Jeong et al.: Effect of the Formation of the Intermetallic Compounds (1/7)

Size: px

Start display at page:

Download "Jeong et al.: Effect of the Formation of the Intermetallic Compounds (1/7)"

Phillip Wright
6 years ago
Views:

1 Jeong et al.: Effect of the Formation of the Intermetallic Compounds (1/7) Effect of the Formation of the Intermetallic Compounds between a Tin Bump and an Electroplated Copper Thin Film on both the Mechanical and Electrical Properties of the Jointed Structures Seongcheol Jeong*, Naokazu Murata*, Yuki Sato**, Ken Suzuki** and Hideo Miura** *Department of Nanomechanics, Graduate School of Engineering, Tohoku University, Aoba, Aramaki, Aoba-ku, Sendai, Miyagi , Japan **Fracture and Reliability Research Institute, Graduate School of Engineering, Tohoku University, Aoba, Aramaki, Aoba-ku, Sendai, Miyagi , Japan (Received August 10, 2009; accepted November 24, 2009) Abstract The mechanical and electrical reliabilities of fine bumps with diameters and heights on the order of scores of microns were studied, considering the growth of the intermetallic compound (IMC) at the interface between a tin bump and a copper thin-film interconnection. It was found that an increase in the thickness of the IMC changed the stress and strain fields around the interface significantly, and thus, changed the fracture mode from a fatigue crack of the solder to a fatigue crack of the copper interconnection or to delamination between the IMC and the copper interconnection. This is because the mechanical properties of the grown IMC differ from those of copper and tin and that a large number of Kirkendall voids appeared around the interface. In addition, the resistance of the bumps increased dramatically with the increment of the IMC layer because of the growth of the Kirkendall voids. Therefore, it is very important to minimize the growth of the IMC in order to assure the reliability of the bump joint structures. Keywords: Flip Chip Technology, Area-arrayed Bump, Joint Reliability, Electroplated Copper Bump, Intermetallic Compound, Residual Stress 1. Introduction Electronic products such as mobile phones and PCs have been miniaturized continuously and their functions have been improved drastically. In highly integrated electronic package systems, the number of I/O pins that connect an LSI chip with a printed-circuit board or another chip has been increasing significantly. The maximum number in one LSI chip will reach 10,000 in the near future.[1] To realize such high-density interconnection structures, an area-arrayed small-bump structure has been employed in multi-chip packages and modules.[1, 2] Both the diameter and the pitch of the bumps will be decreased from a few hundred microns to dozens of microns. To maintain the low resistance of the miniaturized interconnection, the material of the thin film interconnection has been changed from aluminum alloy to electroplated copper. In addition, for environmental reasons, the material of the solder bump has been changed from conventiol lead-tin eutectic solder to lead-free solder that mainly consists of tin. As a result, IMCs such as Cu 6 Sn 5 and Cu 3 Sn easily grow at the interface between the copper interconnections and the lead-free solder bump.[3 7] Since both the mechanical and electrical properties of the IMC are quite different from those of tin and copper, the increase of the thickness of the IMC layer in the fine bump-interconnection joint structures may deteriorate the mechanical and electrical quality of the jointed structures. For example, the increase of the thickness of the IMC layer in a bump-jointed structure with bumps on the order of dozens of microns in size may change the stress-strain field in the structure significantly. Since the amplitude of the local deformation of a silicon chip thinner than 100 μm is a strong function of the mechanical properties of the bump material, such a change of the joint structure should also alter both the electronic functions and the reliability of the jointed LSI chips. Though the life of the jointed struc- 91

$1, 2009 ture has been dominated by the fatigue life of the solder Since the current density during electroplating of thin film bumps, the fracture mode of the jointed structure with a$ interconnections varies from 10 ma/cm2 to 100 ma/cm2 thick IMC layer may be changed to the fatigue life of the depending on the products, the test film was electroplated copper interconnection or to

interconnections varies from 10 ma/cm2 to 100 ma/cm2 thick IMC layer may be changed to the fatigue life of the depending on the products, the test film was electroplated copper interconnection or to

2 Transactions of The Japan Institute of Electronics Packaging Vol. 2, No. 1, 2009 ture has been dominated by the fatigue life of the solder Since the current density during electroplating of thin film bumps, the fracture mode of the jointed structure with a interconnections varies from 10 ma/cm2 to 100 ma/cm2 thick IMC layer may be changed to the fatigue life of the depending on the products, the test film was electroplated copper interconnection or to delamination between the on stainless steel under a constant current density of 20 grown IMC layer and the copper interconnection or ma/cm2 at 30 C. The thickness of the stacked bump was remaining solder bump. In addition, the electrical resis- about 50 μ m, and the thickness ratio between the copper tance of the bump-jointed structure may increase with the and tin was varied by considering the final composition of increase of the thickness of the IMC layer because the the grown IMC layer. In this figure, the thickness of both electronic conductivity of the IMC is much lower than that the copper and tin was about 25 μ m and this stacked layer of copper or tin. was used for the formation of Cu6Sn5. When the expected In this paper, the mechanical and electrical reliabilities composition was Cu3Sn, the thickness of the copper layer of bump-jointed structures, with bumps on the order of doz- was about 40 μ m and that of tin layer was about 10 μ m. ens of microns, are discussed experimentally and analyti- Then these bumps were annealed in pure argon gas at cally. First, the fracture mode of the bump structure was about 260 C or 350 C with various annealing times from 10 estimated using a finite element analysis considering the min. to 3 hrs. After the annealing, the composition of the measured mechanical properties of the grown IMC and top surface of the stacked tin layer was analyzed using an electroplated copper and tin thin films. Then, the change Auger Electron Spectrum (AES) analysis to confirm the of the electrical resistance of the bump-jointed structure growth of the IMC. It was confirmed that copper was dif- was measured as a function of the thickness of the IMC fused into the tin layer and formed the IMC. When the newly grown between the electroplated copper thin film stacked bump was annealed at 260 C for 1 hour, it was and the electroplated tin bump. found that the top layer consisted of Cu6Sn5. When the stacked bump was annealed at 350 C, the grown top layer 2. Test structures consisted of Cu3Sn. In addition, it was found that there was 200-μ m square stacked tin and copper bumps were elec- a several-micron-thick layer between the remaining copper troplated on thermally oxidized silicon wafers as shown in layer and the Cu6Sn5 layer which consisted of Cu3Sn. The Fig. 1. Thin Cr (100 nm) and Au (300 nm) films were depos- rough surface was then polished to a depth of about 1 μ m ited on the wafer by sputtering before electroplating. Elec- for nanoindentation. troplated copper thin films were grown on the sputtered During this annealing, the changes of the internal stress gold thin film. The composition of the plating bath used for of each layer and the grown IMC layer were measured by the electroplating was controlled by diluting 80 g of CuO detecting the change of the surface curvature of a substrate powder and 186 g of H2SO4 with 1000 ml of purified water. which was cut into thin rectangles from the wafer. The When the tin layer was electroplated, a commercial solution electroplated copper thin film shrank upon annealing as (PF-077S, made by Ishihara Chemical Co., Ltd.) was used. shown in Fig. 2 and the calculated change was a tensile stress of about 200 MPa remaining in the film at room tem- Fig. 1 Outlook of the bump-joint model structure (Stacked electroplated Sn/Cu bumps). 92 Fig. 2 Change of warpage of a strip sample of a copper thin film electroplated on a silicon substrate after annealing.

Jeong et al.: Effect of the Formation of the Intermetallic Compounds (3/7) perature after the annealing. No significant change was observed in the electroplated tin layer.

thermal stress between room temperature and the annealing temperature. These values were used for a finite element analysis of the stress-strain field of the bump-interconnection jointed structure.

3 Jeong et al.: Effect of the Formation of the Intermetallic Compounds (3/7) perature after the annealing. No significant change was observed in the electroplated tin layer. A tensile stress of about 50 MPa was measured in the grown IMC layer, and the reaction-induced stress which appeared at the annealing temperature was determined as about 400 MPa by considering the thermal stress between room temperature and the annealing temperature. These values were used for a finite element analysis of the stress-strain field of the bump-interconnection jointed structure. When a silicon chip thinned to about 70 μm was mounted on the area-arrayed bump structure, clear local deformation of about 1 μm was observed, as shown in Fig. 3, after curing of underfill material at 150 C. Such local deformation of a silicon chip causes a large local distribution of residual stress of about 200 MPa and thus, causes a serious disparity between the electronic functions of devices before and after assembly. This local deformation of a silicon chip is a strong function of the thickness of the chip and the difference in the material constants between the metallic bumps and underfill material. Therefore, it is very important to understand both the electrical and mechanical properties of the bump joint structure. 3. Measurement of the Young s modulus of the IMC layers A Nanoindenter DCM-SA2 made by MTS Corp. was used for the measurement of the Young s modulus of the grown IMC layers. The Young s modulus of each sample was measured from its surface. The indentation depth was about 500 nm. Since the thickness of the grown IMC was more than 10 μm, the measured results were considered not to include the effect of the substrate or the under-layer. The resolution of the displacement measurement was 0.2 pm and that of the applied force was 1 nn. The strain rate during the measurement was fixed at 0.05%/s. In addition, the continuous stiffness measurement method was applied to the measurement of the distribution of the Young s modulus of the film along its thickness direction. A small alternative current was applied to a load cell coil of the indenter to make small vibrations at a tip of the indenter. The Young s modulus of a film was measured at the unloading process of each small vibration. No change of the mechanical properties caused by strain hardening due to this vibration occurs in the film because the amplitude of this vibration was sufficiently low. The indentation-depth dependence of the Young s modulus of the electroplated tin and copper films was measured as shown in Fig. 4. The measured Young s modulus of the electroplated tin thin films was constant at about 50 GPa. It agreed well with that of bulk tin. However, the Young s modulus of the electroplated copper thin films showed wide variation from about 40 GPa to 90 GPa at a depth of 500 nm. This variation was caused by the micro texture of the film as shown in Fig. 5. This photo is a scanning electron micrograph of an electroplated copper thin film with its surface partially etched by a focused ion beam. The film mainly consisted of fine columnar grains with porous grain boundaries. Such fine grains whose diame- Fig. 4 Indentation depth dependence of Young s modulus of electroplated tin and copper thin films. Fig. 3 Residual surface morphology of a silicon chip mounted on an area-arrayed fine bump structure. Fig. 5 Scanning electron micrograph of micro texture of an electroplated copper thin film partially etched off by ion beam. 93

[8, 9] Next, the Young s modulus of the grown Cu6Sn5 layer was measured as shown in Fig. 6. The Young s modulus of the film was almost constant at about 110 GPa, regardless of depth and location.

4 Transactions of The Japan Institute of Electronics Packaging Vol. 2, No. 1, 2009 ters are less than 1 μ m and are surrounded by weak grain boundaries, show cooperative grain boundary sliding that varies the mechanical properties of thin film materials significantly.[8, 9] Next, the Young s modulus of the grown Cu6Sn5 layer was measured as shown in Fig. 6. The Young s modulus of the film was almost constant at about 110 GPa, regardless of depth and location. This value agreed well with the results reported by X. Deng[10] and K. Tanida.[11] On the other hand, the Young s modulus of the grown Cu3Sn layer was found to vary significantly as shown in Fig. 7. The measured value varied from 50 GPa to 110 GPa at a depth Fig. 6 Indentation depth dependence of Young s modulus of a Cu6Sn5 layer grown by annealing. of 500 nm. This fluctuation was caused by the fine porous grain structure of the layer as also shown in Fig. 7. It was reported by K. Tanida[11] that the Young modulus of the grown Cu3Sn layer was constant at about 132 GPa. Since the grown layer consisted mainly of coarse particle-like grains and a lot of voids were observed in both the grown IMC layer and the remaining copper in this experiment, this micro texture must have caused the fluctuation of the measured data. These voids were Kirkendall voids and they were caused by the large difference of the diffusion constant between copper and tin during the formation of the Cu3Sn layer from the Cu6Sn5 layer. Since the diffusion constant of copper was much higher than that of tin, a lot of vacancies remained mainly around interface between Fig. 7 Indentation depth dependence of Young s modulus of a Cu3Sn layer grown by annealing. the remaining copper and the grown Cu3Sn layer. Though the grown Cu6Sn5 layer consisted of a homogeneous and densified structure and thus, the mechanical properties of the grown layer agreed well with the bulk material, the grown Cu3Sn layer was more brittle than the bulk material due to the formation of porous grains with a lot of Kirkendall voids. Furthermore, there were large fluctuations in its local mechanical properties. Thus, the growth of such a porous and brittle Cu3Sn layer degrades the quality of the bump joint structures. In addition to the change of mechanical properties of the grown IMC layers, it was found that the mechanical prop- Fig. 8 Example change of the stress-strain curve of an elec- erties of the annealed copper film also changed. Figure 8 troplated copper thin film caused by annealing. shows an example of the measured change of the stressstrain curve of the annealed film. The as-electroplated cop- properties of the electroplated copper films vary drastically per thin films showed brittle characteristics as mentioned depending on their thermal history after electroplating. above. The stress-strain curves of the annealed film Such a change should affect the long-term reliability of the became ductile as shown in this figure. This was mainly bump joint. because of the coarsening of the grains and the densification of the annealed film. The average grain size of the film changed from about 500 nm to 5 μ m when the film was annealed at 260 C for 60 min. Therefore, the mechanical Finite element analysis of the stress-strain filed in the bump jointed structure In order to discuss the effect of the IMC growth on the

5 Jeong et al.: Effect of the Formation of the Intermetallic Compounds (5/7) reliability of the bump-interconnection jointed structure, the change of the stress-strain filed in the joint structure was analyzed by a finite element method as shown in Fig. 9. Since the maximum plastic strain in solder bumps is a strong function of chip size, the fatigue life of solder joints is discussed at the center bump in Fig. 9(a). The chip size was fixed at 700 μm square. Both bump height and bump pitch were varied from 100 μm to 10 μm. The minimum element size was fixed at 0.5 μm. The total numbers of elements and nodes were about 30,000. Table 1 summarizes the material properties used in the analysis. Both the silicon and substrate were assumed to be elastic materials. The plastic deformation of both the copper and tin layers were modeled by a simple linear model. No strain-hardening was considered in this analysis. In addition, it was assumed that the IMC layer consisted of the Cu 6 Sn 5 layer only. This was because the Cu 3 Sn layer was much thinner than the Cu 6 Sn 5 layer and the average material properties of the Cu 3 Sn layer were close to those of the Cu 6 Sn 5 layer. The effect of the growth of the intermetallic compound on the stress-strain filed in the jointed bump-thin-film interconnection structure was analyzed by considering the change of the thickness of the IMC layer. The change of the residual stress in each material caused by the growth of the IMC layer was also taken into account in the analysis. The ambient temperature was varied from 40 C to 125 C to calculate the equivalent plastic strain range for estimating the fatigue life of the solder bump and copper interconnection. The commercial code, CADAS-FEX, made by Hitachi Ltd. was applied to the analysis. Figure 10 shows the effect of the thickness of the grown IMC layer on the shear stress at the interface between the IMC and the remaining copper interconnection. It was found that the shear stress increased drastically when the thickness ratio of the IMC layer in the bump structure increased from 0% to 20%. As shown in Fig. 7, since a lot of Kirkendall voids appeared during the growth of the IMC layer, this increase of the shear stress may cause delamination at the interface between the grown IMC layer and the remaining copper interconnection, and thus, deteriorate the integrity of this interface. The changes of the equivalent plastic strain range in both the tin bump and the copper interconnection are summarized in Fig. 11. The plastic strain range in the tin bump decreased markedly from about 1.8% to 0.7% when the thickness ratio of the grown IMC layer increased from 0% to 20%. This is because the grown rigid IMC layer changed the stress concentration field at the interface between the solder and metal (IMC) and it prevented the deformation of the solder near the interface. Since the plastic deformation of the solder relaxes the stress field in the copper interconnection, the decrease of the plastic deformation of Fig. 9 Finite element analysis of the effect of the growth of intermetallic compound on the stress-strain filed in the jointed bump-thin-film interconnection structure. Table 1 Materials properties used for finite element analysis. Material Young s modulus (GPa) Poisson s ratio Yield Stress (MPa) Thermal expansion coefficient (ppm/ C) Silicon Copper Tin Substrate MC Fig. 10 Effect of the grow of intermetallic compound (IMC) on the shear stress at the interface between the IMC and copper interconnections. 95

6 Transactions of The Japan Institute of Electronics Packaging Vol. 2, No. 1, 2009 Fig. 11 Effect of the grow of intermetallic compound (IMC) on the maximum equivalent plastic strain range in both the tin bump and copper interconnections. Fig. 12 Change of electrical resistance of a Sn/Cu bump caused by the growth of IMC layer. the solder increases the stress in the interconnection. Thus, the plastic strain range in the copper interconnection increased significantly from about 0.3% to 0.8%. This dramatic change indicates that the risk of fatigue cracking of the solder bump decreases with the growth of the IMC layer. This is because the plastic deformation of the bump is strictly limited by the rigid IMC layer. On the other hand, the fatigue life of the copper interconnection must severely decrease and it may dominate the life of the bump-joint structure. This result clearly indicates that the fracture mode of this bump-interconnection jointed structure varies depending on the thickness of the grown IMC layer. Thus, it is very important to control the thickness of the IMC layer during operation of products in order to design or estimate the life of the jointed structure quantitatively. the resistance of electroplated copper thin films was much higher than that of bulk copper.[12] This is because of the porous grains with porous grain boundaries of the electroplated films. Actually, the resistivity of the as-electroplated copper bumps measured by the 4-point probe method varied from Ωm to Ωm, about 1.5 times to 3 times higher than that of bulk copper. This unexpected increase, therefore, can be attributed to the growth of both the porous fine grains with porous grain boundaries and Kirkendall voids around the interface between the IMC layer and the remaining copper as shown in this figure. The fluctuation of the measured change was due to the variation of the initial micro texture of the electroplated copper and the interfacial microstructure between the grown IMC layer and the remaining copper layer. Therefore, 5. Change of electrical resistance of the bump joint Since the electronic conductivity of the IMC is lower it can be concluded that the growth of the IMC layer at the interface between a tin bump and copper interconnection also deteriorates the electrical integrity of the than that of copper and tin, the increase of the thickness of the IMC layer should increase the electrical resistance of the bump-interconnection jointed structure. Thus, the jointed structure. From this point of view, the growth of the IMC layer should be strictly limited to assure the reliability of the structure. change of the resistance was measured by controlling the annealing time of the stacked tin/copper bump structure. A four-point proving method was applied to the measurement 6. Conclusion The mechanical and electrical reliabilities of fine bumps of the resistance of the bump structure. Figure12 shows the measured change of the bump structure as a function of the thickness of the IMC layer. It was found that the resistance of the bump increased monotonically with the increase of the thickness of the IMC layer, as was estimated. The increase ratio, however, was much larger than that estimated using the resistivity of the bulk IMC material. Since the estimated increase rate was about 3 μω/μm, the measured value was about a hundred times higher than the estimated value. It has been reported that with diameters and heights on the order of dozens of microns was studied considering the growth of the intermetallic compound (IMC) at the interface between a tin bump and a copper thin-film interconnection. It was found that an increase of the thickness of the IMC changed the stress and strain fields around the interface significantly, and thus, changed the fracture mode from a fatigue crack of solder to a fatigue crack of the copper interconnection or to delamination between the IMC and the copper interconnection. This is due to the fact that the mechanical 96

7 Jeong et al.: Effect of the Formation of the Intermetallic Compounds (7/7) properties of the grown IMC differ from those of copper and tin and that a large number of Kirkendall voids appeared around the interface. In addition, the resistance of the bumps increased significantly with the increment of the IMC layer because of the growth of fine grains with porous grain boundaries and Kirkendall voids around the interface between the grown IMC layer and the remaining copper interconnection. Therefore, it is very important to minimize the growth of the IMC to assure the reliability of the bump joint structures. References [1] International Technology Roadmap for Semiconductors (2007) on [2] Rao R. Tummala, Fundamentals of Microsystems Packaging, McGraw Hill, (2001), pp [3] X. Deng, N. Chawla, K. K. Chawla and M. Koopman, Deformation behavior of (Cu, Ag)-Sn intermetallics by nanoindentation, ActaMaterialia, Vol. 52, (2004), pp [4] R. J. Fields, S. R. Low III and G. K. Lucey, Jr., The Metal Science of Joining, Tms (1992), pp [5] W. K. Choi and H. M. Lee, Effect of soldering and aging on interfacial microstructure and growth of intermetallic compounds between Sn-3.5Ag solder and Cu substrate, J. of Electronic Materials Vol. 29 (2000), pp [6] T. Y. Lee, W. J. Choi, K. N. Tu, J. W. Jang, S. M. Kuo, J. K. Lin, D. R. Frear, K. Zeng and J. K. Kivilahti, Morphology, kinetics, and thermodynamics of solid-state aging of eutectic SnPb and Pb-free solders (Sn-3.5Ag, Sn-3.8Ag-0.7Cu and Sn-0.7Cu) on Cu, J. of Materials Research, Vol. 17, (2002), pp [7] J. W. Yoon, S. W. Kim and S. B. Jung, IMC growth and shear strength of Sn-Ag-Bi-In/Au/Ni/Cu BGA joints during aging, Materials Transactions, Vol. 45, (2004), pp [8] T. Tamakawa, K. Sakutani and H. Miura, Effect of Micro Texture of Electroplated Copper Thin Films on Their Mechanical Properties, J. of the Society of Materials Science Japan, Vol. 56, No. 10, (2007), pp [9] N. Murata, K. Tamakawa, K. Suzuki and H. Miura, Fatigue Strength of Electroplated Copper Thin Films under Uni-Axial Stress, Journal of Solid Mechanics and Materials Engineering, Vol. 3, No. 3, (2009), pp [10] X. Deng, N. Chawla, K. K. Chawla and M. Koopman, Acta Materialia, Vol. 52, (2001), pp [11] K. Tanida, M. Umemoto, N. Tanaka, Y. Tomita, K. Takahashi, Micro Cu Bump Interconnection on 3D Chip Stacking Technology, Jpn. J. Appl. Phys., Part 1, Vol. 43, (2004), pp [12] W. H. The, L. T. Koh, S. M. Chen, J. Xie, C. Y. Li and P. D. Foo, Study of microstructure and resistivity evolution for electroplated copper films at near room temperature, Microelectronics Journal, Vol. 32, Issue 7, (2001), pp

Shear Strength in Solder Bump Joints for High Reliability Photodiode Packages

Materials Transactions, Vol. 44, No. 10 (2003) pp. 2163 to 2168 #2003 The Japan Institute of Metals Shear Strength in Solder Bump Joints for High Reliability Photodiode Packages Chong-Hee Yu 1, Kyung-Seob