Only the JEDEC JESD22-B116 most recent version that addresses Cu wire bonds, while the other standards are primarily meant for Au wire bonds. Shear.
Bond Pad Effects on the Shear Strength of Copper Wire Bonds Subramani Manoharan [1], Stevan Hunter [1, 2] and Patrick McCluskey [1] [1] Center for Advanced Life Cycle Engineering (CALCE), University of Maryland, College Park, Maryland 20740, USA. [2] ON Semiconductor, 5005 East McDowell Rd, Phoenix, Arizona 85008, USA
Abstract Wire bond evaluation is crucial in determining its quality. With the increase in use of copper wire bonds, this becomes even more important due to its nature to induce defects such as pad cracking. Bond shear testing is a widely used method to assess wire bond quality. Standards for performing shear test are in place (JEDEC, ASTM, IEC, etc.) however, they do not prescribe factors that affect results, making it difficult to compare bonds made in different conditions. Bond pad thickness is a major factor, studied in detail in this work. It is shown from a design of experiment and through finite element analysis, that bond pad thickness affects the bond shear strength value, while the IMC % coverage and shear mode were not dependent on pad thickness.
the ball bond to be tested, is used to apply force in the lateral direction (parallel to the substrate).
Introduction Copper (Cu) and palladium coated copper (PCC) wire bonds are increasingly used in the semiconductor industry as an alternative to gold (Au). The switch in wire bond material was primarily driven by the increasing cost of gold, however Cu wire provides some physical benefits such as higher thermal conductivity, electrical conductivity, lower intermetallic compound (IMC) formation with aluminum (Al) and higher sweep resistance during transfer molding process [1][2].
The shear tool is set to move at a constant speed, while the force encountered to move the tool in its path is noted. As the tool hits the bond and stops, the force is increased until the bond breaks [5]. Peak force that causes the break is recorded as the shear strength of the bond. Bond shear testing is primarily used to optimize the bonding process and then to qualify and monitor wire bonds on production samples. It can also indicate the stress tolerance of the bond, such as stress from thermal expansion of mold compound during thermal cycling conditions.
Copper’s inherent material properties makes it unsuitable for wire bonding in several applications. Its increased hardness requires higher bonding force and ultrasonic energy to be applied, causing higher potential for fracture of the bond pad or the sensitive circuitry under the pad [3]. Aluminum (Al) extrusion on the periphery of the bond, commonly referred to as “splash”, is also a concern, which can lead to electrical shorting in fine pitch bonding. Other failures such as ball bond lift off and ball bond neck crack are also reported [1][4]. These failure modes can all result from non-optimized bonding conditions. It is essential to detect the potential for these failures to assure bond quality and prevent reliability issues.
Bond shear strength depends on several factors such as: 1. Bonded materials 2. Area of bonding 3. Surface roughness of bond pad 4. Oxidation of Al pad surface 5. IMC strength 6. IMC area 7. IMC % area coverage 8. Cracks or voids in the bond interface 9. Pad film ductility and strength
Bond shear strength measurement is widely used in the industry to qualify and monitor wire bonding processes, with high shear force value supposedly indicating a reliable bond. Bond Shear Test Bond shear testing is the process of measuring the bond adhesive strength by performing a destructive test. Figure 1 illustrates the process, where a shear tool, much wider than
Shear Direction
Ball Bond IMC and Al from pad Aluminum Pad Specimen Clamp
Figure 1: Bond shear testing showing bond yielding at the interface.
Thermosonic ball bonding of Cu wire utilizes power, time, and force in the bonding “recipe”, controlling the process to break native oxide on the Al pad surface and cause increase in temperature, promoting formation of a solid-solid solution of Cu-Al IMC. Wherever oxide remains intact on the Al surface, IMC will not form, hence no bonding in that area. IMC area is most significant in determining bond strength, with increased IMC coverage hence higher strength preferred. Bond strength can be measured in a pass/fail sense in the bond pull test. If the wire breaks mid-span while the bond ball
is still intact, then the bond is considered strong. However, this doesn’t predict the bond’s reliability once it is packaged, surrounded by molding compound with a different thermal coefficient of expansion. The bond shear test is used on these strong bonds to provide a value that may be more indicative of bond reliability. IMC area is most significant in determining bond strength, with increased IMC area resulting in higher shear strength. IMC coverage as percentage of bond area can be measured in a destructive test. Companies may decide to set a lower limit for IMC coverage such as 80%, below which reliability may be assumed to be questionable. Several wire bond shear testing standards are in use, such as JEDEC, ASTM, IEC, etc. [6][7][8]. These have been developed to perform standard shear tests facilitating the results to be compared. Only the JEDEC JESD22-B116 most recent version that addresses Cu wire bonds, while the other standards are primarily meant for Au wire bonds. Shear strength values increase with ball diameter, and ball diameter increases with wire diameter. This is due to larger IMC area, from increased free air ball (FAB) diameter in larger diameter wires. Ball bond diameter is typically 2 to 3 times the wire diameter. A non-destructive bond shear test (NDBS) can be used in cases where 100% screening is required. In NDBS testing the shear tool stresses the bond only up to a preset value, which is kept much below the predetermined force required to shear a “good bond”, and then retracts. It has been shown that NDBS does not affect ultimate destructive shear strength (for 25.4um bonds with bond diameter greater than or equal to 2.5 times the diameter). Usually NDBS limit of about 50-60% is used. ASTM (F1269-06) recommends different limits for NDBS tests based on the bond conditions [11]. It was recently shown that bond pad thickness increase can lead to a higher shear force [11][15]. Experimental and finite element simulations are employed to continue studying this effect. Experiment Four different bonding process recipes on four different Al bond pad thickness were used in the experiment to study the bond pad thickness effects on bond shear results. Thermosonic ball bonds were made with Cu wire of diameter 18 µm on Al integrated circuit test pads. Pad thicknesses of 0.5, 1, 3 and 4 µm Al were included, to cover a wide range of thickness used in integrated circuit and power electronic devices. Power and time were varied in the bonding recipe to purposely create bonds that are strong, yet exhibit a range of IMC coverage in the bonds. Table 1 lists the experimental bonding recipe main parameters.
Table 1: Bonding process recipes used in the experiment Bonding Parameters Power (Watts) Time (ms) Force (gF)
Recipe 1 94% (1.78) 67% (37) 100% (47.5)
Recipe 2 97% (1.84) 75% (41.5) 100% (47.5)
Recipe 3 97% (1.84) 84% (46) 100% (47.5)
Recipe 4 100% (1.9) 100% (55) 100% (47.5)
Recipe 4 is optimal for bonding on these pads, as would be used in manufacturing. Recipes with reduced energy are formulated to create strong bonds having less IMC coverage. Lower energy than this results in “non-stick on pad” (NSOP), where the Cu ball pulls off the pad during the wire shaping motions. It’s not feasible to go to higher energy in the experiment because the bond ball diameter begins to exceed the pad width. After wire bonding, the assemblies were aged at 185°C for 4 hours to promote small IMC growth in the region where IMC had already formed, to enhance IMC visibility later. Then the Cu balls etched away with the use of nitric acid to expose the Cu-Al IMC following the method of [13] (also see [14]) Optical microscope, and scanning electron microscope (SEM) with electron dispersive spectroscopy (EDS), were used to identify coverage of IMC at the interface, which was then measured and tabulated. Figure 2, shows the presence of Cu in the central region which corresponds to the CuAl2 IMC.
Figure 2: EDS analysis of the ball bond-bond pad interface region after etching. (Left) SEM image and (Right) Cu map indicating the Cu-Al IMC region.
Bonding area was measured from inspection of pad Al deformation after etching away the Cu ball bonds. Figure 3, shows the bond area versus experimental bond recipe for each of the pad Al thicknesses. Bonding area increases with higher bonding power and time as expected. But at lower power and time, thicker pads caused smaller bonds. We attribute this to the fact that more US energy is being absorbed into the malleable thick Al pad film instead of causing more Cu ball deformation. However, with increased power and time the bond ball shape and bond area became essentially constant at the largest diameter, regardless of pad thickness. This indicates an apparent robustness against pad thickness changes in the optimal bonding recipe.
Figure 3: Ball bond area measurement for all pad thickness with different bonding recipes.
Figure 5: IMC percent coverage measurement for all bond pad thicknesses with different bonding recipes.
Figure 4 shows the IMC area measured after etching the bond. It shows an increase in IMC area with higher energy bonding recipes, as expected. The error bars show the range measured from 0.5 to 4 µm pad thickness. Maximum area of IMC coverage is limited by the bond area. There does not appear to be a relationship between IMC area and pad thickness.
Shear strength and shear mode was tabulated for about 10 bonds for each recipe on each of the pad thicknesses. Figure 6 shows example micrographs of bonds after shear test for the 4 pad thicknesses. All sheared in the Al below the bond. The shear mode is important. In this case, the shear mode tells us that Al ultimate strength is being exceeded, and that all other aspects of the bond are therefore stronger than the Al shear strength, including pad films adhesion to the silicondioxide below, IMC adhesion between Al and Cu, and the Cu ball itself.
Figure 4: IMC area in different pad for different bonding recipes.
%IMC coverage (with respect to bond area) in Figure 5 increases from about 50% to over 80% as bond recipe energy increases. Error bars show the range of measurements made from 0.5 to 4 µm pad thickness. %IMC coverage also becomes consistent at the highest energy recipe (as seen by short error bar) similar to the ball bond area result in Figure 3. The data all support the premise that bonding Recipe 4 is optimal for production.
Figure 6: Shear mode for bonds on different pad thickness. (Clock wise from top left) 0.5um, 1um, 4um and 3um.
Figure 7 is example optical micrographs showing IMC after etching away the ball bonds, for the four pad thicknesses. Each row shows a different bonding recipe, starting with Recipe 1 at the top. Each column is a different pad Al thickness, starting with 0.5 µm on the left. Notice how IMC area increases as bonding energy increases from Recipe 1 to Recipe 4.
bond area for thick pads. However, the higher energy bonding recipes show a more consistent and medium value shear strength. This is due to reduced variation in ball bond area with the highest energy bonding recipe as we saw in Figure 3. Higher bonding power and time causes the bond ball to flatten, with increased bond area and reduced sensitivity to bond pad thickness.
Figure 7: Images of ball bond-bond pad interface post etching on 3µm bond pad thickness. IMC region is highlighted in grey shade and the ball bond area is marked.
Results and Discussion Shear force measurements are shown in Figure 8, which shows a trend of increased shear force with thicker pad for the same bonding recipe. An average of 33% more shear force is required to shear the ball bond off 4 µm Al pad as compared to 0.5 µm pad thickness. Higher power and time (obtained by changing recipe from 1 to 4) in bonding caused an increase in shear force due to increased IMC area, as expected. It is hypothesized that the apparent increase in shear force with increased pad Al thickness is due to higher stress absorption in the thicker pad Al beneath the bond prior to shearing the Al film.
Figure 9: Shear strength as force/(bond area) for bonds on different pad thickness with different bonding parameters. Red line indicates minimum shear strength requirement of JESD57J.
Note that Cu ball bonds show much higher shear force than Au, because on a good Cu bond the shear failure occurs within the pad Al film, whereas a well-bonded Au ball will shear within the Au, which has a lower modulus than the pad Al. But poor IMC coverage in either Cu or Au bonding can result in a ball lift due to inadequate adhesion, with a similar failure mode, hence the same shear strength limit applied to both. Even with poor %IMC coverage on Cu wirebonds from Recipe 1 and Recipe 2, the bonds exhibit acceptable adhesion compared to the JESD47 limit. However, poor IMC coverage is expected to cause higher electrical resistance, leading to additional failure modes in the bond during use conditions. The manufacturing process control practice of optimizing a bond recipe for high shear strength, verifying high %IMC coverage, and operating at high Cpk for shear value, is the best assurance for reliable bonds.
Figure 8: Shear force values for bonds on different pad thickness with different bonding recipes.
Shear strength is calculated by dividing the shear force value by the bond area. A recently revised standard, JEDEC JESD47J for qualification of integrated circuits, states that each Au or Cu ball bond on Al bond pad shall have shear force per bond area greater than 0.0062 gram force/µm2 to be considered reliable. Figure 9 shows the shear strength per bond area obtained in this study, which is much higher than the accepted limit (marked by red line). Shear strength shows an increasing trend with pad thickness for all bonding recipes, which is due to the higher shear force measured and smaller
Finite Element Analysis Finite element analysis (FEA) was performed to investigate stress in the pad Al due to the force of a shear tool, to try and match the physical data and to gain more understanding from the shear test. Figure 10 shows screen captures of a simple model of a bond ball attached to the bond pad. The wire diameter is 18µm, and bond diameter 60µm to roughly match experimental conditions. A flat was designed on the wire bond to facilitate application of shear force.
198 Mpa
Figure 11: (Left) Bond pad top view. Red section line is shown. (Right) Section showing stress above ultimate strength in Al bond pad.
Figure 10: (Clockwise from top-left) Finite element model of ball bond on bond pad. Meshed assembly. Bottom of ball bond. Flat surface created on bonded ball for applying shear force.
The pad Al bottom face was implemented as a bonded “fixed support” for simplicity. Von Mises stress is analyzed for the metals. High stress is observed at the shear tool contact point as expected, with stress reducing with distance away from that point. To observe stress distribution in the Al below the bond in the different pad thickness, Von Mises stress is monitored in the Al 0.25µm beneath the Cu bond. Thick pads show lower stress than thinner pads for the same applied force. In other words, it will require higher shear stress on the Cu ball to cause the Al film below it to yield, when the Al is thicker. Table 2 summarizes data for an applied shear force of 12 gF, for FEA models of 4 different Al pad thicknesses. For the same applied shear force on the Cu ball, FEA predicts much more stress inside the thinnest pad Al film, 43% increase over stress in the thickest pad at the same point below the Cu bond (comparable to the 33% average increase observed experimentally).
Four to five times the amount of force (12 gF) is actually required experimentally to cause the whole bond to shear from the pad, but FEA has sufficiently validated the effect of shear force on pad thickness by this simple modeling. Lower stress in thick pads is due to the increased stress dissipation through the malleable Al. Figure 12 shows aerial views of stress in the pad Al at 0.25 µm beneath the Cu ball bond. As mentioned, the applied 12 gF on the Cu ball causes highest stress just at the ultimate yield point for Al in the thick pad (shown by the yellow region). The thin pad, in the same location, has 283 MPa stress (red region), already far above the yield point. This is another view illustrating that physics of shear testing predicts higher shear values for Cu bonds on thicker pad Al.
Table 2: Relative shear stress for different pad thickness FEA Relative Shear Strength in pad Al for stress on Cu ball of 12gF Pad Thickness (µm) 0.5 1 3 4 Von-Mises Stress (MPa)
283
252.6
212
198
Relative Change (%)
43
27.6
7.1
0
FEA helps explain why higher force is required to shear the Al film apart on thicker pads. Shear stress must exceed the ultimate strength of Al (~205MPa for 99% Al alloy) before it will fracture and then shear apart. 12gF is sufficient to cause a maximum stress point of 198 Mpa, near the ultimate strength in the pad Al. Figure 11 illustrates this case.
Figure 12: Von-Mises stress at a section 0.25µm under the bond pad for an applied force of 12gF.
Conclusion Bond shear testing is an important evaluation process that is used to categorize good quality bonds from bad ones. Several factors contribute to the shear force. We have shown that Al bond pad thickness is a significant factor in shear test results, with Cu bonds on thicker pads producing higher shear strength values than for thinner pads, even for non-optimized Cu wire bond recipes. Shear mode is consistent throughout the experiment, shearing in the pad Al beneath the bond. IMC% coverage was measured and it was concluded that it
did not cause the increase in shear force. A hypothesis that the thick Al pads absorb higher stress than thin pads is proven by performing FEA on a simplified model of the ball bondbond pad interface. For the same applied lateral stress to the Cu bond ball, calculated stress developed in the Al pad film is significantly less for thick pads compared to thin pads. This implies that shear test limits in manufacturing should take pad Al thickness into account when comparing bonds from different devices. References [1] Chauhan, Preeti S., Anupam Choubey, Zhaowei Zhong, and Michael G. Pecht. "Copper Wire Bonding." Copper Wire Bonding (2013): 1-9. Web. [2] Manoharan, Subramani, Gopal Krishnanramaswami, F. Patrick Mccluskey, and Michael G. Pecht. "Failure mechanisms in encapsulated copper wire-bonded devices." 2016 IEEE 23rd International Symposium on the Physical and Failure Analysis of Integrated Circuits (IPFA) (2016): n. pag. Web. [3] Hunter, S., Mallik, A., Whittaker, D., Alldredge, R., and Rodrigues, T. “Simulation of Ball Bonding on Various Bond Pad Structures,” IEEE Electronic Packaging Technology Conference (EPTC), 2013. [4] Loh, K., Pan, Y. J., & Tan, C. E. (2016). Production challenges of TSOP Copper wire bonding. 2016 IEEE 37th International Electronics Manufacturing Technology (IEMT) & 18th Electronics Materials and Packaging (EMAP) Conference. [5] Hueners, J., “Introduction to Wire Bond Pull and Ball Shear Testing”, Palomar Technologies, November 13, 2012. [6] Wire Bond Shear Test, JESD22-B116B, May 2017. [7] Standard Test Method for Destructive Shear Testing of Ball Bonds, ASTMF1269-13, January 2013. [8] Wire Bond Shear Test, AEC-Q100-001 REV-C, October 1998. [9] Qin, I., Xu, H., Milton, B., Mendoza, N., Clauberg, H., Chylak, B., Nakamura, S. (2014). Process optimization and reliability study for Cu wire bonding advanced nodes. 2014 IEEE 64th Electronic Components and Technology Conference (ECTC). [10] Breach, C. (2010) “What is the future of bonding wire? Will copper entirely replace gold?” Gold Bulletin, Volume:43 (Issue No.3), pp. 150-168. [11] S. Manoharan, P. McCluskey and S. Hunter, “Effects of bond pad thickness on the shear strength of copper wire bonds,” High Temperature Network, July 2017, Vol. 2017, pp 68-73. [12] Daniel L., and Wong C.P, “Advanced materials for electronics packaging,” Springer International Publishing, 2017.
[13] G.M. O’Halloran, Arjan van IJzerloo, Rene Rongen, Frank Zachariasse, “Planar Analysis of CopperAluminium Intermetallics”, International Symposium for Testing and Failure Analysis (ISTFA), San Jose, CA, USA, Nov. 2013 [14] Kid, W. B., Lin, Y. J., Leng, E. P., Yun, Y. K., & Yong, C. (2012). Effect of chemical etching solution on Cu/Al IMC result. 2012 14th International Conference on Electronic Materials and Packaging (EMAP). [15] Andrews, D., Hill, L., Collins, A., Hoo, K.I., and Hunter, S. “Copper Ball Bond Shear Test for Two Pad Aluminum Thickness,” IEEE Electronics Packaging Technology Conference (EPTC), 2014.
