BAMFIT – From an Evaluation System to a Remanufacturing Platform for Heavy Wires in Power Electronics

BAMFIT (Bondtec Accelerated Mechanical Fatigue Interconnection Test) was developed over the last ten years to assess the reliability of heavy wire bonds in a timely manner, ideally on the day the assemblies are produced [1,2]. This is important for fast and cost-efficient production monitoring and development of bonding processes. It is an alternative or complementary method to the Active Power Cycling (APC) test, in which, unlike APC, the wire is moved mechanically. The unidirectional, cyclically induced stress is achieved by a standard ultrasonic device used in wire bonding machines and a special gripping unit [3]. The schematic test procedure is shown in Figure 1. The heavy wire bond connection is clamped at a predefined height with the gripping unit and then subjected to ultrasonic (US) in the wire direction and under a low tensile load.

Quantitative evaluation

End-of-life (EOL) Test

The service life or end-of-life (EOL) test involves testing the wire connection until the bond connection pulls off. The result is the test time or, at the known excitation frequency (e.g. 57 kHz), the calculated number of load cycles until failure (Nf). Typical EOL test times are between 0.1 and 10 seconds per wire connection. Czerny et al. investigated the influence of ultrasonic power using shear testing, BAMFIT and power cycling (PC) testing and compared the data with each other [3]. Essentially, a strong deviation was found between the static shear test and the two EOL test methods, while there was good agreement between BAMFIT and the PC test. Figure 2 shows the results of the EOL test of ~50 Al wire connections with a wire thickness of 500 μm for different bonding times and force ramps. In agreement with comparable shear tests, Nf also increases with increasing bonding time (200 ms → 400 ms) due to the increased bonding area. With a further increase in bonding time (400 ms → 1000 ms), Nf decreases. Presumably, the longer exposure time of the ultrasound changes the microstructure, which can affect the microstructure and consequently the mechanical properties and crack behavior within the bonded joint. The heights of the error bars indicate that the original idea of obtaining a clear statement about the reliability of the wire connection is problematic, especially when sampling with a small number of samples. This high dispersion of the measurement data is based on the complex interaction between microstructure and crack propagation. The next chapter describes a method in which the damage-dependent residual strength of the connection can be determined with lower measurement dispersion.

Fig. 1: BAMFIT functional principle

Fig. 2: Bond time-dependent end-of-life (EOL) test results:
● insufficient bond time,
● reference bond and
● overbonded

Crack Robustness Verification (CRV)

Crack Robustness Verification (CRV) CRV represents an improvement on the EOL test. Here, after pre-damage using BAMFIT, the remaining shear strength τ (shear force Fτ) of the connection is measured. The BAMFIT termination criterion is not based on a fixed number of cycles, but on a definable decrease in the electrical ultrasonic transducer current IUS, which is representative of the actual crack propagation. This allows wire connections with as similar pre-damage as possible to be tested, which leads to a reduction in the standard deviation. This advantage can be seen in the significantly smaller error bars in Figure 3. Here, the shear forces as a function of the BAMFIT pre-damage for the first and second bonds of a 500 μm Al wire are shown. The first data point in each case represents the reference value of undamaged wires – these are exclusively measured values from a shear test.

Qualitative evaluation

In addition to the quantitative assessment of the bond by Nf or the damage-dependent shear strength τ(IUS), BAMFIT also enables a qualitative assessment of the bonding area and crack propagation, thus contributing significantly to a better understanding of the bonding process and load-dependent crack propagation.

Fig. 3: Test results for Crack Robustness Verification (CRV) from ● 1. Bond und ● 2. Bond

Bonding area

In contrast to the shear test, in which material in the bonding zone is smeared, BAMFIT exposes the bond interface without smearing by applying vertical tension away from the interface. In regions where no bonding has taken place, the color and structure of the bond surface are almost unchanged. Good bond connections are usually small, isolated areas having a size of <5% of the total area (see Fig. 4, image taken at approx. 1000x magnification using light microscopy). If there is a poor bond connection (e.g. due to local contamination or a particle), this can be clearly demonstrated by the appearance of the interface. Unbonded regions are then significantly larger and can be easily quantified visually. This detection can be further facilitated and its significance improved by applying optical 3D measurement methods [4]. The fracture plane within a heavy wire that has been detached using BAMFIT runs approx. 20-30 μm above the bond interface (see next chapter). A height section within the 3D image of a BAMFIT fracture surface allows unbonded areas to be clearly distinguished from bonded areas (bonded areas with adhering Al residues). With the help of such advanced methods for interface evaluation, bond parameter studies can be supplemented with information on the actual bonded area and the shape of the bond interface – important data that can be used to further improve the robustness of a wire bonding process.

Crack propagation

The shear force reduction determined using CRV (see previous chapter) correlates with a reduction in the bonding area and progress. However, the exact quantification of the crack length in the top view after the test has been carried out is very difficult (e.g. using 3D topography analyses [4]) and does not provide any precise insights into the crack structure. A cross-section is much better suited for this type of analysis. Figure 5 shows such a crack progression in the cross-section of an Al heavy wire wedge on a Cu surface. The crack runs within the wire material, above the wire-bond pad interface. The crack in Figure 5 is only slightly branched, runs partly in the direction of the bond interface and then continues its path close to the interface. In the central region, no structure damaged by cracks is yet visible. The crack propagation changes depending on the selected bonding parameters and the microstructure in the bond contact, which is altered by the bonding process. Very fine-grained microstructures that are strongly altered by the bonding process produce a more branched crack propagation, while the crack spreads over very large areas close to the interface in unbonded regions. After a predefined load using BAMFIT, the degree of damage can be determined based on the total crack length (from both sides). This allows comparative statements to be made, e.g. between different wire types with very similar material compositions or different bonding parameters. It should be noted that a cross-section only shows one plane in the bond contact. It must therefore be ensured that the planes examined are prepared in a comparable manner. In addition, ≥10 crack structures per state are required for a statistically relevant statement.

Fig. 4: Exposed bond interface of a 300 μm Al wire on a Direct Copper Bonded
(DCB) surface metallised with Electroless Nickel Immersion Gold (ENIG)

Fig. 5: Crack propagation within an Al heavy wire bond after BAMFIT pre-damage

Fig. 7: Wire-bonded battery module consisting of 16 × 21700 battery cells, a
battery management system (BMS) and the mechanical structure. The red marking
symbolizes a defective cell that is relevant for remanufacturing. In (a), the
extracted wire bridge is disposed of in a controlled manner.

Remanufacturing

The obvious result of an EOL-BAMFIT test is the destruction of the bond connection(s) or the removal of a bonding wire. However, this secondary effect can also be used specifically for remanufacturing. To do this, a new wire is ultrasonically bonded after the wire has been removed with BAMFIT. These additional steps are shown again schematically in Figure 6 (f–h).

Remanufacturing of batteries

With the increasing demand for batteries – especially for the automotive industry – the question arises as to whether defective battery modules should be disposed of or reprocessed. It has been shown that a defective battery module can still consist of ~89% functional battery cells (see Fig. 7), which suggests reprocessing not only from an ecological but also from an economic point of view [5]. Compared to resistance-welded or laser-bonded ribbons, wire-bonded connections can be removed and rebonded much more easily. Removing the wire with a shearing tool carries the risk of an unintentional short circuit between the anode and cathode – especially if particles or entire wires come loose. This can be prevented with BAMFIT by first separating the two wire connections in <100 ms and then removing the wire bridge (which is fixed in the clamp at all times) in a controlled manner. The near-surface detachment and clean surface ensure good bondability. This means that wires for rebonding can be placed not only next to, but also directly on the original positions.

Remanufacturing of control units

The same principle – wire removal using BAMFIT, followed by wire bonding – can also be used for electronic assemblies. The remanufacturing of heavy wires in control units, as shown in Figure 8, for example, is intended to represent heavy wire remanufacturing in power electronics. Here, the original wires (a) were removed with BAMFIT (b) and rebonded (c). Electronic assemblies are embedded in gel to protect them from environmental influences and to ensure mechanical stability. This complicates the remanufacturing process, but it is not a showstopper. The combination of shearing and subsequent rebonding is a common method of achieving this. In this study, however, the wires are removed again using BAMFIT. As shown in Figure 9, the gel is mechanically removed (a → b) to expose the affected wires (c). Visual inspection shows that gel residues are still present. Residue-free chemical removal, which is preferred for bonding, was not considered in this feasibility study. The two wires were removed with BAMFIT and could still be rebonded afterwards. In (d), you can see the wire removed with BAMFIT (front) and the rebonded wire (back).

Fig. 6: Remanufacturing process flow consisting of the BAMFIT process (a-e) and
the rebonding process (f-h)

Summary

It all began with the development of BAMFIT, an evaluation system for the real-time prediction of the service life of heavy wires. The continuous evolution of the BAMFIT system allows for versatile insights into the connection formation of wire bonding processes as well as stress-dependent material properties of the bonded connection. A sustainable benefit is the use of this platform for reprocessing, whether for battery modules, control units or in power electronics in general. Table 1 summarises the main possibilities and combinations with BAMFIT.

References
[1] Czerny, B., Mazloum-Nejadari, A., Khatibi, G., & Zehetbauer, M. (2016). Fatigue testing method for
fine bond wires in an LQFP package. Microelectronics Reliability, 64, 270-275. https://doi.org/10.1016/j.
microrel.2016.07.068
[2] Czerny, B., & Khatibi, G. (2017). Accelerated mechanical fatigue interconnect testing method for electrical
wire bonds. TEME – Technics, Technologies, Education, 10(3), 121-131. https://doi.org/10.1515/
teme-2017-0131
[3] Czerny, B., & Khatibi, G., Highly Accelerated Lifetime Testing in Power Electronics. In Proceedings of
the 54th Int Symposium Microelectron (IMAPS 2021), 390-396. San Diego, CA, USA, October 11-14,
2021. https://doi.org/10.4071/1085-8024-2021.1.000390
[4] Schmitz, S. et al. Advanced bonding interface inspection technique for process optimization in heavy
wire bonding. Int Symposium Microelectron 2021, 332–338 (2021). https://doi.org/10.4071/1085-
8024-2021.1.000332
[5] Kampker, A., Wessel, S., Fiedler, F., & Maltoni, F., Battery pack remanufacturing process up to cell level
with sorting and repurposing of battery cells. J Remanufactur 11, 1–23 (2021). https://doi.org/10.1007/
s13243-020-00088-6