Process Development of Al-alloy Wire Bonding for High-Temperature Power Electronics

As a cost-effective, flexible, and reliable interconnect method, Al wire bonding remains the mainstream interconnect technology in power electronics and semiconductor industries. With innovation of new devices made of silicon carbide (SiC) or gallium nitride (GaN) and Ag sintering die attach process, power electronics devices can be operated over 400 °C. Unfortunately, the conventional Al wires are not suitable for high-temperature applications since pure Al wires become soft when operation temperature is higher than 150 °C. New wires for chip topside interconnects have to be developed to accommodate high-temperature applications. Cu wire has high mechanical strength, higher electrical and thermal conductivities, and can be operated in a higher temperature, which is a promising candidate for interconnect in high-power, high-temperature power electronics [1], [2]. Several companies have been using Cu wire in their high-end products. However, due to its hardness, Cu wire requires special die-top metallization and still has a few challenges in the bonding process [3]. Those challenges have to be overcome before Cu wire can be widely adopted in the power electronics industry. Thus Al-alloy bonding wire with better mechanical properties and less demand for bonding process is desired.

Besides the high temperature electronics, Al-alloy wire is also a good candidate for few other applications. Electric vehicles have gained more and more market share in the automotive industry. Electric vehicle battery interconnect becomes one of the fast-growing markets for wire bonding technology. The wires for battery cells interconnect need high mechanical strength because they may be subjected to strong vibration, especially when they are located near the engine [4]. Furthermore, the wires not only interconnect individual cells to form a battery pack, but also act as fuses to protect the battery in case an electrical short occurs. Although Cu wire has high strength and electrical conductivity, its high fusion current prevents its use in this application. The batteries are not manufactured in a cleanroom like semiconductor devices. The battery surface is usually contaminated or oxidized to a certain degree. It will cause wire bonding problems if the contamination becomes more severe. To avoid additional cleaning process, some companies increase bond power to have an extra bond strength to overcome surface contamination. Excessive bond power likely will cause excessive deformation on soft Al wires and forms a weaker heel. Therefore, an Al wire with higher strength and hardness is required.

Al-alloy wires such as AlX wire [5] and Al-Sc [6] wire from Heraeus and 2N-Al wire from Tanaka [7], with modified hardness and microstructure, have improved performance and can be operated up to 300 °C. Those wires generally have 99% Al purity and <1% alloying particles. With finer grain size and alloying particles acting as obstacles for grain coarsening propagation, those wires have higher heat resistance, stronger tensile strength, and higher thermal fatigue resistance than conventional Al wires, thus are suitable for high-temperature power devices. Unlike Cu wires, Al-alloy wires do not require special die-top metallization and are easier to implement in current mass production. Due to their higher strength and hardness, however, those wires present higher risk of die damage than conventional Al wires during wire bonding process. Bonding process development and optimization is critical to minimize die damage and maximize device reliability. In this paper, the bonding process of 300 μm diameter 2N-Al wire bonded on SEMIKRON IGBT devices is studied and optimized, and the mechanical properties of 2N-Al wire bonds are compared with 4N-Al wire bonds on Panasonic CGR18650DA battery.

Samples were bonded with an OE3600plus wire bonder equipped with an 80 kHz large-wire rear-cut bond head using standard tungsten carbide bonding wedges (Kulicke & Soffa Industries, Inc.).

The wire being used in this study is 2N-Al wire from Tanaka. Alloyed with <1% metal element, this wire has much finer grain size than standard Al wire, as shown in Fig. 1.

Due to precipitation hardening, fine particles of alloyed materials could be participated in the material while temperature increases, thus reduce the grain coarsening of Al at high temperatures. Table 1 shows that 2N-Al wire itself exhibits much better thermal and mechanical properties than conventional 4N-Al wires. The tensile strength of 2N-Al wire is almost twice that of standard 4N-Al. Thermal cycling test results from Tanaka show that the thermal fatigue life time of the 2N-Al wire is about 2 times as that of a standard 4N-Al wire.

For parameter optimization, the 2N-Al wires were bonded on IGBTs and DBCs in power modules provided by SEMIKRON, as is shown in Figure 2.

For the mechanical strength comparison, 2N-Al wires and standard 4N-Al wires were bonded on Panasonic CGR18650DA battery cells, as is shown in Figure 3.

The process optimization was conducted in 3 steps through design of experiments (DoE). A regular bonding process has 8 bond parameters – Touch Force (TF), Start Force (SF), Bond Force (BF), Start Power (SP), Bond Power (BP), Start Ramp Time (SRT), Bond Ramp Time (BRT), and Bond Hold Time (BHT). The process optimization was to define an optimal value for each bond parameter so that an optimal bond quality would be achieved. The criteria to evaluate bond quality were pull force, pull failure mode, shear force, shear nugget coverage, and die damage rate.

Die damage was initially tested by measuring the electrical resistance between the emitter, collector and gate every time before and after an IGBT module was bonded. In the confirmation step, more accurate measurement of die damage was conducted with a Gate Tester, which detects gate leakage current.

The pull and shear tests have been done with Nordson Dage Series 4000. Shear tests were performed using a shear height of 30 μm and a shear speed of 300 μm s⁻¹. Pull tests were performed using a pull speed of 500 μm s⁻¹. Both pull and shear forces were recorded with the highest load.

In the first step, factor screening, 4 parameters were identified as the most significant factors out of the 8 bonding parameters. This step started with defining lower and upper limit of each of the 8 bonding parameters by scanning the parameters from low to high. 12 runs of Plackett-Burman screening design [8] were created when the range of each bonding parameter was defined. Plackett-Burman design is a fractional factorial design in which interactions of the factors are considered negligible and the number of runs is minimized. The detailed design matrix for factor screening is shown in Table 2. The aforementioned criteria was used to make a Pareto Chart of Effects. Combining the 5 Pareto Charts, a ranking list of effect was generated. The four most significant factors were: Bond Force, Bond Power, Start Power, and Bond Hold Time, which would be optimized in the next step.

In the second step of the optimization, an optimal value was identified for each of the important bonding parameters. A central composite design (CCD) [9], [10] of the 4 parameters was created, which contained 31 runs of tests. Table 3 shows the rough design matrix of the CCD. Pull and shear forces were used as responses for optimization. Two fitting models were created from these test results. Analysis of Variance was used to make sure the fitting models were good.

From each of the fitting models, an optimized set of parameters considering pull or shear strength was obtained. By combining the two models and maximizing both pull and shear strength simultaneously, a third set of optimized bonding parameters was obtained.

Then, 4 runs of test were conducted to confirm that the optimized parameters produce optimal bond quality. 250 bonds were made in each run to have enough sample size to evaluate die damage rate. No die damage was detected with either measurement methods.

In electric automobiles, the battery cells are subjected to strong vibration, and mechanical loads such as bending force would be applied to the interconnect wires. The mechanical strength of 2N-Al wire itself has already been studied by Tanaka. However, the strength comparison of bonds made of 2N-Al and standard Al wire (4N-Al) were not performed on real battery cells. In this paper, comparison tests were performed to compare the mechanical strength of the two wires and bonds. Bond process windows of both wires on CGR18650DA cells were explored. Pull and shear tests were conducted on the samples.

The response surface DoE procedure consisted of 31 runs of tests. The boxplot of pull and shear forces of all the 31 runs are shown in Figure 4. All pull forces fall between 1000g and 1200g and the shear forces fall between 1600g and 2500g. Figure 5 presents the bond appearance and nugget coverage after shearing of bonds of the 31 runs.

Figure 6 shows the optimization plots of shear force, pull force and the combination of pull and shear forces. Each row describes how the response changes with the four significant parameters. Individual desirability of each response (pull force or shear force) was calculated and both responses were assigned with equal weights in the calculation. When the composite desirability is set at the maximum, shown in the first row of Figure 6, the optimal parameters are obtained and shown with the red color. With the optimal parameters (BF=957.6, SP=74.5, BP=56.6, BHT=107.3), the bonds are predicted to have both high pull force (1154.3g) and high shear force (2274.1g).

To verify the calculated optimal parameters can produce such high pull and shear forces, a confirmation run was performed. Two parameter sets that produced lowest and highest bond deformation in the above 31 runs were also included in the confirmation run as a comparison. The description of each parameter set is shown in Table 3. The parameter set that produced the highest pull force was excluded from the confirmation run because the pull force is not very sensitive to bond parameters in this application.

Figure 7 presents an overview of the bond appearance and the bonded area after shear test. All bonds shows acceptable lateral deformation. The parameters set Low presents no nugget coverage after shear test, while the other three parameter sets present more than 75% of nugget coverage.

Pull and shear force are shown in Figure 8 for the confirmation run. After optimization, the bond quality is improved and >1,100 g pull force and >2,200 shear force are achieved. The Low parameter presents normal pull force but low shear force, while the High parameter presents both acceptable pull and shear force.

With the four sets of parameters, measurement of die damage was performed. No die damage was found with all the bonding parameters. Thus, conclusion could be drawn that this wire has wide process parameter window on SEMIKRON's IGBT devices.

Pull and shear force tests were performed on the bonds made with 2N-Al and 4N-Al wires on CGR18650DA battery cells. Images of bonds made with the two wires are presented in Figure 9, which shows little difference between the bond appearance of the two wires. Figure 10 presents how the pull and shear forces of the bonds change with the bond power. Both the pull and shear forces of 2N-Al wire are ~20% higher than those of 4N-Al wires, which indicate that 2N-Al wire would be able to resist higher mechanical loads, such as the bending force caused by vibration. At high power (≥75), the bond strength of 4N wire drops while the strength of 2N wire is maintained at a high level. This result confirms that 2N-Al wire has wider bond process window than 4N-Al wire at high power side, which would benefit the wire bonding on contaminated battery cells where a higher bond power may be required.

The 2N-Al wire has better mechanical strength and higher thermal fatigue resistance than conventional 4N-Al wires. The paper showed a 2-step DoE procedure of optimizing the bonding parameters of the 2N-Al wire to high-temperature power electronics. Pull and shear forces are the two indicators that are used to evaluate the quality of the bonds. Die damage is monitored after wire bonding. Results of confirmation run show that the bonding process parameters are optimized without causing any die damage and this wire has a wide process window on IGBT devices.

Further comparison between the Al-alloy wire and standard Al wire was also conducted. This study shows that a change from standard Al wire to Al-alloy wire yields a decent improvement in terms of mechanical strength when bonded on battery cells. These conclusions indicate the potential of this wire to be used in automotive industry, where higher mechanical strength of the wire is required.