cu electrode, this layer delaminated from the interface and cracks formed in this layer.
Fig. 7. Cross-section SEM image of the solder joint after 2500 h without applying current at 423 K.
(2) Sn diffused from Ni3SnP IMC to these cracks in the P-rich layer and formed Cu3Sn and (Ni,Cu)3Sn4 IMCs from the Cu electrode side. (3) Thermal diffusion provided Sn from the solder side and formed Cu3Sn and (Cu,Ni)6Sn5 IMCs on the Cu electrode interface. P from the P-rich layer and Cu formed the Cu–P IMC. (4) Sn at the cathode side was diffused by electromigration and broke down from the Cu–P IMC interface.
4. Conclusions Electromigration in the joint between the Ni–P plating and Sn–Cu solder has been investigated at 423 K with current densities lower than 10 kA/cm2. (1) The times to failure for the electromigration tests were 1060 and 2320 h with current densities of 7.5 and 5.0 kA/cm2 respectively. The electromigration breakdown mode occurred even with current density lower than 10 kA/cm2. The breakdown surface was the interface between Cu–P IMC at the cathode side and the solder. (2) The P-rich layer at the cathode side grew faster than that at the anode side and than that without applying current. The diffusion direction of Ni was as same as the electron flow. Therefore, electromigration accelerated Ni diffusion at the cathode side. (3) The P-rich layer at the anode side grew slower than that at the cathode side and than that without applying current. The diffusion direction of Ni toward the solder was opposite to that of
Fig. 8. The comparison of Ni–P reaction layer thickness with 5.0 kA/cm2 at 423 K.
2558
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559
Fig. 9. (a) Cross-section SEM image of the solder joint after 2500 h with 5.0 kA/cm2 at 423 K; (b) enlarged image of cathode side; (c),(d) enlarged images of P-rich layer.
the electron flow. Therefore, electromigration suppressed Ni diffusion at the anode side. (4) The P-rich layer reached the Cu electrode, cracks were formed. Sn was diffused from Ni3SnP IMC to these cracks in the P-rich layer and formed Cu–Sn IMCs with the Cu electrode.
Acknowledgments The authors would like to thank students in Chukyo University and colleagues in Toyota Motor Corporation for their helpful discussions. References
Fig. 10. Schematic of electromigration failure mechanism at the cathode.
[1] J.R. Black, Electromigration — a brief survey and some recent results, IEEE Trans. Electron Devices (1969) 338–347. [2] J.R. Black, Physics of electromigration, Annual Proceedings of Reliability Physics 1974, pp. 142–149. [3] H.B. Huntington, Electromigration in Metals, in: A.S. Nowick, J.J. Burton (Eds.), Diffusion in Solids Recent Development, Academic Press, New York 1975, pp. 303–352. [4] I.A. Blech, Electromigration in thin aluminium films on titanium nitride, J. Appl. Phys. 47 (4) (1975) 1203–1208. [5] J.R. Black, Electromigration of Al–Si alloy films, Annual Proceedings of Reliability Physics 1978, pp. 300–307. [6] J.R. Lloyd, K.N. Tu, J. Jaspal, The physics and material science of electromigration and thermomigration in solders, in: K.J. Puttlitz, K.A. Stalter (Eds.), Handbook of Lead Free Solder Technology for Microelectronic Assemblies, Marcel Dekker, New York 2002, pp. 827–850. [7] K.N. Tu, Recent advances on electromigration in very-large-scale-integration of interconnects, J. Appl. Phys. 94 (9) (2003) 5451–5473. [8] T. Matsubara, H. Yaguchi, T. Takaoka, K. Jinno, Development of new hybrid system for compact class vehicle”, Proceedings of JSAE 2009, Yokohama, Japan 2009, pp. 21–24. [9] N. Nozawa, T. Maekawa, E. Yagi, Y. Terao, Development of power control unit for compact class vehicle, Proceedings of 22th ISPSD 2010, Hiroshima, Japan 2010, pp. 43–45. [10] N. Hirano, K. Mamitsu, T. Okumura, Structural development of double-sided cooling power modules, Denso Tech. Rev. 16 (2011) 30–37. [11] Y. Sakamoto, Assembly technologies of double-sided cooling power modules, Denso Tech. Rev. 16 (2011) 46–56. [12] S. Miura, Y. Ookura, Y. Okabe, S. Momota, Development of power devices for power cards, Denso Tech. Rev. 16 (2011) 38–45. [13] K. Hamada, Present status a future prospects for electronics in EVs/HEVs and expectations for wide bandgap semiconductor devices, Mater. Sci. Forum 600–603 (2009) 889–893.
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559 [14] S. Hirose, Power electronics technology for the next generation environmentallyfriendly vehicles, Proceedings of the 24th Symposium Microelectronics, Osaka, Japan, Sep. 2014 2014, pp. 37–40. [15] Kadoguchi T, Suzuki Y, Kaji M, Nakajima K, Miyoshi T, Kawashima T, Okumura T. Semiconductor Module. U.S. Patent 8,810,026, Aug. 19, 2014 [16] Kadoguchi T, Iwasaki S, Kawashima T, Okumura T, Nishihata M. Semiconductor Device and Manufacturing Method Thereof. U.S. Patent 8,884,411, Nov. 11, 2014 [17] L.D. Chen, M.L. Huang, S.M. Zhou, Effect of electromigration on intermetallic compound formation in line-type Cu/Sn/Cu and Cu/Sn/Ni interconnects, Proceedings of 60th ECTC 2010, Las Vegas, NV 2010, pp. 176–181. [18] J.W. Jang, L.N. Ramanathan, D.R. Frear, Electromigration behavior of lead-free solder flip chip bumps on NiP/Cu metallization, J. Appl. Phys. 103 (12) (2008) 123506. [19] K.H. Kuo, J. Lee, C. Stan, F.L. Chien, R. Lee, J. Lau, Electromigration performance of printed Sn0.7Cu bumps with immersion tin surface finishing for flip chip applications, Proceedings of 62th ECTC 2012, Sparks, NV 2012, pp. 698–702. [20] L.N. Ramanathan, T.-Y.T. Lee, J.-W. Jang, S.-H. Chae, P.S. Ho, Current carrying capability of Sn0.7Cu solder bumps in flip chip modules for high power applications, Proceedings of 57th ECTC 2007, Reno, NV 2007, pp. 1456–1461. [21] S. Peng, L. Li, A comparison study of electromigration performance of Pb-free flip chip solder bumps, Proceedings of 59th ECTC 2009, San Diego, CA 2009, pp. 1456–1461. [22] M. Lu, P. Lauro, D.-Y. Shih, R. Polastre, C. Goldsmith, D.W. Henderson, Comparison of electromigration performance for Pb-free solders and surface finishes with Ni UBM, Proceedings of 58th ECTC 2008, Orlando, FL 2008, pp. 360–365. [23] Y.-S. Lai, Y.-T. Chiu, C.-W. Lee, Y.-H. Shao, J. Chen, Electromigration reliability and morphologies of Cu pillar flip-chip solder joints, Proceedings of 58th ECTC 2008, Orlando, FL 2008, pp. 330–335.
2559
[24] S.-H. Chael, J. Im, T. Uehling, S.H. Paul, Effects of UBM thickness, contact trace structure and solder joint scaling on electromigration reliability of Pb-free solder joints, Proceedings of 58th ECTC 2008, Orlando, FL 2008, pp. 354–359. [25] K. Yamanaka, Y. Tsukada, K. Suganuma, Solder electromigration in Cu/In/Cu flip chip joint system, J. Alloys Compd. 437 (2007) 186–190. [26] K. Yamanaka, Y. Tsukada, K. Suganuma, Studies on solder bump electromigration in Cu/Sn–3Ag–0.5Cu/Cu system, Microelectron. Reliab. 47 (2007) 1280–1287. [27] J.K. Dong Wook Kim, J. Lee, M.-J. Lee, S.Y. Pai, S. Chen, F. Kuo, Evaluation of electromigration (EM) life of ENEPIG and CuSOP surface finishes with various solder bump materials, Proceedings of 60th ECTC 2010, Las Vegas, NV 2010, pp. 1841–1845. [28] Y.-S. Lai, J.-M. Song, Electromigration reliability with respect to Cu content in solder joint system, Proceedings of 58th ECTC 2008, Orlando, FL 2008, pp. 1160–1163. [29] C.E. Ho, C.W. Fan, W.Z. Hsieh, Pronounced effects of Ni(P) thickness on the interfacial reaction and high impact resistance of the solder/Au/Pd(P)/Ni(P)Cu reactive system, Surf. Coat. Technol. 259 (2014) 244–251. [30] C.E. Ho, W.Z. Hsieh, T.H. Yang, Depletion and phase transformation of a submicron Ni(P) film in the early stage of soldering reaction between Sn–Ag–Cu and Au/ Pd(P)/Ni(P)/Cu, Electron. Mater. Lett. 11 (1) (2015) 155–163. [31] C.-W. Hwang, K. Suganuma, Interface microstructure between Ni–P alloy plating and Sn–Ag–(Cu) lead-free solders, J. Mater. Res. 18 (11) (2003) 2540–2543. [32] C. Baldwin, T.E. Such, Plating rates and physical properties of electroless nickel/ phosphorus alloy deposits, Trans. Inst. Met. Finish. 46 (1968) 73–80. [33] K. Parker, Effects of heat treatment on the properties of electroless nickel deposits, Plat. Surf. Finish. 68 (12) (1981) 71–77.
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/mr
Electromigration behavior in Cu/Ni–P/Sn–Cu based joint system with low current density Takuya Kadoguchi a,d,⁎, Keisuke Gotou b,1, Kimihiro Yamanaka c, Shijo Nagao d, Katsuaki Suganuma d a
Power Electronics Development Division, Toyota Motor Corporation, Japan School of Information Science and Engineering, Chukyo University, Japan School of Engineering, Chukyo University, Japan d The Institute of Scientific and Industrial Research, Osaka University, Japan b c
a r t i c l e
i n f o
Article history: Received 30 April 2015 Received in revised form 2 October 2015 Accepted 4 October 2015 Available online 16 October 2015 Keywords: Electromigration Solder Power module Ni–P Sn–Cu
a b s t r a c t Although electromigration in solder joints has great influence on reliability, few study has been reported on the Cu/Ni–P/Sn–Cu based joint system electromigration with realistic current density range lower than 10 kA/cm2. We investigated a Cu/Ni–P/Sn–0.7Cu/Ni–P/Cu joint with current densities of 5.0 and 7.5 kA/cm2 at 423 K. Solder joint breakdown at the cathode side was detected for both stress conditions. Ni–P plating disappeared completely at the cathode side and a Cu–Sn intermetallic compound (IMC) formed at the interface. Cu–P IMC formed on the solder breakdown interface. Ni diffusion in Ni–P plating at the cathode was accelerated and the P-rich layer grew thicker than at the anode side before breaking down under electromigration stress. The P-rich layer reached the Cu electrode resulting in cracking along the interface between solder layer and Cu. Sn was diffused from the Ni3SnP IMC to the P-rich layer cracks and formed Cu3Sn IMC with the Cu electrode. Thus, the electromigration mechanism in an electroless Ni–P plating/Sn–Cu based joint system with low current density was clarified. © 2015 Elsevier Ltd. All rights reserved.
1. Introduction Electromigration in the semiconductor field has been studied extensively in recent decades to understand and to avoid any fatal damage caused by this phenomenon in metallization. Electromigration is the movement of metal atoms in the direction of strong electron flow, which is termed “electron wind”, and results from momentum transfer between conducting electrons and diffusing metal atoms. The mean time to failure caused by electromigration decreases as the temperature or current density increases, and has been expressed well by Black's equation [1–7]. For example, the current density of a fine pitch solder bump between a chip and substrate has been one of the major concerns in the semiconductor field. It is well known that current densities of 10 kA/cm2 may cause serious electromigration. Although power modules with insulated gate bipolar transistors (IGBTs) and diodes do not require joints as small as the flip chip joints'. Power modules for hybrid electric vehicles are usually operated above 200 A. In addition, the service temperature of engine compartments is very high. Thus, there are serious concerns about electromigration failure of power modules for hybrid electric vehicles. The current density of power devices with solder in the ⁎ Corresponding author at: Power Electronics Development Division, Toyota Motor Corporation, Japan. E-mail address: [email protected] (T. Kadoguchi). 1 Present address: IHI Inspection & Instrumentation Corporation, Ltd., Japan.
http://dx.doi.org/10.1016/j.microrel.2015.10.003 0026-2714/© 2015 Elsevier Ltd. All rights reserved.
double-sided cooling power modules varies from 0.24 kA/cm2 to 0.4 kA/cm2. It is believed that the influence of solder electromigration is minimal in the current operating environment [8–12]. Environmentally friendly vehicles must achieve a good fuel efficiency and a high power density simultaneously. To achieve these demands, low energy consumption, high heat dissipation and high thermostability are required in power modules. The current density of solder joints will increase according to the demand for high thermostability SiC power devices and device miniaturization [13,14]. The miniaturization of power modules can be achieved by integrating 6-in-1 and 2-in-1 circuit compositions, which operate a three-phase alternating current motor, into one package [15,16]. Kadoguchi et al. reported that an emitter electrode in the upper arm IGBT usually jointed a collector electrode in the lower arm IGBT for 2-in-1 power modules. The joint is smaller than that of power devices, so the current density increases and solder electromigration occurs easily [16]. Therefore, it is necessary to investigate the mechanism of solder electromigration and to enhance the electromigration resistance for power modules. Several studies reported that joints with under bump metallurgy, such as Ni and Ni/Ti, could enhance the electromigration reliability of Sn–Cu, Sn–Ag and Sn–Pb solders with high current. However, few study exists on solder electromigration with current density lower than 10 kA/cm2, which is a realistic value [17–28]. The Ni–P plating is expected to have a diffusion barrier effect to suppress electromigration [18,29,30]. We therefore studied electromigration in an electroless Ni–P plating/Sn–0.7Cu based joint system with current densities lower than
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559
2555
10 kA/cm2. The diffusion behavior of each element caused by electromigration was mainly identified by comparing the solder joint interfaces at an anode and at a cathode. 2. Experimental procedure Fig. 1(a) and (b) show an overview of the sample and a schematic cross-section of a cathode side respectively. Oxygen free copper (C1020) was used for the electrodes and processed into 1-mm-long 0.4 mm × 0.4 mm cross sections. Soldering pad for the Cu electrode was finished by electroless Ni–P/Au plating. The Ni–P plating thickness varied from 7 to 14 μm. Two electrodes were jointed with a 0.5-mm-diameter solder ball. The solder composition was M725 (Sn–0.7Cu–Ni–P, wt.%) from Senjyu Metal Industry Co., Ltd. Fig. 2 shows the scanning electron microscope (SEM) image of the cross-section of a joint interface. The intermetallic compound (IMC) at the interface was ~ 2.5 μm thick. Energy dispersive X-ray spectroscopy (EDX) analysis showed that the IMC consisted of 33.7 at.% Cu–21.4 at.% Ni–44.9 at.% Sn. Therefore, the IMC composition can be estimated as (Cu,Ni)6Sn5. A P-rich layer was ~1.3 μm thick, and this layer consists of mixture of Ni3P and Ni [31]. The sample composition was confirmed in Fig. 1(b). We applied currents of 8.0 and 12 A to the solder joint to generate current densities of 5.0 and 7.5 kA/cm2 respectively. This test sample was placed on a hot plate, which was controlled so that the joint temperature was maintained at 423 K when a specified constant direct current passed through the joint. The temperature profile of the test sample was determined by thermocouple and by measuring the resistance change in the junction line between the Cu/solder/Cu joints. The change in resistance caused by electromigration was measured by monitoring voltage. The failure criterion was defined as a 20% increase in resistance from the initial value. The failed sample cross-section was polished and studied using a SEM. The electromigration failure mode was investigated by analyzing the elemental composition with EDX. To check the electromigration process phenomenon, the joint was studied after applying a stress of 423 K and 5.0 kA/cm2 for 250, 500, 750, and 2500 h. 3. Results and discussion 3.1. Electromigration stress test Fig. 3 shows the change in resistance with time for the stress condition at 423 K with current densities of 7.5 and 5.0 kA/cm2. The failure
Fig. 1. EM test sample. (a) Overview; (b) Schematic of cathode.
Fig. 2. Cross-section SEM image of the solder joint interface after reflow.
times were 1060 and 2320 h respectively. The failure samples after the electromigration stress tests are shown in Fig. 4. The solder of both failed samples was detached from the Cu electrode at the cathode side. The samples maintained at 423 K without applying current were not failed at the joint interface. Therefore, the failure mode of both samples with applied current was caused by electromigration. Fig. 5 shows a SEM image of the failed sample for a stress condition of 423 K and 7.5 kA/cm2. This sample was failed at the solder of a cathode side as shown in Fig. 5(a). (Cu,Ni)6Sn5 IMC existed entirely in the solder layer. Fig. 5(b) and (c) show enlarged images of the cracking area. The Ni–P plating and P-rich layers, which remained after reflow, disappeared. The Cu electrode reacted with the solder and formed Cu3Sn and (Cu,Ni)6Sn5 IMCs. Breakdown by electromigration occurred in the solder though continuous Cu–P layer was detected by EDX measurements on the solder surface. EDX point analysis on the layer indicated the atomic ratio of Cu and P was about 3:1, suggesting the compound may be Cu3P IMC. Jang et al. reported that electromigration makes P in the Ni–P plating at a cathode side diffused and reacted with the Cu electrode (without Ni–P plating) at an anode resulting in the formation of Cu–P IMC [18]. The position of Cu–P IMC in this sample coincided with that of the P-rich layer. Therefore, it is thought that electromigration makes Cu of the electrode diffuse to the P-rich layer to form Cu–P IMC. (Cu,Ni)6Sn5 IMC which remained after reflow was migrated by electron and failure occurred at the Cu–P IMC interface.
Fig. 3. Change of resistance ratio under electromigration stress.
2556
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559
Fig. 4. Overview of failure samples (a) 423 K, 7.5 kA/cm2, 1060 h; (b) 423 K, 5.0 kA/cm2, 2320 h.
Fig. 5(d) shows a cross-section of a joint interface at the anode side. The P-rich layer was ~1.8 μm thick and grew 1.4 times thicker than the initial one. However, the Ni–P plating remained and the influence of electromigration at the anode side was less than that at the cathode side. 3.2. P-rich layer behavior with electromigration stress To confirm the P-rich layer behavior influenced by electromigration, test sample cross sections were analyzed after applying a stress of 423 K and 5.0 kA/cm2 for 250, 500, 750, and 2500 h. No samples exceeded the current breakdown criterion of 20% resistance increase from the initial value. Fig. 6 shows cross-section SEM images after applying a stress of
423 K and 5.0 kA/cm2. The thickness of the (Cu,Ni)6Sn5 IMC at the cathode side was ~ 15 μm after 250 h. It grew approximately six times thicker than initial value. After 500 h, the (Cu,Ni)6Sn5 IMC became thinner. After 2500 h, the (Cu,Ni)6Sn5 IMC disappeared from the joint interface and Ni3SnP IMC formed over the joint instead. Thus, electromigration caused the decomposition of (Cu,Ni)6Sn5 IMC. The Prich layer at the cathode side grew with time, and all the Ni–P plating changed into the continuous P-rich layer (~8 μm in thickness) resulting in crack formation. Delamination at the cathode interface of the Cu electrode was confirmed. The Ni–P plating layer (P: 9.5 wt.%) was amorphous and compressive stress was applied. The volume shrinkage by transformation from an amorphous (Ni–P) layer to a crystal (Ni3P) one caused cracks and delamination [32,33]. The P-rich layer at the anode side was ~2.0 μm thick after 2500 h. It grew thicker after reflow but the rate at the anode side was slower than at the cathode side. Fig. 7 shows a cross-section SEM image of a joint with annealing only at 423 K for 2500 h. A Ni–P plating layer remained and the interface of the Cu electrode was undetached, although the (Cu,Ni)6Sn5 IMC and P-rich layer grew thicker than initial value. The P-rich layer was ~4.0 μm thick. Electromigration almost doubled the growth rate of the P-rich layer at the cathode side after 2500 h and halved that at the anode side after 2500 h. Fig. 8 shows the comparison in Ni–P thickness with time at 423 K with current density of 5.0 kA/cm2 and without current density. The reaction rate of Ni–P at the cathode side at 423 K with current density was faster than that at 423 K without current density until 2500 h. Therefore, it is believed that electromigration enhances diffusion of Ni from the Ni–P plating to the solder layer. The P-rich layer at the anode side increased slowly than that with only annealing. Thus, the electromigration suppressed the diffusion of Ni in the Ni–P plating at the anode toward the solder. 3.3. Formation of Cu–Sn IMC under P-rich layer As shown in Fig. 5, after current stressing, Sn diffuses into the Cu electrode passing through the Ni–P layer and formed Cu–Sn IMC. However, as shown in Fig. 6, Ni–P plating and P-rich layers still existed on the Cu electrode, and so the solder did not contact the Cu electrode directly. Fig. 9(a) shows a cross-section SEM image of a joint after 2500 h at 423 K with current density of 5.0 kA/cm2. The Ni–P plating changed into a P-rich layer and cracks existed in Fig. 9(b). These cracks opened the diffusion paths of Cu and Sn atoms, and may accelerate Cu migration
Fig. 5. (a) Cross-section SEM image of the solder joint after 1060 h with 7.5 kA/cm2 at 423 K; (b), (c) enlarged images of cathode side; (d) enlarged image of anode side.
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559
2557
Fig. 6. Cross-section SEM images of the solder joint at the cathode and anode with 5.0 kA/cm2 at 423 K.
toward the anode. Sn was detected by EDX measurements inside the cracks of the P-rich layer on Ni3SnP IMC or (Cu,Ni)6Sn5 IMC as shown in Fig. 9(c). Though Sn was transferred from each IMC, the transfer direction of Sn was opposite to the electron flow in this electromigration test. It is thought that Sn is transferred by thermal diffusion. Two types of IMC were confirmed inside cracks of the P-rich layer as shown in Fig. 9(d). One IMC in the Ni3SnP side was estimated to be (Ni,Cu)3Sn4 IMC because its composition was 34.1 at.% Ni–10.5 at.% Cu–55.5 at.% Sn from elemental EDX analysis. The other IMC was estimated to be Cu3Sn IMC because its composition was 75.2 at.% Cu–24.8 at.% Sn from elemental EDX analysis. It is believed that the Sn of the Ni3SnP IMC is transferred by thermal diffusion, reacts with the Cu electrode and forms these IMCs. 3.4. Breakdown mechanism of the cathode side with electromigration stress Fig. 10 shows an estimated breakdown mechanism of the cathode side by analyzing the electromigration failure mode and test samples with electromigration stress. (1) Electromigration diffused Ni in Ni–P plating and accelerated the formation of a P-rich layer. When the P-rich layer reached a Cu electrode, this layer delaminated from the interface and cracks formed in this layer.
Fig. 7. Cross-section SEM image of the solder joint after 2500 h without applying current at 423 K.
(2) Sn diffused from Ni3SnP IMC to these cracks in the P-rich layer and formed Cu3Sn and (Ni,Cu)3Sn4 IMCs from the Cu electrode side. (3) Thermal diffusion provided Sn from the solder side and formed Cu3Sn and (Cu,Ni)6Sn5 IMCs on the Cu electrode interface. P from the P-rich layer and Cu formed the Cu–P IMC. (4) Sn at the cathode side was diffused by electromigration and broke down from the Cu–P IMC interface.
4. Conclusions Electromigration in the joint between the Ni–P plating and Sn–Cu solder has been investigated at 423 K with current densities lower than 10 kA/cm2. (1) The times to failure for the electromigration tests were 1060 and 2320 h with current densities of 7.5 and 5.0 kA/cm2 respectively. The electromigration breakdown mode occurred even with current density lower than 10 kA/cm2. The breakdown surface was the interface between Cu–P IMC at the cathode side and the solder. (2) The P-rich layer at the cathode side grew faster than that at the anode side and than that without applying current. The diffusion direction of Ni was as same as the electron flow. Therefore, electromigration accelerated Ni diffusion at the cathode side. (3) The P-rich layer at the anode side grew slower than that at the cathode side and than that without applying current. The diffusion direction of Ni toward the solder was opposite to that of
Fig. 8. The comparison of Ni–P reaction layer thickness with 5.0 kA/cm2 at 423 K.
2558
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559
Fig. 9. (a) Cross-section SEM image of the solder joint after 2500 h with 5.0 kA/cm2 at 423 K; (b) enlarged image of cathode side; (c),(d) enlarged images of P-rich layer.
the electron flow. Therefore, electromigration suppressed Ni diffusion at the anode side. (4) The P-rich layer reached the Cu electrode, cracks were formed. Sn was diffused from Ni3SnP IMC to these cracks in the P-rich layer and formed Cu–Sn IMCs with the Cu electrode.
Acknowledgments The authors would like to thank students in Chukyo University and colleagues in Toyota Motor Corporation for their helpful discussions. References
Fig. 10. Schematic of electromigration failure mechanism at the cathode.
[1] J.R. Black, Electromigration — a brief survey and some recent results, IEEE Trans. Electron Devices (1969) 338–347. [2] J.R. Black, Physics of electromigration, Annual Proceedings of Reliability Physics 1974, pp. 142–149. [3] H.B. Huntington, Electromigration in Metals, in: A.S. Nowick, J.J. Burton (Eds.), Diffusion in Solids Recent Development, Academic Press, New York 1975, pp. 303–352. [4] I.A. Blech, Electromigration in thin aluminium films on titanium nitride, J. Appl. Phys. 47 (4) (1975) 1203–1208. [5] J.R. Black, Electromigration of Al–Si alloy films, Annual Proceedings of Reliability Physics 1978, pp. 300–307. [6] J.R. Lloyd, K.N. Tu, J. Jaspal, The physics and material science of electromigration and thermomigration in solders, in: K.J. Puttlitz, K.A. Stalter (Eds.), Handbook of Lead Free Solder Technology for Microelectronic Assemblies, Marcel Dekker, New York 2002, pp. 827–850. [7] K.N. Tu, Recent advances on electromigration in very-large-scale-integration of interconnects, J. Appl. Phys. 94 (9) (2003) 5451–5473. [8] T. Matsubara, H. Yaguchi, T. Takaoka, K. Jinno, Development of new hybrid system for compact class vehicle”, Proceedings of JSAE 2009, Yokohama, Japan 2009, pp. 21–24. [9] N. Nozawa, T. Maekawa, E. Yagi, Y. Terao, Development of power control unit for compact class vehicle, Proceedings of 22th ISPSD 2010, Hiroshima, Japan 2010, pp. 43–45. [10] N. Hirano, K. Mamitsu, T. Okumura, Structural development of double-sided cooling power modules, Denso Tech. Rev. 16 (2011) 30–37. [11] Y. Sakamoto, Assembly technologies of double-sided cooling power modules, Denso Tech. Rev. 16 (2011) 46–56. [12] S. Miura, Y. Ookura, Y. Okabe, S. Momota, Development of power devices for power cards, Denso Tech. Rev. 16 (2011) 38–45. [13] K. Hamada, Present status a future prospects for electronics in EVs/HEVs and expectations for wide bandgap semiconductor devices, Mater. Sci. Forum 600–603 (2009) 889–893.
T. Kadoguchi et al. / Microelectronics Reliability 55 (2015) 2554–2559 [14] S. Hirose, Power electronics technology for the next generation environmentallyfriendly vehicles, Proceedings of the 24th Symposium Microelectronics, Osaka, Japan, Sep. 2014 2014, pp. 37–40. [15] Kadoguchi T, Suzuki Y, Kaji M, Nakajima K, Miyoshi T, Kawashima T, Okumura T. Semiconductor Module. U.S. Patent 8,810,026, Aug. 19, 2014 [16] Kadoguchi T, Iwasaki S, Kawashima T, Okumura T, Nishihata M. Semiconductor Device and Manufacturing Method Thereof. U.S. Patent 8,884,411, Nov. 11, 2014 [17] L.D. Chen, M.L. Huang, S.M. Zhou, Effect of electromigration on intermetallic compound formation in line-type Cu/Sn/Cu and Cu/Sn/Ni interconnects, Proceedings of 60th ECTC 2010, Las Vegas, NV 2010, pp. 176–181. [18] J.W. Jang, L.N. Ramanathan, D.R. Frear, Electromigration behavior of lead-free solder flip chip bumps on NiP/Cu metallization, J. Appl. Phys. 103 (12) (2008) 123506. [19] K.H. Kuo, J. Lee, C. Stan, F.L. Chien, R. Lee, J. Lau, Electromigration performance of printed Sn0.7Cu bumps with immersion tin surface finishing for flip chip applications, Proceedings of 62th ECTC 2012, Sparks, NV 2012, pp. 698–702. [20] L.N. Ramanathan, T.-Y.T. Lee, J.-W. Jang, S.-H. Chae, P.S. Ho, Current carrying capability of Sn0.7Cu solder bumps in flip chip modules for high power applications, Proceedings of 57th ECTC 2007, Reno, NV 2007, pp. 1456–1461. [21] S. Peng, L. Li, A comparison study of electromigration performance of Pb-free flip chip solder bumps, Proceedings of 59th ECTC 2009, San Diego, CA 2009, pp. 1456–1461. [22] M. Lu, P. Lauro, D.-Y. Shih, R. Polastre, C. Goldsmith, D.W. Henderson, Comparison of electromigration performance for Pb-free solders and surface finishes with Ni UBM, Proceedings of 58th ECTC 2008, Orlando, FL 2008, pp. 360–365. [23] Y.-S. Lai, Y.-T. Chiu, C.-W. Lee, Y.-H. Shao, J. Chen, Electromigration reliability and morphologies of Cu pillar flip-chip solder joints, Proceedings of 58th ECTC 2008, Orlando, FL 2008, pp. 330–335.
2559
[24] S.-H. Chael, J. Im, T. Uehling, S.H. Paul, Effects of UBM thickness, contact trace structure and solder joint scaling on electromigration reliability of Pb-free solder joints, Proceedings of 58th ECTC 2008, Orlando, FL 2008, pp. 354–359. [25] K. Yamanaka, Y. Tsukada, K. Suganuma, Solder electromigration in Cu/In/Cu flip chip joint system, J. Alloys Compd. 437 (2007) 186–190. [26] K. Yamanaka, Y. Tsukada, K. Suganuma, Studies on solder bump electromigration in Cu/Sn–3Ag–0.5Cu/Cu system, Microelectron. Reliab. 47 (2007) 1280–1287. [27] J.K. Dong Wook Kim, J. Lee, M.-J. Lee, S.Y. Pai, S. Chen, F. Kuo, Evaluation of electromigration (EM) life of ENEPIG and CuSOP surface finishes with various solder bump materials, Proceedings of 60th ECTC 2010, Las Vegas, NV 2010, pp. 1841–1845. [28] Y.-S. Lai, J.-M. Song, Electromigration reliability with respect to Cu content in solder joint system, Proceedings of 58th ECTC 2008, Orlando, FL 2008, pp. 1160–1163. [29] C.E. Ho, C.W. Fan, W.Z. Hsieh, Pronounced effects of Ni(P) thickness on the interfacial reaction and high impact resistance of the solder/Au/Pd(P)/Ni(P)Cu reactive system, Surf. Coat. Technol. 259 (2014) 244–251. [30] C.E. Ho, W.Z. Hsieh, T.H. Yang, Depletion and phase transformation of a submicron Ni(P) film in the early stage of soldering reaction between Sn–Ag–Cu and Au/ Pd(P)/Ni(P)/Cu, Electron. Mater. Lett. 11 (1) (2015) 155–163. [31] C.-W. Hwang, K. Suganuma, Interface microstructure between Ni–P alloy plating and Sn–Ag–(Cu) lead-free solders, J. Mater. Res. 18 (11) (2003) 2540–2543. [32] C. Baldwin, T.E. Such, Plating rates and physical properties of electroless nickel/ phosphorus alloy deposits, Trans. Inst. Met. Finish. 46 (1968) 73–80. [33] K. Parker, Effects of heat treatment on the properties of electroless nickel deposits, Plat. Surf. Finish. 68 (12) (1981) 71–77.