Ag Plating and Its Impact on Void-Free Ag/Sn Bumping

Transcription

1 Ag Plating and Its Impact on Void-Free Ag/Sn Bumping Hirokazu Ezawa, Kazuhito Higuchi, Msaharu Seto, Takashi Togasaki, Sachiko Takeda* and Rei Kiumi* Toshiba Corporation Semiconductor Company Advanced ULSI Process Engineering Department 8, Shinsugita-cho, Isogo-ku, Yokohama, , Japan *Ebara Corporation Precison Machinery Group phone: , fax: Abstract We have already developed the eutectic Sn-Ag solder bumping process by alloying Ag/Sn electroplated metal stacks to overcome some problems concerning Sn-Ag alloy plating. As the dimensions of solder bumps shrink, the effect of voids in the solder bumps on electromigration resistance must be discussed. For the Sn-Ag alloy plated bumps, voids in the solder bumps as reflow processed are difficult to be avoided. The large amount of degassed species from the alloy-plated bumps due to strong chemical agents trapped in the plated films has a close relation to generation of voids in the bumps. In contrast, though the stack plating process shows less degassing, micro-voids would be left at the interface of the plated stack as embryos of residual voids in the solder bumps. In this study, surface roughness of the underlying Ag films and degassing behavior of the Ag/Sn stack-plated bumps has been investigated using different types of Ag plating solutions. Gas analyses and X-ray imaging inspections were performed for the Ag/Sn plated stacks. Surface roughness of the underlying Ag layer was also characterized by a laser scanning and Scanning Electron Microscopy. From the experimental results, it has been confirmed that less degassing is the most important issue for the Ag/Sn plated stacks. In addition, improvement of surface roughness of the underlying Ag plated films must not be neglected. Introduction With aggressive miniaturization of feature sizes in ultra large scale integrated (ULSI) circuits on Si devices, flip chip interconnection with more than several thousand solder bumps on a chip has become a reality in first level packaging. For high operating System-On-Chips (SoCs), the concern of soft-error rates due to alpha-particle emission from the radioactive Pb of solder joints has emerged as a liability issue in high functional applications. From environmental concerns, the legislative and market requirements for the ban on Pb materials are crucial to complete the development of Pb-free bumping in time for a 2006 European deadline. Among a wide variety of Pb-free solders, Sn-Ag or Sn-Ag-Cu alloys manufactured by electroplating are successful candidates to replace Pb-containing bumps with Pb-free ones. Sn-Ag alloy plating has the fundamental difficulty in preparing the plating solutions due to a large difference in electrochemical potentials between Sn and Ag. The strong complexing agents must be added in the alloy plating solutions to control reduction of the metal ions. The additives unavoidably remain in the alloy-plated bumps. Voids could be generated in the solder bumps due to desorption and expansion of the gasified additives during solder reflow. In high volume producing lines, the concentration of the metal ions and the additives should be precisely monitored and frequently maintained. This impedes improvement of productivity. Under these circumstances, we have already developed the eutectic Sn-Ag bumping process by alloying Ag/Sn electroplated metal stacks [1]. It has been also confirmed that the Ag/Cu/Sn plated stacks can be completely transformed to the ternary solder alloys [2]. On the basis of the investigation into alloying behavior of the electroplated metal stacks, the novel approach to employ multi-stack electroplating has been qualified to exhibit wide versatility of solder materials and compositions. In the design rule for upcoming high-performance SoCs, each flip-chip solder joint will be required to carry higher than 0.4A. As the bump diameter is shrunk to less than 50µm, electromigration in the solder joints will be a major reliability problem. Current crowding at the cathode end where electron flow enters a solder bump from a global metal line via a final metal pad on a chip has been discussed in terms of nucleation and growth of void [3]. However, if voids reside in solder bumps before high current is applied, they are afraid to be nucleation points for electromigration failure. The random location of the voids in the bumps will cause a wide distribution of the mean-time-to-failure, limiting a maximum current density per solder bump. In the previous publication, evolution behavior of degassing has been characterized for the Sn-Ag alloy plated bumps by thermogravimetry-gas-chromatography/massspectrometry (TG-GC/MS) analysis, together with the X-ray imaging inspection to observe residual voids in the bumps [4]. The results show that the amount of degassed species from the Ag/Sn plated stacks is small compared with that from the Sn-Ag alloy-plated bumps, suggesting that the plated stack has potentially an advantage to reduce residual voids in the bumps because each of the plating solutions in the stack plating does not need the specific chemical additives. However, in the plated Ag/Sn stack, micro-voids are easy to reside at the interface of the metal stack due to rough surface of the underlying Ag films. Figure 1 shows an example of the micro-voids at the rough interface of the Ag/Sn plated stack when a Ag plating condition fluctuates. They will be afraid to be the embryos of growth of voids in the solder bumps during reflow. In this study, the effects of surface roughness of the plated Ag films and degassing behavior of the Ag/Sn stack plated bumps on generation of voids in the bumps has been investigated using different types of Ag plating solutions /05/$ IEEE 107

2 Sn (50µm) Ag (2µm) Figure 1. Cross-sectional SEM image of a Ag/Sn electroplated metal stack under an inadequate plating condition. Bumping process The Ti/Ni/Pd barrier film stack of a thickness of 0.45µm was deposited on thermally oxidized (100) Si wafers by a magnetron sputtering system with a base pressure of less than 10-6 Pa. A nega-type photoresist with a thickness of more than 50µm was then applied to the wafer, and conventional lithographic exposure and wet chemical development decided the size and location of the openings for electroplating in the resist mask. To fill the openings, electroplating of 2µm thick Ag was first performed and, then, 50µm thick Sn was deposited in the openings. After the resist mask was removed, exposing the plated metal stacks, the barrier film stack was selectively etched off and removed all the film from the field area between the plated metal stacks. The details of the process are described in the other publication [1]. To know the degassed species from electroplated films, TG-GC/MS analysis was performed for the single plated Sn bumps and the Ag/Sn stack-plated bumps. The sample chips were rectangles with a size of 6.0mm 10.0mm, sawed from the 200mm bump fabricated wafers. The sample chip had area-arrayed bumps with a pitch of 200µm. Each measurement used more than 10 chips including about bumps. TG-GC/MS involved heating the samples up to 300 C with a rate of 10 C/min. Solder bumps were inspected by X-ray imaging to observe voids. The inspection ignores small voids with less than 10µm in diameter due to the lack of resolution. (a) (b) Experimental In this study, two types of Ag plating were employed. A conventional silver alkane sulfonic acid (Ag(CH 3 SO 3 )) based solution adding a nonionic surfactant was prepared. The Ag metal content was 5g/l. The temperature of the plating solution was controlled at 20 C. The other Ag plating was performed using a silver iodide (AgI) solution with a potassium iodide electrolyte at 30 C. The Ag metal content was 30g/l. The Sn(CH 3 SO 3 ) based plating solution adding a nonionic surfactant was also prepared. The Sn metal content was 60g/l. Sn electroplating was performed with a current density of 10mA/cm 2 at 20 C. Polarization curves were obtained for the Ag solutions using the reference electrode of Hg/HgSO 4 /H 2 SO 4 and the working electrode of Pt that rotates at 0 to 4000rpm to know stability of the solutions under high current densities. The surface morphology of the Ag plated films was characterized by a laser scanning microscopy and Scanning Electron Microscopy (SEM). Laser scanning was performed for 80µm squared areas of the surface of the Ag plated films. Surface roughness is defined in terms of the equation, Ra=1/L dx, where Ra is the centerline average roughness, x is the axis of the scan, L is the profile sample distance in the x-direction, and is the surface deviation measured from the centerline. 108 Figure 2. Polarization curves for non-cyanide Ag plating solutions based on (a) Ag(CH 3 SO 3 ) and (b) AgI. Results and discussion Ag plating Figure 2 shows the polarization curves for Ag plating solutions. The potential, at which deposition starts, and the limiting current density at 0rpm are 0.02V and 1.4mA/cm 2 for the Ag(CH 3 SO 3 ) solution, respectively. For the AgI solution, they are -0.83V and 9.0mA/cm 2, respectively. This result has

3 confirmed that the AgI solution shows a wide process window for Ag plating. Figure 3 shows the SEM images of the surfaces of the Ag plated films under different current densities. In the Ag plating using the Ag(CH3SO3) solution, as shown in Figure 3(a), porous Ag films were obtained at current densities of higher than 10mA/cm2. No significant change of surface roughness was observed for the Ag films plated by the AgI solution in Figure 3(b). In Figure 4, the surface roughness of the plated Ag films, determined by laser-scanning, is plotted against the plating current density. At less than 10mA/cm2 plating current densities, a slight decrease in Ra was observed with increasing the plating current density for the Ag films plated using the Ag(CH3SO3) solution. In contrast, the Ra value of the Ag film plated by the AgI solution was not significantly changed at increased plating current densities up to 50mA/cm2. In general, the amount of chemical additives trapped in plated films increases with increasing plating current density. To know the effect of the Ag film properties on generation of voids in the Ag/Sn plated bumps, the Ag plating by the AgI solution can extend the investigation up to higher plating current densities. Figure 4. Surface roughness of Ra, measured by a laser scanning vs. plating current densities for the Ag films plated using the Ag(CH3SO3) solution and the AgI solution, represented by the open circles and the closed circles, respectively. Figure 3. SEM images of the surface of Ag films plated using (a) the Ag(CH3SO3) solution and (b) the AgI solution at different current densities. Organic gasses of masses over 100 were not detected. Desorption of H2O and CO2 can be identified by the spectra of masses 18 and 44, respectively. Though the other masses would be considered as hydrocarbon gasses, they cannot be clearly identified at present. Mass peaks of m/z=43, 57, 71 and 85 emerged with a periodic pitch of mass 14 (N, CH2), Degassing behavior and X-ray inspection which might be caused by decomposition of some organic Figure 5 shows the mass-spectra of degassed species from compounds. The mass peak of m/z=27 at 258 C could also the single plated Sn bumps detected at 70 C, 178 C and result from decomposition of mass 44, freeing mass 16 (O, 258 C. The evolutions of m/z=18, 41, 43, 44, 57, 71 and 85 CH4, NH2). More precise investigation must be required to were detected by indexing mass numbers in comparison with know if unidentified organic gasses are due to decomposition predictable distributions of mass-peaks as close as possible. of chemical additives in the plating solution. Figure 6 shows a typical example of the mass spectra of degassed species from the Ag/Sn plated stack. The underlying Ag was electroplated at a current density of 50mA/cm2 using 109

4 the AgI solution. The evolutions of m/z=18, 41, 43 and 44 were detected for the Ag/Sn plated stack. It should be noted that the intensity of the other peaks observed for the single plated Sn bumps in Figure 5 looks weak at the expense of the higher intensity of the major peaks in Figure 6. No mass peaks of masses over 100 were detected for the Ag/Sn plated stack. Figure 7 shows the degassing profiles for m/z=18, 41, 43 and 44 with increasing the temperature of the Ag/Sn plated stacks. In the profile of H 2 O for the sample with the Ag plating current density of 50mA/cm 2, steep evolution started around 230 C where melting of the bumps would occur and, then, a maximum evolution rate was observed at approximately 245 C. In the profile of mass 41, a maximum was observed around 270 C. Similar degassing behavior was observed for the Ag/Sn stack with the Ag plating current density of 10mA/cm 2. By integrating the evolution-rate curves, the amounts of the degassed species in the unit of wtppm are summarized in Table 1. The total amount of degassing species slightly increases with increasing the Ag plating current density from 10mA/cm 2 to 50mA/cm 2. After the gas analyses, X-ray inspection was performed for the bumps on the degas-measured samples. Figure 8 shows the X-ray images for Sn-Ag bumps by the Ag/Sn plated stacks with the underlying Ag films plated using the AgI solution. Large voids were observed in the Ag/Sn stacks with the Ag film plated at a current density of 50mA/cm 2. In contrast, in the case of the Ag film plated at a current density of 10mA/cm 2, tiny voids were found near the edges of the Ag/Sn stack plated bumps. As shown in Figure 3 and Figure 4, the underlying Ag films plated using the AgI solution do not show significant changes of surface roughness within the range of the plating current density from 10mA/cm 2 to 50mA/cm 2. Thus, the effect of surface roughness of the Ag plated films on generation of voids in the bumps can be ruled out in the discussion about the results of X-ray images. Degassing behavior can dominantly account for the significant difference in the appearance of voids in the Ag/Sn plated stacks as shown in Figure 8. It should be noted that, in these experiments, the Ag/Sn plated stacks have undergone a simulated reflow operation with heating the samples at a rate of 10 C /min up to 300 C in a specific chamber for TG-GC/MS, where no reflow flux was actually coated on the plated metal stacks. In this study, the impact of reflow flux and soldering on generation of voids in solder bumps is not considered. However, the results of the gas analyses together with the X-ray images suggest that less degassing of the plated stacks leads to fewer voids in the bumps. Figure 5. Mass spectra of degassed species from the Sn plated bumps at 70 C, 178 C and 258 C. Figure 5. Mass spectra of degassed species from the plated Sn bumps at 70 C, 243 C and 258 C Figure 6. Mass spectra of degassed species from the Ag/Sn plated stack at 246 C and 271 C. The underlying Ag was plated at 50mA/cm 2 using the AgI solution. 110

5 Figure 7. Degassing profiles for masses 18, 41, 43 and 44 with a heating rate of 10 C/min for the Ag/Sn plated stacks. Figure 8. X-ray images of Sn-Ag bumps by the Ag/Sn plated stacks with underlying Ag plated at the current densities of (a) 10mA/cm 2 and (b) 50mA/cm 2 using the AgI solution, observed after degas analyses. In the previous study, degassing behavior of the Sn-Ag alloy plated bumps using a conventional CH 4 SO 3 based solution has been characterized by evolutions of organis gases of over 100 massses[4], summarizing the previous results in Table 2 for the Sn-Ag alloy plated bumps and the Ag/Sn stack plated bumps with underlying Ag plated using a Ag(CH 3 SO 3 ) solution. Figure 9 also shows the previous results of X-ray void inspection for them[4]. According to the previous study, large amounts of strong complexing agents trapped in the Sn- Ag alloy plated bumps would be degassed during annealing for gas analyses, resulting in generation of some large voids in the bumps. The Ag/Sn stack plating process does not need specific chemical agents in plating solutions. This may cause reduction of the amounts of degassed species from the Ag/Sn plated stacks as shown in Table 2, leading to no detection of voids in the bumps as shown in Figure 9. In this study, in the Ag/Sn stack plating process, the Ag(CH 3 SO 3 ) solution is replaced with the AgI solution, as is described above. As shown in the tables, the amount of degassed H 2 O for the stacks with Ag plated by the AgI solution is less than that by the Ag(CH 3 SO 3 ) solution. In contrast, less degassing of masses 41, 43 and 44 is remarkable for the stacks with Ag plated by the Ag(CH 3 SO 3 ) solution. This suggests that less chemical addtives trapped in the underlying Ag films will be required to realize void-free bumping. Comparing Figure 8(a) with Figure 9(b) at a plating current density of 10mA/cm 2 for the underlying Ag layer of the Ag/Sn stacks, the AgI solution provides more voids in the stacks than the Ag(CH 3 SO 3 ) solution after gas analyses. The appearance of voids in the bumps may be due to the exisitence of micro-voids at the interface of the plated stacks before annealing.the difference in surface roughness between the Ag films, as shown in Figure 4, must be also considered to realize void-free solder bumping. For future fine-pitch solder joints of flip chip interconnect, a wide distribution of electromigration failure rate due to the exisitence of voids in solder bumps will restrict the design rule for SoCs. For the reason, a void-free scheme will be crucial to the viability of solder bumping for future device generations. Conclusion The effects of degassing behavior of the Ag/Sn plated stacks and surface roughness of the underlying Ag films on generation of voids in the solder bumps have been investigated in this study. Compared with a Ag(CH 3 SO 3 ) plating solution, the Ag plating using a AgI solution extended the investigation to higher plating current densities. The experimental results suggest that less degassing leads to fewer residual voids in the stack plated bumps. It has been confirmed that the appearance of voids in the Ag/Sn plated stacks after annealing depends strongly on degassing behavior of the plated stacks. However, surface roughness of the underlying Ag films must be also considered to eliminate embryos of voids which would reside around the interface of the plated stacks. 111

6 Figure 9. X-ray images of Sn-Ag bumps by (a) Sn-Ag alloy plating and (b) the Ag/Sn stack plating with underlying Ag plated using the Ag(CH 3 SO 3 ) solution, observed after degas analyses [4]. Acknowledgments The assistance from Ebara Corporation in helpful discussions with Nobutoshi Saito, Dr. Fumio Kuriyama of Precision Machinery Group is gratefully appreciated. The authors also wish to thank Toray Research Center, Inc. for TG-GC/MS analyses. References 1. H. Ezawa, M. Miyata, S. Honma, H. Inoue, T. Tokuoka, J. Yoshioka, and M. Tsujimura, Eutectic Sn-Ag Solder Bump Process for ULSI Flip Chip Technology, IEEE Trans.Elecronics Packaging Manufacturing, Vol. 24,No.4 (2001),pp H.Ezawa, M. Miyata, M. Seto, and S. Honma, Pb-free Bumping by Alloying Electroplated Metal Stacks, Proceedings of 53 rd Electronic Components & Technology Conference, 2003, pp K. N. Tu, Recent advances on electromigration in verylarge-scale-integration of interconnects, J. Appl. Phys. Vol.94, No.9(2003),pp H.Ezawa, M.Seto, and H.Katsumata, Pb-Free Bumping for High-Performance SoCs, Proceedings of 54 th Electronic Components & Technology Conference, 2004, pp Table 1. The amounts of degassed species under different Ag plating conditions using the AgI solution. m/z Degassed species Ag(50mA/cm 2 ) / Sn stack Ag(10mA/cm 2 ) / Sn stack 18 H 2 O Hydrocarbon Hydrocarbon CO Total amount (wtppm) Table 2. The amounts of degassed species for Sn-Ag alloy plated bumps and Ag/Sn plated stacks [4]. m/z Degassed species Sn-Ag alloy-plated Bumps Ag(Ag(CH 3 SO 3 ))/Sn stack 18 H 2 O Hydrocarbon CO Organic gas 26.9 Not detected Total amount (wtppm)