Characterization of Cold Drawn Gold bonding Wire with EBSD

Transcription

1 499 Material Science Forum Vols (2002) pp Trans Tech Publications, Switzerland Characterization of Cold Drawn Gold bonding Wire with EBSD Jae-Hyung Cho 1,2, J.S. Cho 3, J.T. Moon 3, J. Lee 3, Y.H. Cho 4, A.D. Rollett 1 and K.H. Oh 2 1 Carnegie Mellon University, Pittsburgh, PA Seoul National University, Seoul Korea MKE Electronics, Pogok-Myeon, Yongin-Si, Kyunggi-Do, Korea 4 Korea Institute of Science and Technology, Seoul, Korea Keyword: Gold bonding wire, Grain Boundary, Misorientation, EBSD Abstract. Cold drawn gold bonding wires have been investigated with Electron Back Scatter Diffraction (EBSD). The textures of drawn gold wires contain major <111>, minor <100> and small fractions of complex fiber components. The <100> oriented regions are located in the center and surface of the wire, and the complex fiber component regions are located near the surface. The <111> oriented regions occur throughout the wire and have large Taylor factors and would be expected to have higher stored energy as a result of plastic deformation compared to the <100> regions. Large misorientations (angles > 40 ) are located between the <111> and <100> regions, which means that the boundaries between them are likely to have high mobility. Boundaries within the <111> regions are predominantly <111> tilt grain boundaries with large misorientations, similarly, the <100> regions have <100> tilt grain boundaries with smaller misorientations. It appears that the stored energy as indicated by geometrically necessary dislocation content in the subgrain structure is similar in all orientations despite the large differences in Taylor factor. Introduction Fine wire of pure Au, Cu or Al, is used for interconnection in semiconductor packaging. Recently, as packaging technology continues to advance, the improved properties of bonding wire are needed. In particular, the ball shape, breaking load, elongation and the homogeneity of microtexture and microstructure of wires are important for bonding wire. These characteristics are related to the purity of the original materials, the drawing process and the annealing process. High purity gold ( %wt Au) is too soft and unstable for obtaining good properties for bonding when it is drawn and annealed. Generally, the annealing and recrystallization temperature for pure gold is in the range 150~200 C and it has been reported that highly deformed pure gold will show recovery and recrystallization at room temperature [1]. Therefore bonding wire commonly has various dopants at the parts per million (ppm) level in order to obtain better thermal and mechanical properties. Impurities, even at these low levels are important in controlling the final mechanical properties and microstructures of gold wire by raising the recrystallization temperature and preventing the grain growth [2, 3]. Such small concentrations of impurities are known to strongly affect the migration rate of grain boundaries in many materials [4, 5]. Recrystallization, recovery and grain growth all occur during annealing, and they affect the microstructures, microtextures and mechanical properties of gold wires [6~9]. The wire drawing textures of fcc metals typically have <111> and <100> fiber texture components. The textures of aluminium, copper and brass wires have been investigated for cyclic symmetry [10]. The textures of drawn silver wires have radial symmetry, which is related to twinning [11]. Near the surface, recrystallized grains are formed, due to frictional heating in the die. Heizmann et al. have reported that the strength of the cyclic texture increases as the die angle increases [12]. Recrystallization textures in drawn wires are also <111> and <100> fibers, which are similar to the deformation textures. The ratio of the <111> to the <100> varies with the annealing time and temperature.

2 500 Texture of Materials Previous research on gold bonding wires has focused on their mechanical properties and recrystallization behavior. In this research, the microtexture and microstructure of drawn gold wire are investigated with SEM/EBSD. In order to understand the grain boundary characteristics, the misorientation distribution function (MDF) and CSL boundaries are calculated from EBSD data. Experimental Material and Sample Preparation. The purity of gold wire used in this research is more than 99.99% and it has some (intentional) dopants, such as Ca and Be that total less than 50 ppm by weight. A typical recrystallization temperature of this gold is 320 C. The original cast gold bar was drawn through a series of diamond dies to a von Mises equivalent strain of EBSD Measurement. The bonding wire was mounted in epoxy and then sectioned and polished. The polished specimens were cleaned with ion milling. EBSD maps were measured of cross sections. For generating orientation and grain boundary maps, a grain tolerance angle of 15 was used, which distinguishes the low angle and high angle grain boundaries. The orientations of grains are shown in inverse pole figure map, and the misorientation distribution function (MDF) is used to characterize the grain boundary characteristics. Taylor factor maps are also shown because they clearly reveal the fiber texture components. The Taylor factor is calculated based on the standard slip systems for fcc metals and velocity gradient appropriate to uniaxial extension [13, 14]. {111}<110> slip systems are assumed and the velocity gradient is as follows, ε ij = (1) Statistical Analysis of Microstructures. The fluctuations of material microstructures can be described by so-called second-order characteristics such as the variance of the volume of a microsturctural component or phase. If f (x) and f (y) are the probabilities of random variables, x and y, the covariance of the pair ( f ( x), f ( y)) is given, cov ( f ( x), f ( y)) = Ε[( f ( x) Εf ( x))( f ( y) Εf ( y))]. (2) The normalized covariance function is called the correlation function, and takes values between 1 and +1. ρ = cov ( f ( x), f ( y)) f ( x) f ( y) σ σ. (3) f ( x) f ( y) where, σ f (x) and σ f (y) are the standard deviations of f (x) and f (y), respectively. A value of 1.0 or +1.0 indicates perfect linear prediction between x and y, whereas a value of zero indicates no linear predictive value [15]. Results

3 Materials Science Forum Vols <111> <100> All a) <High> <Low> All b) c) d) e) <111> <100> complex f) g) h) Figure 1. Image for a cold drawn gold bonding wire (30µm). a) Image maps for <111>+<100> regions b) Image maps for <high>+<low> Taylor factor c) Complex regions d) Schematic plot for Structure of wire e) Image quality f) <111> region inverse pole figure g) <100> region inverse pole figure h) complex region inverse pole figure Transverse Section of Wire (30mm); As Drawn. In order to analyze the fiber texture of drawn wire, cross sections of wires were measured with EBSD. The main fiber components observed during drawing are <111>//ND and <100>//ND. These two texture components are known as the typical fibers of fcc wires. A set of EBSD maps of the wire as drawn is shown in fig 1. Each grain in the wire can be partitioned into two types by their Taylor factors (TF) or by their crystallographic orientation, i.e. <111> or <100>//ND. Using a tolerance angle of 15 (misorientation), the <111> and <100> fiber regions are separated in fig 1-a; Taylor factors show a very similar separation of high and low value regions in fig 1-b. The average Taylor factor of cold drawn wire was calculated using the standard set 12 slip systems for fcc metals, and its value was found to be Regions with a Taylor factor lower than 2.87 are predominantly <100> oriented, whereas regions with Taylor factor greater than 2.87 are predominantly <111>. As expected, the images partitioned by orientation or by Taylor factor are very similar to each other. The region between the center and surface regions has complex structure and orientations, which are deviated from <111> and <100>

4 502 Texture of Materials a) b) Figure 2. Misorientation distribution of the drawn wire. a) <111> regions b) <100> regions R3 value ( 2-1, 2-1,0) 17b 1/4~1/3 1/5 1/8~1/6 1/11~1/9 0 <111> regions 19b 7 37c 33c 3 25b 11 GBs between <111> and <100> 9 39a 13b 21a 31a (0,0,0) 41a 25a 37a 13a 17a <100> regions 5 29a ( 2-1,0,0) Figure 3. Projected CSL grain boundaries in the Rodrigues space in the <111>, <100> regions and between them. by more than 15. Fig 1-c shows this difference by complex grains that lie more than 15 from either <100> or <111>. Fig 1-d shows the overall structure of the wire after the drawing process. Most of the wire is oriented parallel to <111>, whereas the center and some parts of the surface have a <100> fiber component. The complex regions are located under the surface. The image quality map of the transverse section of the cold drawn wire is shown in fig 1-e and suggests that <100> regions are associated with high image quality (IQ). Using the orientation, a Taylor factor map was calculated, and then the image quality map was combined as shown in equation 3. The resulting value was 0.22, which is a mild negative correlation. This suggests that high Taylor factor regions have low pattern quality and low Taylor factor regions have high pattern quality in keeping with the qualitative observation made previously. Misorientations in the As Drawn Wire. Misorientation distributions in Rodrigues-Frank space based on the partitioning shown in fig 1-a are given in fig 2. Grain boundaries in the <100> regions have lower misorientation angles than in the <111> regions. These maps show that the misorientation distributions of <111> fibers, as expected, are concentrated on <111> misorientation axes and their Rodrigues vector components, (R1, R2, R3) take values from ( 0, 0, 0) to

5 Materials Science Forum Vols (,, ). The latter Rodrigues vector is equivalent to a 60 <111> misorientation angle/axis pair The length of the Rodrigues vector is equal to the tangent of half the misorientation angle, therefore <111> regions have a large range of misorientation angles, i.e. from 0 to 60 around the <111> misorientation axis [16]. By contrast, the <100> regions are concentrated on <100> misorientation axes and their Rodrigues vectors are located between ( 0, 0, 0) and ( 2 1, 0, 0). The latter Rodrigues vector is equivalent to a 45 <100> misorientation. Grain boundaries in the <100> regions have lower misorientation angles than in the <111> regions. Fig 3 shows a projection of the fundamental zone in Rodrigues space and the location of most of the low-sigma Coincident Site Lattice (CSL) boundary types. In this fig, the < 111> axis falls on top of the <110> axis along the hypotenuse of the triangle. The <100> axis projects along the lower, horizontal edge. CSLs along the <111> axis, fig 3, with Σ3, 7, 13b, 21a and 31a are found frequently in the <111> fiber regions. The most frequent CSLs in the <100> component are Σ5, 13a, 17a, 25a and 29a on the <100> axis. Almost all of the boundaries between the <111> and <100> components have misorientation angles above 40 so that the CSLs are mainly Σ3, 9, 11, 17b, 25b, 31b and 33c. The misorientation angle distribution for cold drawn gold wire is shown in fig 4. Three different distributions are plotted separately based on each of the fiber regions, i.e. <111>, <100> and intermediate orientations. All three exhibit non-random distributions. Most misorientation angles in the <100> regions are less than 40, whereas the <111> regions exhibit angles up to 60. High misorientations predominate for boundaries between the <111> and <100> regions. The peak in the misorientation distribution for boundaries between the <111> and <100> regions is located between 45 and 60. The large misorientation angles of boundaries between the <111> and <100> regions mean that these boundaries will tend to have higher energy and, possibly, mobility than the average boundary in this system <100> fibers <111> fibers Bet. <111> and<100> Fraction number Misorientation angle ( o ) Figure 4. Misorientation angle distribution of cold drawn wire. Conclusion In this study, cold drawn gold bonding wires have been investigated with EBSD. 1. The cold drawn bonding wire has a major <111> fiber component and a minor <100> component. The <100> orientated grains are located in the center and the surface regions.

6 504 Texture of Materials 2. There is a weak correlation between image quality and misorientation in the cold drawn wire, and it shows lower stored dislocation density in <100> than <111> grains statistically. 3. The misorientation angle distributions show that grain boundaries within the <111> fiber have larger misorientation angles than in the <100> component. 4. Other than low angle boundaries, CSL boundaries in the <111> regions are predominantly of <111> axis type. Similarly the CSL boundaries in the <100> regions are of the <100> misorientation axis type. The CSLs between <111> and <100> have large misorientation angles greater than 40. Acknowledgement This research is supported by BK21 project of Ministry of Education & Human Resources Development in Korea and MKE Electronics. Partial support of the Mesoscale Interface Mapping Project at Carnegie Mellon University under NSF grant No is also acknowledged. References [1] T.H. Ramsey: Solid State Technology, Vol. 16 (1973), pp [2] S.Omiyama, Y.Fukui: Gold Bulletin, Vol. 15 (1980), p. 43. [3] B.L. Gehman: Solid State Technology, Vol. 23 (1980), pp [4] K. T. Aust and J. W. Rutter: Trans, TMS-AIME, Vol. 215 (1959), pp [5] E.M. Fridman, C.V. Kopezky and L.S. Shvindlerman: Z. Metallkunde, Vol. 66 (1975), p [6] K.Busch, H.U.Kunzi, B.Ilschner: Scripta Metall., Vol. 22 (1988), pp [7] K. Hausmann, B. Ilschner, H.U. Kunzi: DVS Berichte, Vol. 102 (1986), pp [8] R. Hofbeck, K. Hausmann, B. Ilschner, H.U. Könzi: Scripta Metall., Vol. 20 (1986), p [9] G. Qi and S. Zhang: Journal of Materials Processing Technology, Vol. 68 (1997), p [10] G. Linßen, H.D. Mengelberg, H.P. Stüwe: Z. Metallkunde, Vol. 55 (1964), pp [11] E. Aernouldt, I. Kokubo, H.P. Stüwe: Z. Metallkunde, Vol. 57 (1966), pp [12] J.J. Heizmann, C. Laruelle, A. Vadon, and A.Abdellaoui: ICOTOM , p [13] OIM, Software for analysis of electron backscatter diffraction patterns, User manual, TSL, [14] S.I. Wright, B.L. Adams and K. Kunze: Mat. Sci. Eng., Vol. A160 (1993), pp [15] J.Ohser and F. Müchlich: Statistical analysis of Microstructures in Materials Science, John Wiley & Sons, Ltd., 2000, Chap 5. [16] V. Randle: The measurement of Grain Boundary Geometry, Institute of Physics publishing, 1993, Chap [17] A. Heinz and P. Neumann: Acta Cryst., Vol. A47 (1991), pp