Inhibition of Electromigration in Eutectic SnBi Solder Interconnect by Plastic Prestraining

Transcription

1 J. Mater. Sci. Technol., 2011, 27(11), Inhibition of Electromigration in Eutectic SnBi Solder Interconnect by Plastic Prestraining X.F. Zhang 1), H.Y. Liu 1), J.D. Guo 1) and J.K. Shang 1,2) 1) Shenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese Academy of Sciences, Shenyang , China 2) Department of Materials Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA [Manuscript received July 1, 2010, in revised form September 23, 2010] Plastic prestraining was applied to a solder interconnect to introduce internal defects such as dislocations in order to investigate the interaction of dislocations with electromigration damage. Above a critical prestrain, Bi interfacial segregation to the anode, a clear indication of electromigration damage in SnBi solder interconnect, was effectively prevented. Such an inhibiting effect is apparently contrary to the common notion that dislocations often act as fast diffusion paths. It is suggested that the dislocations introduced by plastic prestraining acted as sinks for vacancies in the early stage of the electromigration process, but as the vacancies accumulated at the dislocations, climb of those dislocations prompted recovery of the deformed samples under current stressing, greatly decreasing the density of dislocation and vacancy in the solder, leading to slower diffusion of Bi atoms. KEY WORDS: Electromigration; Interfacial segregation; Prestrain; Dislocation; Vacancy 1. Introduction With continued downward scaling of feature sizes in electronic devices and the corresponding increase in the current density, the reliability of solder joints under electromigration becomes a crucial problem in advanced interconnects [1]. Although many studies have addressed the electromigration issues of the solder interconnects, very few solutions are available to mitigate the electromigration damage. So far, elemental doping is the only successful way to retard the electromigration in solder interconnects [2,3]. However, the alloying elements usually affect other properties of the solders, such as solderability, melting point, and mechanical properties, etc. Therefore it is imperative that alternative methods for mitigating electromigration in solder interconnects be developed. Plastic deformation usually introduced a large amount of defects into the metallic alloys, such as Corresponding author. Prof., Ph.D.; Tel./Fax: ; address: jkshang@imr.ac.cn (J.K. Shang). dislocations, vacancies, and so on. These defects may greatly affect the diffusivity of atoms, and then affect the electromigration process. In this study, we designed an experiment to probe the interaction of deformation with electromigration damage in SnBi by intentionally introducing pre-strain to the alloy immediately prior to electromigration experiments. Under current stressing, Bi atoms are prone to migrate toward the anode and form a continuous Bi layer at the anode side, which may decrease the reliability of the solder interconnects [4,5]. For the SnBi interconnect, the electromigration mechanism was controlled by Bi migration rather than by vacancy-mediated Sn diffusion [4,5]. By conducting electromigration tests after prestraining to various levels, an inhibiting effect was observed, namely Bi interfacial segregation to anode was effectively prevented rather than enhanced, as expected from a vacancy mechanism. 2. Experimental The Sn 58 wt%bi alloy used in this study was pre-

2 X.F. Zhang et al.: J. Mater. Sci. Technol., 2011, 27(11), Fig. 1 Characterization of the sample preparation process and the temperature profile adopted for the reflow process pared by melting pure Sn and pure Bi in a vacuum quartz tube at 300 C for 30 min. The solder alloy has a melting temperature of about 138 C. Prior to soldering, two copper cubes, 10 mm wide, 10 mm thick and 10 mm long, were ground and carefully polished to form smooth surfaces. Between the two adjoining surfaces, Mo wires, 280 μm in diameter, were placed for control of the solder bond thickness. The two copper cubes were then aligned and fixed before the assembly was heated in an oven where the solder was reflowed at 180 C for 60 s. After air cooling, the asreflowed solder joints were cut into slender bars with a cross section roughly 500 μm 500 μm by electric discharge machining. These bar specimens were further polished to about 330 μm 330 μm in cross section. Figure 1 characterized the sample preparation process and the temperature profile adopted for the reflow process. These samples were prestrained on a tensile testing machine in strain control at a rate 10 3 s 1 and 25 C, to total strains of 0, 2%, 8%, 20%, respectively. In order to keep the prestrained state of the samples before current stressing and minimize the effect of additional factors on the experimental results as much as possible, these samples were kept in refrigerator at 20 C before current stressing. Electromigration testing of the solder interconnect was conducted by applying direct currents to both ends of the bar samples using copper wires as the electric leads. The entire fixture was then immersed in an oil bath held at a constant room temperature of 25 C. The sample temperature was monitored by a K-type thermocouple placed next to the solder interconnect. The current density in the samples was A/cm 2, the temperature of the samples was about 55 C, and the time under current stressing was 180 h. After current stressing, the interfaces between solders and substrates were examined by scanning electron microscopy (SEM). The composition of the precipitates and phases between solders and substrates was analyzed by energy dispersive spectroscopy (EDS). After electromigration, tensile strengths of the solder joints were measured on a micromechanical testing system at a constant strain rate of 10 3 s 1 and 25 C. In comparison with the electromigration effect, tensile strengths of the solder joints without electromigration were also investigated in the same conditions. In order to reduce measurement error, each experimental point was made more than three samples tested. Because the Bi segregation is uneven, it is difficult to measure the thickness of the Bi-rich layer directly. In the present work, a self-developed measurement software was used to measure the thickness of Bi-rich layer. The SEM images were first input into the Photoshop software; and then the area of the total Bi-rich layer was measured by pixel analysis. The average thickness of the Bi-rich layer was finally calculated by the following formula: H = S/L where H, S, and L are the average thickness, area

3 1074 X.F. Zhang et al.: J. Mater. Sci. Technol., 2011, 27(11), Fig. 2 Microstructures of Sn-58Bi solder interconnect before current stressing and length of the Bi-rich layer, respectively. Therefore, no matter what the shape of Bi-rich layer would be,anaveragethicknesscanbeobtainedwithaminor error. Only the thickness of a continuous Bi layer was measured. All the thickness mentioned in the article refers to a continuous Bi layer thickness. 3. Results and Discussion The initial microstructure of the Sn-Bi/Cu interconnect at the as-reflowed state is shown in Fig. 2, where Cu substrates appeared in a heavy dark color. The solder itself consisted of the Sn-rich phase (dark color) and the Bi-rich phase (bright color). The two phases were distributed in the solder. At the SnBi/Cu interface, a thin layer of Cu 6 Sn 5 compound (55Cu- 45Sn in at.% by EDS analysis) can be seen to adhere to the Cu pad. After application of electric current, the microstructure in the solder interconnects experienced significant changes. Figure 3 presents the changes of the interfacial microstructure at different applied prestrains, after current stressing for 180 h at a current density of A/cm 2 at 55 C. The images on the left are for the anode side and those on the right are for the cathode. For samples without prestrain, a continuous Bi layer about 4.6 μm in thickness was accumulated at the anode side, as shown in Fig. 3(a). At the cathode side (Fig. 3(b)), the interfacial microstructure turned asymmetric to the anode side and no continuous Bi layer was observed at the interface; compositional analysis indicated that the region next to the interface was Sn rich, containing 99 at.% Sn. This segregation of Bi to the anode side is similar to those found in the previous studies [4,5]. When prestrain was introduced into the solder interconnects, the Bi interfacial segregation under current stressing was effectively retarded. For the sample with 2 % prestrain, the Bi-rich continuous layer at the anode side is only 2.8 μm in thickness, which is 40 % thinner than that without prestrain, as shown in Fig. 3(c); while at the cathode side, the interfacial microstructure remained almost unchanged compared to the as-reflowed state, as shown in Fig. 3(d). No continuous Bi-rich layer or Sn-rich region existed at the cathode interface except for some coarsening of the Bi phase. As the prestrain increased to 8% and 20%, the thickness of Bi-rich layer at the anode side decreased to 2.4 μm and 1.8 μm, respectively, as shown in Fig. 3(e) and (g). Since less Bi segregation to the anode is obtained at the higher prestrain, the change of the microstructure in the cathode (Fig. 3 (f) and (h)) are also slight as well. Figure 4 presents the relationship between the Bi segregation and prestrain under current stressing. Remarkably, the greater the prestrain imposed on the samples, the less the amount of Bi segregation to the anode side. From above, it is clear that the Bi segregation under current stressing was inhibited by imposing prestrains on the sample. The effect of electromigration on the tensile strength of the solder joint is shown in Fig. 5. From the tensile stress-strain curves shown in Fig. 5(a), it is evident that the prestraining only or electromigration alone reduced the tensile strengths of the solder joint, but current stressing after prestraining recovered the tensile strengths to nearly the level of the as-reflowed state. However, the elongation of these samples after current stressing was somewhat decreased. Figure 5(b) shows the tensile strength vs. prestrain before and after electromigration. It can be seen that the tensile strength of the solder joints without prestrain after electromigration was reduced by about 16 MPa (from 72 MPa to 56 MPa) in comparison with that in the as-reflowed state. However, tensile strengths of the prestrained solder joints after electromigration are only reduced by about 10 MPa, 2 MPa, 3 MPa at strains of 2%, 8%, 20%, respectively. Above 8% prestrain, the tensile strength of the solder joint after electromigration approached the level without electromigration. Generally, the atomic flux [6,7] under current stressing can be expressed as J = J chem + J em = CD ln C (kt kt x + Z ee) where J chem is the diffusion term due to the chemical potential gradient, J em the drift term due to electron momentum transfer effect, C the atomic density, D the diffusion constant, k Boltzmann constant, T the absolute temperature, Z theeffectivechargenumber, e the electronic charge, E the electric field and E=ρj, ρ the resistivity and j the current density. For Cu/Sn58Bi/Cu solder joints, Bi is the dominant diffusing species during the electromigration process, especially for the interfacial segregation [4,5]. As the concentration gradients at the anode side and the cathode side were symmetrical, the migration of Bi atoms under current stressing was mainly attributed to J em = CD kt Z ee, which means that the Bi

4 X.F. Zhang et al.: J. Mater. Sci. Technol., 2011, 27(11), Fig. 3 SEM images showing the microstructure of the samples at the anode side (a,c,e,g) and at the cathode side (b,d,f,h) after current stressing, (a), (b) no prestrain; (c), (d) 2% prestrain; (e), (f) 8% prestrain; (g), (h) 20% prestrain Fig. 4 Relationship between the Bi segregation and prestrain under current stressing atomic flux was only determined by the diffusion constant D at a certain current density and a certain temperature. Based on the common view point, the density of defects, such as dislocations, vacancies, and so on, should be increased after deformation. Since dislocations are known to be fast diffusion paths, the diffusion constant D should increase after tensile test, and the Bi atomic flux enhanced. However, our experimental results showed otherwise. In the early stage of the electromigration experiment of the prestrained sample, vacancies are created by the electromigration force and attracted toward dislocations. Those vacancies may interact with the dislocations further by either exerting an osmotic pressure to drive dislocation climb or by diffusing rapidly along the dislocation core as in pipe diffusion. However, since dislocations are randomly oriented and distributed from the plastic prestraining, the pipe diffusion leads to very little directional flow of the electromigration species, Bi, which is essential in producing the electric polarity effect. On the other hand, the climb of dislocations can result in the recovery of a deformed metal [8]. During the recovery Fig. 5 Variations of tensile properties of Sn-58Bi solder joints with prestrains: (a) tensile stress-strain curves at 20% prestrain, (b) tensile strength vs. prestrain before and after electromigration process, dislocations are able to glide, cross-slip and climb. If two dislocations of opposite sign meet then they effectively cancel out and their contribution to the stored energy is removed. Therefore, the density of defects, primarily dislocations, introduced by plastic deformation may be greatly decreased. Although the temperature of the sample under current stressing

5 1076 X.F. Zhang et al.: J. Mater. Sci. Technol., 2011, 27(11), is only 55 C, the melting point of the solder is also very low (138 C), thus the temperature of the solder interconnect (corresponding to homologous temperature of over 79%) is high enough to make recovery and even recrystallization possible. In addition, the current induced recovery and recrystallization usually has a much lower dislocation and vacancy density than that of common annealing [9,10]. Thus, the density of dislocation and vacancy in the pre-strained samples were even less than that in the samples without prestrain. Due to the lack of favorable neighboring site and rapid diffusion paths, the migration of Bi atoms to anode would be effectively retarded even under a high magnitude of electromigration driving force, as observed experimentally (Figs. 3 and 4). Furthermore, the recovery process may be more complete at the higher pre-strain level; thus, the inhibition effect of pre-strain on the electromigration is increased with the increase of pre-strain. AsthediffusionrateofBiatomsslowsdownand vacancies move to the sinks at the dislocations or with the dislocation climb, the density of vacancies formed under current stressing in the pre-strained joints may be less than the joints without pre-straining, so that the decrease of strength for the pre-strained joint after current-stressing was less than that without prestraining, as shown in Fig Conclusion Bi interfacial segregation to the anode was successfully retarded by imposing prestrains on solder interconnects before current stressing. In the early stage of the electromigration process, dislocations apparently acted more as vacancy sinks and as more vacancies accumulated at the dislocations, climb of the dislocations caused the recovery of the deformed samples under current stressing, which decreased the density of dislocation and vacancy in the solder and slowed the diffusion rate of Bi atoms. The suppression of the electromigration damage by introducing pre-strain into solder promises a potential new solution for enhancing electromigration resistance of solder interconnects. Acknowledgements This work was supported by the National Natural Science Foundation of China (Grant No ), the National Basic Research Program of China (Grant No. 2010CB631006) and the Natural Science Foundation of Liaoning Province, China (Grant No ). The authors would like to thank H.Y. Guo for helpful discussion. REFERENCES [1 ] K.N. Tu: J. Appl. Phys., 2003, 94, [2 ] C.M. Chen and C.C. Huang: J. Alloy. Compd., 2008, 461, 235. [3 ] C.C. Wei and C.Y. Liu: J. Mater. Res., 2005, 20, [4 ] Q.L. Yang and J.K. Shang: J. Electron. Mater., 2005, 34, [5 ] C.M. Chen, L.T. Chen and Y.S. Lin: J. Electron. Mater., 2007, 36, 168. [6 ] I.A. Blech and C. Herring: Appl. Phys. Lett., 1976, 29, 131. [7 ] H. Conrad: Mater. Sci. Eng. A, 2000, 287, 227. [8 ] R.W. Cahn and P. Haasen: P hysical Metallurgy, 4th edn, Elesvier Science B. V., 1996, [9 ] H. Conrad, N. Karam and S. Mannan: Scripta Metall., 1983, 17, 411. [10] Y.Z. Zhou, S.H. Xiao and J.D. Guo: Mater. Lett., 2004, 58, 1948.