.loumlflof Materials Proce ssh g Technology ELSEVIER Journal of Materials Processing Technology 68 (1997) 288-293 Recrystallization of gold alloys for producing fine bonding wires Guojun Qi, Sam Zhang * Gintic Institute of Manufacturing Technology, Nanyang Technological University, 71 Nanyang Drive, Singapore 638075, Singapore Received 28 December 1995 Abstract RecrystaUization maps correlating deformation, temperature and hardness have been constructed for two types of gold alloys for producing 'hard' and 'medium-hard' fine bonding wires. Recrystallization initiation temperatures (TJ and recrystallization temperatures (Tr) were inferred from the maps. The values of Ti and Tr of the alloy for the 'hard' wire are about 80-100 K greater than those for the 'medium-hard' wire, which was attributed to the different dopants present in the materials. However, when fully annealed, the two types of gold materials possess similar hardnesses of between 35 to 40 HV or HK. The recrystallization map can serve as a general guide in choosing appropriate annealing conditions in gold bonding wire manufacturing processes. © 1997 Elsevier Science S.A. Keywords: Gold; Gold alloy; Annealing; Recrystallization; Bonding wire 1. Introduction Fine wire bonding is still the most popular inter-connection method used in packaging electronic devices, despite extensive research and development efforts on new processes such as tape bonding, flip chip mounting and beam lead joining. Wire bonding machines of increasingly greater speed and complexity, plus the general trend ef downsizing in electronic packages which favors bonding patterns with finer pitches, demand improved ,'operties of gold bonding wires in terms of yield strength, homogeneity and metallurgical integrity during application. The properties of the gold wires are inheri~ed from the raw material and acquired in the productioa processes. Extremely pure gold ( > 99.999%wt Au) is simply too soft and unstable for successful wire drawing or wire bonding. Therefore, it is a common practice to add various dopants at ppm level (by weight) to high purity gold to cater for different thermo-mechanical properties. Variation iv_ the chem* Co,, responding author. Tel.: + 65 7991336; fax: + 65 7922779; e-mad: szhang@gintic.gov.sg in terms of tensile strength, or breaking load as commonly used in the wire making industry, a "hard' wire of 25.4 lam in diameter would, when annealed to around 5% elongation at rupture, have a breaking load of 9.5-12 g, a 'medium-hard" wire 8-9.5 g and a 'soft' wire below 8 g. 0924-0136/97/$I 7.00 © 1997 Elsevier Science S.A. All rights reserved. PII S0924-0136(96)00034-9 istry of a material leads to various types of commercial bonding wires: 'hard', 'medium-hard' or 'soft' wires ~. The user can choose the wire type according to his/her package design requirements on loop height, loop length, bond pull strength, etc. Although the detailed working mechanism is not fully understood, it is known that dopants can strengthen gold wires by raising the recrystallization temperature and restraining grain growth [1-3]. The production of fine bonding wires is basically a cold drawing process. In conventional production, properly refined and doped gold is cast into cylindrical bars, which are subsequently rolled or swaged into wires of millimeters in diameter. The wires are then cold drawn through various dies to final sizes of around 25 microns diameter. Recently continuous-casting techniques have found application in replacing the casting and rolling/swaging processes so that the material can be drawn directly after casting. When the wire is drawn, energy is stored in the metal as a result of cold-working and the wire becomes harder and more brittle as the deformation proceeds. Therefore the drawing process must be stopped at particular stages to release the accumulated energy and homogenize its microstructure through recovery and annealing. These heat-treatment operations have been proven to play a very important, if not the decisive, role in achieving good drawability of G. Qi, S. Zlzang /Journal qf Materials Processing Technology 68 (I997)288-293 the material, desired mechanical properties and their homogeneity along the wire. Insufficient annealing will produce too hard a material for the subsequent drawing process, whilst over-annealing will result in a non-homogeneous microstructure due to grain growth. Such improper heat treatment has been identified as one of the culprits responsible for some malfunctioning of bonding wires during bonding operations (G.J. Qi, Priv. coramun., 1993). However, there is little published work on the annealing of gold materials in the intermediate stages of wire drawing. Tomiyama et al. [2] reported temperature ranges in which different gold materials with 99% cold work may recrystallize. Busch et al. [4] studied the annealing and recrystallization kinetics of bonding wires using modulus and resistivity measurements. Some review papers [5,6] and handbooks [7] give general rules of thumb that, e.g., gold recrystallizes in the range of 423-473 K or that the annealing temperature for gold is 573 K. Apparently the information from these sources refers either to final products only or is too inaccurate for practical use. This work is aimed at understanding the behavior of gold materials during annealing at intermediate drawing stages through the construction of recrystallization maps for two representative types of gold alloys. The results of the study can serve as a general guide in choosing appropriate heattreatment conditions in production processes. A recrystaltization map is a three-dimensional diagram that reveals the relationship between deformation (strain), annealing temperature and one of the mechanical properties, e.g., hardness or tensile strength. From this map, the correlation of recrystallization temperature vs. deformation can be inferred and the mechanical behavior after recrystailization can be better understood. Amongst the mechanical properties of metallic materials, hardness is perhaps the easiest to measure and is, in fact, used in practical production as one of the SPC (Statistical Process Control) factors. As such, in this study, microhardness (HV and HK) was measured with respect to deformation and annealing temperature. 2. Experimental 2.1. Material and sample preparation Two types of gold alloys for producing the so-called 'hard' and 'medium-hard' bonding wires were chosen for the study, these being pure gold (99.999%) doped with different amount and combination of Be, Ca, etc., all in ppm level by weight. The samples were obtained from a wire drawing production line. The initial materials were fully annealed ~ 3 mm wires. Plastic deformation expressed by the percentage 289 decrease in cross-sectional area was used as an index of cold work (CW), i.e.: %CW = Plastic deformation = ( A o - AO/A(, x 100 = (4~,,- ~bf)/~,, x l o o where Ao and ~/~ arz the original cross-sectional area and diameter of the wire, respectively, and Af and 4)t the same measurements of the wire after drawing. Nine levels of deformation from the initial size were selected (%CW = 25.7, 46.6, 61.7, 72.5, 80.3, 85.8, 89.8, 92.4, and 93.3), which were determined by the sizes of the drawing dies. Samples of about 5--6 cm length were taken at the desired drawing stages and annealed immediately after drawing to avoid possible recovery effects at room temperature. 2.2. Annealing The recrystallization temperature refers to an approximate temperature at which a highly cold-worked material completely recrystallizes in I h [8]. Accordingly, the annealing time for all the samples was fixed at 1 h in this work. The annealing was carried out in a horizontal oven with a constant temperature zone of about i5 cm at various temperatures from 573 to 673 K. At each temperature nine pieces of samples, one from each deformation level, were bound together with a piece of thin gold wire and placed in the middle of the hot zone. Temperature was measured with a calibrated K-type thermocouple, the tip of which was put into contact with the samples to ensure accurate temperature measurement. The temperature fluctuation was controlled to within + 1 K for an entire expep.'menta! run. Preliminary experhnents verified that it took about 2 min for the samples to reach within 5 K below the equilibrium temperature. Timing started, therefore, after the samples had been, placed in the hot zone for 2 rain. The samples were quenched in water after annealing for 1 h. 2.3. Hardness measurement and metatlography Ideally, hardness measurements should be made on a fiat surface. In the SPC practice in the wire drawing industry, however, this is usually done on the round surface of a wire for convenience of operation and to avoid possible errors introduced by work hardening if a fiat surface is prepared by grinding and polishing. To obtain information directly comparable to practical operations, the microhardness (HV and HK) was measured on the round surfaces of the annealed samples using a Matsuzawa Seiki microhardness tester, Model DMH-2. The load was 100 gf (0.981 N) and the loading time was 10 s in the measurements. Each datum point was the average of about ten measurements. In the case 290 G. Qi, S. Zhang /Journal of Material,, Processing Technology 68 (1997) 288-293 of Knoop hardness, care was taken to ensure that the longer diagonal of the indenter was parallel to and or, top of the wire. Systematic errors arising from the round surface effect were estimated to be less than 3% of the measured values in HV measurements and ew:n less in HK measurements. Selected samples were p:epared metallographically and etched to reveal the mierostructure. The microhardness is plotted against annealing temperature at a constant deformation, as shown in Fig. l(a) and l(b), which results are typical of the experiment data. A combination of such plots for all of the deformations gives rise to the so-called recrystallization map, as shown in Fig. 2 for the 'hard' gold alloy and in Fig. 3 for the 'medium-hard' gold alloy, respectively. Both the Knoop and the Vickers hardness were taken for all of the samples. It was noted that the hardness values from the two measurement techniques were very 80 RecrystaSzabon 70 60 q) 50 Reoovep] Grain Growth 40 Ti 30 (a) bS0 = -7. , i i ! 570 590 610 630 Tr , , 650 670 ,590 0 701 o "o. 401~ 3, Results and discussion 2 =S . s0t _8 • 20 0 70 70 / 0 3o 2O 563 583 603 623 643 663 683 Annealing Temperature (g) Fig. 2. Recrystallization map for the 'hard' gold alloy, showing the correlation amongst microhardness (HV), temperature and deformation (annealing time: I h). close to each other, with a hardness number range of 30-80. An annealing process may be divided into three stages [8]: recovery, recrystallization and grain growth. Recovery is primarily a low temperature process and property changes do not cause an appreciable change in microstructure. When the temperature is sufficiently high to provide the necessary energy for atoms to overcome the rigidity of the distorted lattice and rearrange themselves to form strain-free grains, recrystallization starts. As the temperature increases, the rigidity of the lattice decreases and small grains combine to form larger grains as, thermodynamically, larger grains have lower free energy. The limit of the grain growth at a particular temperature is the equilibrium between the thermodynamic driving force and the lattice rigidity. The three stages of annealing can be identified easily for the gold alloys. As illustrated in Fig. 1, for every 80 Recfystallization fa 7o 5O o2 :S Recovery 30 b50 (b) G rain G rowth 40 , 570 4) Ti , , i 590 610 ! 630 , 'Tr, 650 I 670 690 Temperature (K) '~' 30 I" 473 Fig. 1. Microhardness (HV) of tile 'hard' gold alloy annealed for i h at different temperatures (cold work: 80.3%). Indicated in the plot are three regions (recovery, recrystallization, and grain growth) and determination of reerystallization starting (Ti) and finishing (T~) temperatures. - • , I 523 , • - • ' . . . . S73 , 623 Annealing Temperature (K) Fig. 3. Recrystallization map for the 'medium-hard' gold alloy, showing the correlation amongst microhardness (HV), temperature and deformation (annealing time: 1 h). G. Qi, S. Zhang / Journal ~f Mate~ial.s Proees~'#lg Teclmoh~gy 68 (1997)288-293 291 Fig. 4. SEM micrographs of the metallographically prepared samples, showing typical microstructures at various annealing stages: (a) during the recovery period; (b) at the start of recrystallization: 4c1 at the completion of recrystallization: and (d) during grain growth. particular deformation there exists a temperature at which the hardness begins to drop drastically, which latter signals the start of massive recrystaUization in the material, this temperature being defined as the recrystallization initiation temperature (Tj). Only recovery took place below Ti and there was not much change in the rnicrostructure from the as-drawn state. Nucleation started from Ti and grew quickly as the temperature increased. Along the curve at the higher temperature side there exists a temperature above which the hardness change becomes much slower, which latter indicates that the recrystallizafion had become completed after the material had been held at the temperature for 1 h, this temperature being the recrystallization temperature (Tr) for the deformation. Equi-axial crystals formed at Tr and, when the amlealing temperature was greater than Tr, large grains appeared, indicating the progress of grain growth. The evolution of the microstructure can be seen from the micrographs presented in Fig. 4, which show typical microstructures of the metallographicaUy prepared samples at various an- t~ealing stages. It is also noted that there exists a hump in the recrystallization region on the hardness vs. temperature curve for many deformation levels (Fig. l(b) or Fig. 2 for the 'hard' alloy and curves for high deformation levels in Fig. 3 for the 'medium-hard' alloy). A satisfactory explanation is not yet available as to what caused the slope change. However, it is suspected that this might be associated with the special mode of deformation in wire drawing. This slope change was ignored when the Ti and Tr were determined. The values of Ti and Tr for the two types of material were plotted against the deformation, as shown in Fig. 5. For both types of material, the values of Ti and Tr are high at low deformation; decrease slowly before CW < 80%; and then decrease more rapidly when CW > 80%. It has long been established [8] that: T~ = [,4/NG3] ''4 where A is a constant, N is the recrystallization nucleation rate and G is the rate of growth of the recrystallized embryos. As a high deformation state is associated G. Qi, S. Zhang /Journal of Materials Process#lg Technology 68 (1997) 288-293 292 with more severe distortion of the initial strain-free crystal structure, the values of N and G will be greater compared with those of the low deformation state, which will result in lower recrystallization temperature according to the above equation. Another observation from Fig. 5 is that the values of a n d Tr for the 'hard' alloy are about 80 to 100 K greater than those of the 'medium-hard' alloy, depending on the deformation, which indicates that the mierostructure of the 'hard' alloy is more heat resistant or, in other words, a relatively greater temperature is needed to bring the deformed alloy back to a strain-free state. The difference is a result of the different dopants used in the materials. It is accepted generally that impurities will cause distortion of the matrix and impinge movement of dislocations, and hence raise the recrystallization temperatur,=. However, the effect will vary with the elements involved. Elements with more significant differences from gold in atomic radius, crystal structure and electrical conductivity, etc., will have a more profound doping effect. In the case of the two types of alloys under study, in addition to the beryllium present in both alloys, the 'hard' alloy had calcium and indium as dopants, instead of the platinum and palladium in the 'medium-hard' material. The difference in the recrystallization temperatures of the materials at the early stages of drawing is relevant to the behavior of the bonding wires in bonding applications. Usually, bonding wires in most applications are not fully stress-relieved or recrystallized. In a ball-bonding process, a piece of wire is melted to form a ball and then connected to a bond pad by thermo-compressing or thermo-sonic welding. During the ball formation, a portion of the heat from the heat source will pass to the wire immediately above the ball through conduction, causing recrystaUization or grain growth. The length of wire where appreciable grain growth can be detected is 800 ~_ 750 j -- HardAlloy: Start Hard Alloy: Finish Medium Hard Alloy: Start MediumHardAlloy: Finish 700 i S I,o= 650 i 550 600 500 2'0 42 "P 40 O .~ 34 32 30 ~ 20 | i 40 i i ,| 60 Deformation i 80 i 100 (%) Fig. 6. Microhardness (HV) vs. deformation for the fully annealed "hard' and the 'medium-hard' gold alloys. called the heat-affected zone (HAZ). Since the heat input to form a ball of a particular size is essentially the same, a 'hard' wire will have a shorter HAZ and smaller grains in the HAZ compared with a 'mediumhard' wire, as the 'hard' wire needs a greater tempe;ature to recrystallize or to initiate grain growth. The microhardness at the recrystaUization temperatures is plotted against deformation, as shown in Fig. 6. It is seen that there is a steep increase in the hardness of the fully annealed alloys after about 80%CW. This is believed to be related to a decrease in grain size resulting from more severe distortion of the initial structure. The phenomenon, together with the more rapid decrease in recrystallization temperatures after 80%CW (Fig. 5 and its discussion), seems to suggest that ~ 80%CW is a critical deformation region, around which more obvious property changes can be expected. Another important observation from Fig. 6 is that, although there are significant differences in recrystallization temperatures (Fig. 5), the two types of fully recrystallized material have hardness values that are very close to each other, i.e., between 35 and 40 HK (or HV) (100 gf, 10 s). These hardness values may serve as a guide for choosing appropriate annealing conditions in production, where the oven and methods to handle the material could be all different from those used in the present work, i.e. properly annealed gold alloys for the two types of bonding-wire production should have a microhardness value between 35 and 40 (HK or HV). 4. Conclusions I¢ 45O ---e--- MediumHard Alloy - - o - - HardAlloy 44 ;o 6'o 8'o ,oo Deformation (%) Fig. 5. Correlation of T.~and Ti vs. deformation for the two types of gold alloy. Recrystallization maps correlating deformation, temperature and hardness for two types of gold alloys for producing fine bonding wires have been established. Recrystallization initiation temperatures (Ti) and re- G. Qi, s. Zhang ~Journal o/ MateriaL~ ProcessmE Tech*wloKv 68 (t997)288 293 crystallization temperatures (T~) were inferred from the maps for the alloys. The values of T~ and T~ for the 'hard' alloy are about 80 to 100 K greater than those for the 'medium-hard' alloy, which is attributed to the different dopants present in the materials. However, when fully annealed, the two types of" gold alloys possess a similar hardness, between 35 to 40 HV or HK, which should be the controlling target value in a production process where the alloys need to be fully recrystallized and yet restrained from undesirable grain growth. [2] [3] [4] [5] [6] [7] References [8] [I] P. Douglas, Metallurgical Fundamentats of Gold Bonding Wire, 293 Technical Report # 4, American Free Wire Co., (1980s: exact year of publication not known). S. Tomiyama, Y. Fukui, Go!d bonding wire for semiconductor applications, Gold Bull. 15 /1982) 43. B.L. Gehman, Gold wire for automated bonding, Solid State Technd. IMarch)119801 84 91. K. Busch, H.U. Kunzi, B. llschner, Annealing and recrystallizalion kinetics oJ ullrathin gold wires: Modulus and resist;vit} measurements, Scripta Metall. 22 {1988) 51)1-505. T.H. Ramsey, MetalJurgicai behavior of gold wire m thermal compression bonding, Solid Stale Technol. {October) 11973) 43 -47. M. Grimwade, The metallurgy of gold, lnterdiscip. Sci. Rev. 17 (1992) 371 38l. ASM Handbook (Formerly Tenth Edition of Metals Handbook), Vol. 2, 1992. J.D. Verhoeven, Fundamentals of Physical Metallurgy, Wiley, Inc.. 1975.