Fabrication of arrays of (100) Cu under-bumpmetallization

Transcription

1 Fabrication of arrays of (100) Cu under-bumpmetallization for 3D IC packaging Wei-Lan Chiu 1, Chia-Ling Lu 1, Han-Wen Lin 1, Chien-Min Liu 1, Yi-Sa Huang 1, Tien-Lin Lu 1, Tao-Chi Liu 1, Hsiang-Yao Hsiao 1, Chih Chen 1, Jui-Chao Kuo 2 and King-Ning Tu 3 1 Department of Materials Science and Engineering, National Chiao Tung University, Hsinchu, Taiwan 30010, Republic of China. 2 Department of Materials Science & Engineering, National Cheng Kung University, Tainan, Taiwan, Republic of China. 3 Department of Materials Science and Engineering, University of California at Los Angeles, Los Angeles, CA, USA. Abstract Due to the thousands of microbumps on a chip for 3D ICs, the precise control of the microstructure of all the material is required. The nearly <111>-oriented nanotwinned and fine-grained Cu was electroplated on a Si wafer surface and annealed at C for 1 h, many extremely large <100>oriented Cu crystals with grain sizes ranging from 200 to 400 μm were obtained. The <111>-oriented Cu grains were transformed into super-large <100>-oriented grains after the annealing. In addition, we patterned the <111>-oriented Cu films into pad arrays of 25 to 100 μm in diameter and annealed the nanotwinned Cu pads with same conditions. An array of <100>-oriented single crystals Cu pads can be obtained. Otherwise, single-crystal nanowire growth displays a process by one-dimensional anisotropic growth, in which the growth along the axial direction is much faster than in the radial direction. This study reported here a bulk-type two-dimensional crystal growth of an array of numerous <100>-oriented single crystals of Cu on Si. The growth process in 3D IC has the potential for microbump applications packaging technology. Keywords <100>-oriented Cu; nanotwinned copper; abnormal grain growth. I. INTRODUCTION To approache the limit of Moore s law, the microelectronic industry is turning to extend the limit using threedimensional integrated circuits (3D ICs) in packaging, which focus on the merging of chip technology and packaging technology. Due to the requirement of higher I/O density in electronic devices, the diameter of flip chip bumps continuously shrinks. In a 100um diameter bump, the volume of solder is much larger than the volume of under bump metallization (UBM). Once the diameter of solder shrinks from 100 to 10μm, the volume of solder consequently decreases by 1000 times. In this condition, the volume of UBM in microbumps became larger than the volume of solder. During reflowing and electric current flowing, the solder would transform full intermetallic compounds. It has been reported that sputtering control the orientation of grain and electroplating by pulse or dc could control grain size and structure. Inculding of fine-grained or nanotwins could increase the mechanism of Cu. Nanotwins could increase the strain hardening capability and be able to combine the high tensile strength and ductility in mechanism. Because of twins with nanoscale lamellar spacings and coherent twin boundaries (TBs), the nano-twinned structure allows room for the dislocation multiplication and storage to continue to large plastic strains. Nanotwins had fabricated by electroplated by pulse in Cu film; they also showed prefer oriented (111) of nanotwinned Cu (nt-cu) have higher hardness and strength than regular Cu.[1-2] But nanotwins are unstable in thermal treatment, they are easy to be grain growth and transfer into another orientation. Controlling grain starts from crystal growth, which is one of the most important topics in multi-disciplinary sciences because the crystal growth controls the properties of semiconductors, metals and ceramics in thermal processing.[3-10] Several researchers have reported that giant grains up to several hundred micrometers can be grown in Al, Ag and Cu thin films.[11-15] However, it is unclear about the grain crystallization in nanotwinned Cu. This study is a new type of abnormal grain growth that results from the extremely anisotropic two dimensional crystal growth of Cu via nearly unidirectionally oriented nano-twins; the growth in the vertical direction is much slower than that in the lateral directions. II. EXPERIMENTAL This experimental that included columnar nanotwins in Cu film and different diameter about Cu pads. First, the nt-cu samples are divided into the pad and film. TiW of 200 nm thickness and Cu of 200 nm thickness were sequentially sputtered onto Si substrate. Ti was used as an adhesive layer, and Cu was as a seed layer. The fabrication process of nt-cu pads induced one photolithographic procedures. The size was the diameter of the Cu microbump, which varied from 25 to 100 μm were deposited by electroplating on the patterned substrate. The nt-cu films were 8 μm in thickness. The stirring rate to grow the oriented nt-cu was 600 r.p.m., and the current density was 50mAcm 2. The duty cycles were T on=0.02 s and T off=1.5 s. The deposition rate was 1.2 nm s 1 under this electroplating condition..the electroplating solution of Cu is composed of CuSO 4 5H 2O (196 g/l), HCl (1 ml/l), additive (3 ml/l), and deionized water. For the preparation of the equivalent Cu layer without the nanotwins, we used the same electroplating solution at 200mA cm 2. In thermal treatment research, the Cu films were subjected to annealing at temperatures ranging from 400 to 500 C for extended periods in vacuum oven at torr. All the electroplating and reflowing processes were under air condition. Using scanning electron microscope (SEM) imagined microstructural features and focused ion beam (FIB) polished cross-sections. X-ray diffraction (XRD) was used to analyze the preferred texture of the Cu films. Electron Probe X-ray Micro-Analysis (EPMA) were used to study the 518

2 individual grain orientation in the Cu film and pads. III. RESULTS AND DISCUSSION Preferred orientation of Cu could be fabricated by pulse electroplating. The spacing of the nanotwins ranged from 10 nm to 100 nm. We fabricated highly oriented <111> nt-cu using oriented-<111> Cu seed layer, likes as Figure 1a and 1b.They show the plan-view of the inverse pole figure map from EBSD. Fig. 1c and 1d show the Cu film has extremely high <111> texture and the average 6.4 misorientation in the statistical misalignment angles of the Cu(111) grains. After annealing at 450 C, we obtained exceptionally large grains about μm diameters and about ~ 65 times larger than the surrounding grains which was called an abnormal grain; see in Fig. 2a. Fig. 2b shows large grain merged small grains in an enlargement and we observed two peaks of {111} and {200} diffraction for various times by X-ray diffraction spectra in Fig. 2c. To convert the abnormal grain growth to solid-state crystal growth, we patterned an array of nt-cu bumps with diameters ranging from 100 to 25 μm and thicknesses of 10 μm using a lithographic technique. Figs. 3a and b present cross-sectional EBSD and FIB images, respectively. There is an anisotropic growth of a <100> grain annealed at 350 C for 30 min. Figs. 3c and d show a <100> grain had observed by annealed at 400 C for 10min. Fig. 3e shows <100> grains grown to 170 and 207 μm, respectively, after annealing at 400 C for 20 min. It is the most significant anisotropic growth in this condition. As-fabricated columnar nt-cu is showed by Fig. 4a. We annealed at 400 C for 20min and polished the cross-section by Figs. 4b and c. The <100> grain on the left-hand side grew closely to 290 μm. Figs. 4b shows the plan-view of the inverse pole figure map from EBSD and indicated oriented- <111> transferred into <100> grains and grain size is distributed for the grains in d. The anisotropic grain growth has a very slow growth in the vertical thickness of 5 μm, which is at least 50 times slower than the lateral growth. In the <100>-oriented grains, no nanotwins are detected. Furthermore, the anisotropic grain growth started from the bottom interface rather than from the free surface. We produced an array of pads with diameters of 75, 50 and 25 μm as-fabrication, as shown in Figs. 5a-c. This figure demonstrates the precise control of growth of an array of single-crystal Cu microbumps. Figs. 5d-f were completely transformed into a <100>-oriented single-crystal of Cu, and the other three were very near complete transformation. The 75-μm Cu pads in Fig. 5d occupy ~ 55% of the surface area. and the 25-μm Cu pads in Figure 5f occupy only ~8% of the surface area. There are some electron-charging in the edge of Cu pads and more severely occur in the sample with 25-μm Cu pads. However, the anisotropic growth is more clearly observed in different diameter of Cu pads. Fig. 1 (a) Plan-view of the inverse pole figure map from for the <111>-oriented surface grains with color coding of the inverse pole figure of the as-deposited Cu film. (b) The inverse pole figure of the <111>-oriented Cu film. (c) X-ray diffraction for the as-deposited Cu film displays orientations. (d) The statistical results showing the numerical fraction of grains as a function of the misalignment angle from the [111] direction. Fig. 2. (a) Plan-view inverse pole figure map of the large <100> grain obtained from electron backscattered diffraction in two-dimensional anisotropic crystal growth in <111> nt-cu films. (b) Enlarged inverse pole figure map for the dashed rectangle. (c) Evolution of the X-ray diffraction intensities for the samples annealed at 450 C for various times. 519

3 Fig. 3 (a) Cross-sectional map of the inverse pole figure from EBSD for the extremely anisotropic growth of the <100> Cu grains annealed at 350 C for 30 min. (b) The corresponding cross-sectional FIB image. (c) Cross-sectional map of the inverse pole figure from EBSD for the sample annealed at 400 C for 10 min. (d) The corresponding cross-sectional FIB image for the sample in c. (e) Cross-sectional FIB images for the sample after annealing at 400 C for 20 min. Fig. 4. (a) Cross-sectional FIB image of the as-fabricated nt-cu sample. (b) Cross-sectional inverse pole figure map of the large <100> crystal with a lateral crystal size of 290 μm that showed extremely anisotropic growth of the <100> Cu crystals. (c) Crosssectional FIB image of the sample in b. (d) Plan-view of the inverse pole figure map from electron backscattered diffraction for the sample annealed at 450 C for 25min. (e) Grain size distribution for the grains in d. Furthermore, we prepared randomly oriented nanotwins to compare the <111>-oriented nanotwins in achieving the grain growth of <100> crystals. Fig. 6a presents a FIB crosssectional view of the randomly oriented nanotwins in Cu, and Fig. 6b presents the corresponding X-ray diffraction spectrum, where show the {111}, {200} and {220} reflections. The normal grain growth occurred after annealing at 450 C for 30 min,. Figs. 6c and d present the FIB cross-sectional image and the corresponding X-ray diffraction spectrum, respectively. In addition, an inverse pole figure map of the top view of the grains obtained from EBSD in Fig. 6e reveals no extremely anisotropic grains. Fig. 6f presents the average grain size is 6.6 μm.by the grain size distribution. By the results, we could define the necessary and sufficient conditions to obtain extremely anisotropic crystal growth below. The first condition is the presence of a seeding layer with a low density of <100>-oriented seeds. So we prepared a seeding layer with a very high density of <111>-oriented 520

4 Fig. 5 Different diameter of arrays of <111> nt-cu are (a) 75-μm diameter, (b) 50-μm diameter and (c) 25-μm diameter. After annealing at 450 C for 60 min, the <111> Cu crystals transformed into <100> single crystals; like as (d) 75-μm diameter, (e) 50-μm diameter and (f) 25-μm diameter. seeds, we could obtain a very high intensity of <111>oriented nt-cu; however, upon annealing, we could hardly detect the growth of <100> grains because there were no <100>-oriented seeds. The second condition is a high density of <111>-oriented nanotins, which provide the stored energy to serve as the driving force in the growth of the anisotropic and untwinned {100} single crystals. Randomly oriented nanotwins will not do serve as this driving force. The third condition is that the substrate should have a large difference in thermal expansion coefficient from that of the metal film, such that a biaxial strain in the in-plane directions of the film occurs under high-temperature annealing. The abnormal grain growth in Cu has been explained by strain relaxation. Strain will clearly have a role in this study of extremely abnormal grain growth. In addition to the biaxial thermal strain mentioned above, it has been reported that in pulse-electroplating nt-cu, in situ strain measurement using the bending beam method reveals that the nt-cu film is under in-plane biaxial tensile strain.[16] We will not attempt to explain the mechanism of abnormal grain growth here because what we have is the microstructural evolution of an elastically anisotropic inclusion (a two-dimensional grain) within an elastically anisotropic matrix (the oriented nt-cu). We have applied this anisotropic growth to produce an array of numerous <100>-oriented single crystals of Cu microbumps with sizes from 100 to 25 μm on Si wafer surfaces. An extremely abnormal grain growth of <100>- oriented single crystals of Cu in a matrix of <111>-oriented and nt-cu columnar grains had formed; the lateral growth is approximately two orders of magnitude faster than the vertical growth. Fig. 6 (a) Cross-sectional FIB image for the as-fabricated Cu film at 200mA cm 2. (b) The corresponding XRD spectrum for the Cu film in a. (c) Cross-sectional FIB image for the Cu film after annealing at 450 C for 30min and (d) the corresponding XRD spectrum for the Cu film in c. (e) Plan-view of the inverse pole figure map from EBSD for the Cu grains annealed at 450 C for 30min. (f) Grain size distribution for the grains in e. The average grain size is about 6.6 μm, which is approximately the thickness of the Cu film. ACKNOWLEDGEMENTS Financial support from the Ministry of Science and Technology, Taiwan under the contracts E MY3 is acknowledged. We would also like to thank the Center for Micro/Nano Science and Technology at National Cheng Kung University for assistance with the analytical equipment. REFERENCES [1] H. Y. Hsiao, C. M. Liu, H. W. Lin, T. C. Liu, C. L. Lu, Y. S. Huang, C.Chen, and K. N. Tu, Unidirectional growth of microbumps on (111)-oriented andnanotwinned copper, Science, vol. 336, pp , (2012). [2] G. T. Lui, D.Chen, and J. C. Kuo, EBSD characterization of twinned copper using pulsed electrodeposition, J. Phys. D. Appl. Phys., vol. 42, pp , (2009). [3] C. V. Thompson, Grain growth and evolution of other cellular structures, Solid state physics. Adv. Res. Appl., vol 55, pp , (2001). [4] S. Schmidt, S. F. Nielsen, C. Gundlach, L. Margulies, X. Huang, and D. J. Jensen, Watching the growth of bulk grains during recrystallization of deformed metas. Science, vol. 305, pp , (2004). [5] T. J. Rupert, D. S. Gianola, Y. Gan and K. J. Hemker, Experimental observations of stress-driven grain boundary migration. Science, vol. 326, pp , (2009). [6] R. D. Flack, Fundamentals of Jet Propulsion with Applications Ch. 8 (Cambridge Univ.Press: New York, NY, USA, 2005). [7] J. E. Burke and D. Turnbull, Recrystallization and grain growth. Prog. Metal Phys. vol. 3, pp , (1952). [8] F. J. Humphreys and M. Hatherly, Recrystallization and Related Annealing Phenomena (Elsevier: Oxford, UK, 2004). 521

5 [9] E. A. Holm and S. M. Foiles, How grain growth stops: a mechanism for grain-growth stagnation in pure materias. Science, vol. 328, pp , (2010). [10] M. Hillert, On the theory of normal and abnormal grain growth. Acta Metall. Vol. 13, pp , (1956) [11] A. M.Gusak, and K. N. Tu, Theory of normal grain growth in normalized size space, Acta Mater., vol 51, pp , (2003). [12] J. Greiser, P. Mullner, and E.Arzt, Abnormal growth of giant grains in silver thin films, Acta Mater., vol49, pp , (2001). [13] A. Gangulee and F. M. D'Heurle, Anomalous large grains in alloyed aluminum films I. Secondary grain growth in aluminum-copper films. Thin Solid Films, vol. 12, pp , (1972). [14] A. Gangulee and F. M. D'Heurle, Anomalous large grains in alloyed aluminum films II. Electromigration and diffusion in films with very big grains. Thin Solid Films vol. 16, pp , (1973). [15] J. Koike, M. Wada, M. Sanada and K. Maruyama, Effects of crystallographic texture on stress-migration resistance in copper thin films. Appl. Phys. Lett. vol. 81, pp , (2002). [16] D. Xu, V. Sriram, V. Ozolins, J. M. Yang, K. N. Tu, G. R. Stafford and C. Beauchamp, In situ measurements of stress evolution for nanotwin formation during pulse electrodeposition of copper. J. Appl. Phys. vol. 105, pp , (2009). 522