Evaluation of Low Cost, High Temperature Die and Substrate Attach Materials for Silicon Carbide (SiC) Power Modules Open Access

SiC power devices have demonstrated significant performance advantages as compared with traditional Silicon (Si) devices. A direct comparison of SiC vs Si power devices can be found in [1] [2]. Many studies have also projected a promising future for SiC power devices based on their desirable properties [3] [4]. The benefits of SiC technology over Si have also been tested at a system level [5]. They can operate reliably at higher maximum junction temperatures as compared with their Si counterparts, with stable and efficient operation at higher temperatures. This has several advantages. A reduced cooling demand, due to lower losses and higher junction temperature, means that the thermal management system can be simplified. A streamlined and efficient cooling system will enhance the system power density significantly. A higher operating temperature also implies a higher current rating for a given system volume. The on-resistance and switching loss of SiC devices is also significantly lower than comparable Si devices. This further enhances the ability of SiC devices to support a high power density with minimized system losses.

However, to achieve these benefits, the packaging of the SiC device must withstand the high stresses associated with operating at elevated temperatures and high temperature distributions across multiple thermal-mechanical layers. The properties of most materials available for power electronics packaging are limited or rated to the maximum capability of Si devices. With the increased demands associated with SiC packaging, materials need to be meticulously identified and evaluated for performance. Of particular interest are die and substrate attach materials. These material interfaces are situated closest to the die, and constitute the primary heat path from junction-to-case [Fig. 1]. It is imperative that these layers meet and exceed the aggressive thermal and reliability requirements of a SiC power module.

One of the major directions for SiC power modules is being able to process high power within a very small volume. As power devices are packed more densely, managing thermo- mechanical stress at the die and substrate attach interfaces becomes a challenge. The material choices to achieve performance targets either need to be engineered or existing solutions need to be verified. Additionally, the increasing requirement to meet RoHS and/or REACH compliance in global markets today, adds another layer of complexity to the material selection challenge. High-reliability solders [6], transient liquid phase bonding [7] [8], and sintering [9] [10] are examples of materials suitable for attach processes in SiC power modules. There have been several studies demonstrating the superior performance of these materials [6] [10]. However, more often than not, the conditions under which these materials are tested are very different from a specific power module application. For example, the die- attach material under test may be the same, but the material of the die, substrate material, plating specifications, temperature swings, dwell times, etc. may be entirely different. For a meaningful prediction of reliability, it is important to develop customized testing for the end application.

In this paper, a comprehensive customized set of experiments for the evaluation of an attach material is described. These experiments were tailored to the HT-4000 power module platform from Wolfspeed. One of the major goals for this power module was to maximize the ratio of SiC active area to available module footprint. The HT-4000 power module also enables a maximum junction temperature of 175 °C. It is capable of supporting ~ 500 A within a very compact industry-standard SP1 footprint. This makes the HT-4000 module an ideal example of the elevated thermal demands associated with next-generation SiC power modules. Fig. 2 shows a photograph of the HT-4000 power module. The module measures 51.6 mm × 40.8 mm × 17.6 mm.

The attach interfaces in this module include the die attach, substrate attach, and contact attach layers. All of these layers will be under varying levels of thermal stress depending on the operating conditions. This paper presents the efforts undertaken to evaluate an attach material for its suitability to comply with the aggressive thermal demand of the HT-4000 power module, based on its specifications.

Section II provides a description of the specifics and quality assessment of the die-attach process. Section III presents the results of a temperature dependent die shear test. Section IV describes the process and results of conducting a temperature-dependent contact pull test using the material of choice. The final section, Section V, describes the effect of thermal cycling and power cycling on the quality of the layers.

The selected material was used for performing the die, substrate, and contact attach process in a single reflow process. The thickness of the attach layer was determined by on the component geometries of the module and thermal resistance target. The goal was to minimize the thickness of the layers while meeting the performance and quality targets.

The attachment process underwent several iterations involving different thermal profiles, until an optimized result was obtained. Once the attachment was completed, three different quality inspections were performed on the modules – an initial visual inspection, x-ray imaging, and C-mode scanning acoustic microscopy (C-SAM). The visual inspection was targeted toward examining adequate fileting and for any obvious signs of structural damage, placement inconsistencies, or cosmetic issues during the attachment process. The x-ray scan was performed to identify cracks or delamination in the attach layers, which may have occurred due to a defective part or damage during the attachment process. A C-SAM scan is a standard technique for identifying and quantifying the voiding in the die and substrate attach layers. A combination of these inspection techniques provides a comprehensive estimate of the quality of the attachment process.

Fig. 3 shows a photograph of a typical test die after the attach process. The visual inspection process checks for a good fillet around the die and contacts, and proper component location, as shown in Fig. 3. Another important defect is the attach material spreading excessively to the wire bonding areas. This is usually an indication of an unoptimized process and will result in an inadequate attach quality, inferior long-term performance or scrapped modules in production. Finally, the visual inspection process helped identify any major cosmetic defects – like discoloration or mechanical damage – during the attach process. The appearance of all visible surfaces should remain uniform and defect-free.

Fig. 4(a) shows a typical x-ray image of the test module. The darker areas indicate the die attach layer. The comparatively lighter background shows an overlay of the module stack up – baseplate, substrate attach, and substrate. Any delamination, cracks, or voiding would show as bright white spots or light areas in the x-ray image. For example, Fig. 4(b) shows a module with a large substrate attach void around the lower periphery. This area appears brighter than the rest of the image.

Fig. 5(a) shows a typical C-SAM scan conducted on a test module. A good attach layer under the die should appear in grayscale as seen in the figure. Any voiding present within the attach layer would appear as bright white. The bright white area seen under the die in Fig. 5(b) is an example of a void formation. The die with worst case voiding is outlined in red. To verify the veracity of the C-SAM scan, this area was cross-sectioned and observed under a microscope. A large gap in the cross section was observed under the region corresponding to the white spot observed in the C-SAM scan [Fig. 5(c)].

An optimized reflow process is the key to producing high-quality parts in a repeatable manner. A good understanding of the process parameters and process windows is important. The largest single void in each layer did not exceed 3%. After the first couple of hundred parts were reflowed using this profile, it was found that the die attach voids occasionally demonstrated higher than desirable voids [see Fig. 6(a)]. This would sacrifice yield as the production volume increased. The die attach was found to be more susceptible to occasionally high voiding. A second round of optimization was conducted to obtain a repeatable and high-quality die attach. As a result of implementing some minor process changes, the total die attach voiding was reduced to < 3% across the next ~ 100 modules as shown in Fig. 6(b).

For this experiment four HT-4000 power modules were assembled with four die per switch position. The die measured 4.36 mm × 7.26 mm. It is recommended that the dimension of the die shear tip equals or exceeds the dimensions of the die edge being sheared. Thus, a 10 mm shear tip was used to shear each die along the longer edge. The test samples were verified to have characteristically low voiding using our optimized process.

The parameters for the shear process are shown in Table 1. For the heated shear test, a temperature controlled stage was used. The stage was calibrated using a thermocouple prior to mounting samples. The temperature on the surface of the substrate was then verified to match the desired test temperature. The heated stage employed vacuum suction holes and an anvil to securely hold the module in place while providing a good thermal contact. After the heater reached the set temperature, the module was mounted on the stage and the vacuum was engaged. The module temperature was allowed to stabilize for a period of 15 minutes before the test was conducted. Fig. 7 shows the die shear test setup, with the test module mounted on the heated stage.

Four die were sheared per temperature step. The results of the die shear test are presented in Fig. 8. Up to 125 °C, the shear limit of 100 kg was reached, and no bond breaks were observed. At the intended service temperature of 175 °C, the bonding strength was found to reduce marginally, but was still very strong at around 80 kg. At temperatures above 200 °C, however, there was more than 50% degradation of the bonding strength. This was expected since our materials were chosen and designed to operate at a maximum temperature rating of 175 °C. Fig. 8(a) shows a plot summarizing these results. Fig. 8(b) shows the Force – Displacement profile of the test at the various temperatures of interest.

MIL-883 method 2019.19 is a good reference for the die- attach bonding strength. Up to 200 °C, the die bonding strength was found to exceed the recommended minimum value of 25 kg by 2×.

The results from this test proved that the die attach material under test met the mechanical requirements of the module. The next step was to verify if the bonding strength of the contacts were adequate.

The contact pull test was conducted on various identical samples, under different conditions. Fig. 9 shows a photograph of a test module mounted on the pull tester. The terminals of the opposite edge of the module were sawed off to ensure that the jaws of the pull tester can be secured to the contacts being pulled without interference. The pull rate was set to 200 μm/sec. The pull was continued until failure.

As shown in Fig. 2, this module contains three larger power contacts and two pairs of thinner signal contacts. Results for these two types of contacts are grouped and presented separately to help identify the variation of bonding strength with contact geometry. In addition, contacts from two different vendors were evaluated.

• Test A: Contacts from Vendor 1 at room temperature, 175 °C, and 200 °C.

• Test B: Contacts from Vendor 2 at room temperature, 175 °C, and 200 °C.

The results of test A are presented in Fig. 10. Fig. 10(a) shows the results for the power contact. The initial bonding strength of the power contacts was > 1100 N. This was only reduced to 900 N at the desired operating temperature of 175 °C. Even at 200 °C, the mean strength of the power contact was close to 800 N. However, there was a lot of variation of this value from sample-to-sample. This possibly occurs because the material becomes unevenly soft from sample to sample. According to IEC60068-2-21, 40 N is the passing requirement for the bonding strength of a terminal.

The result for the signal contacts is presented in Fig. 10(b). As expected, the absolute bonding strength of the signal contact was much lower than the power contacts owing to the smaller area of contact. The normalized bonding strength in MPa was found to be nearly identical for both types of contacts across all the temperature steps. Even the weakest recorded bond strength at 200 °C was found to exceed the bonding strength requirement outlined in IEC60068-2-21. As with the power contacts, the variation on the bonding strength at 200 °C was spread across a wider range. At the desired service temperature of 175 °C, the bonding strength was found to be nearly 200 N.

These pull test results showed that both the attach material and the process parameters were very well suited for a strong bond. This was expected to yield a favorable reliability performance, which will be presented in a subsequent section.

It was important to verify the robustness of the contact attach process, and the ability of the process to withstand a change in minor part variations. The baseplate and contacts for this test were obtained from a separate source. The geometric and material specifications of the parts were kept identical. Fig. 11 summarizes the results of this test.

Fig. 11(a) shows the results for the power contacts. The average bonding strengths were very similar to the previous results in Fig. 10(a). However, the confidence of the measurements was much higher and they were contained within a very narrow range of variation. Fig. 11(b) shows the results for the signal contacts and echoes the earlier results from Fig. 10(b). It can be concluded that the attach material and process were highly likely to repeatedly demonstrate a high bonding quality.

The next set of tests focused on subjecting the attach material to reliability testing. A good initial bonding strength is a necessary condition for superior reliability, but not a sufficient condition. It was important to obtain reliability data to be able to verify whether the material would be able to withstand the stresses associated with long term operation.

Thermal cycling and power cycling tests were chosen to simulate the worst-case scenario during field operations. C- SAM images for the die and substrate attach layers were captured before and after the tests to gauge degradation.

Thermal cycling was performed per IEC 60749-25, from −55 °C to 150 °C, for a total of 1000 cycles (1 hour per cycle). To our knowledge, these standards are more rigorous than any other contemporary power module manufacturer. The modules for this test were fully built to completion and confirmed to be electrically functional after the build process. A total of five samples were assembled for this test. C-SAM scans for the die and substrate attach layers, x-rays, and electrical characterization results were captured before and after the test. A “fail” condition was reached if the voiding after thermal cycling exceeded 25%.

Fig. 12 shows the results for the die attach layer. The images on the left side were taken before the test, and the images to the right were taken after 1000 cycles of thermal cycling. It was observed that the x-ray images were nearly identical and seemed unaffected as a result of the test.

In the post-cycling images, a circular ghost image could be seen. This is a shadow of a hole feature on the top of the module. The x-ray images captured post-cycling were of a fully built module with a housing having a circular hole. The housing was absent when the pre-cycling images were taken, and hence the ghost image was not seen for the images on the left. The substrate attach layer covered a much larger area compared with the die attach. This makes the substrate attach layer more susceptible to cracking under thermal stress. However, no cracking was observed via x-ray.

For the final test after thermal cycling, a contact pull test was conducted on some representative samples after 1000 hours of thermal cycling. Fig. 13 shows the results. The samples were tested at 175 °C to represent the most aggressive operating condition.

It can be observed from the results that there was some degradation in the strength of the power and signal contact attaches after rigorous thermal cycling. However, the minimum bonding strength of 40 N from the IEC60068-2-21 standards was met. This confirmed the high bonding quality demonstrated in Section IV. The thermal cycling test is an isothermal stress subjecting the entire module to the same temperature. For the next experiment, a power cycling test was performed, which subject the module stack to a sharp temperature gradient, with hot spots at the die location. This test was more representative of what a module would experience in real world application.

The power cycling test was performed across 45,000 cycles. The maximum junction temperature was maintained at 175 °C, with a junction to case temperature difference of 80 °C. The die were turned on for 2 seconds, and turned off for the next 4 seconds. A “fail” condition was defined as the thermal impedance of the module increasing by 20%, or the forward voltage of the body diode changing by 5%. A delamination of the die or substrate attach layer would directly manifest itself as a change in the thermal resistance of the module. Six completed modules were used for the power cycling test. The x-ray images of the modules are shown in Fig. 14. Similar to the thermal cycling test, there was no perceived change in the attach quality in any of the samples post power cycling. This was, of course, with the exception of the circular ghost image – and an explanation for this was provided in Section V, A.

As mentioned earlier, the thermal impedance of the module during the power cycling test was a good indicator of the state of the attach quality after testing. The change in the thermal resistance of all the tested modules was well within the 20% limit. All the tested modules passed the power cycling criteria outlined earlier in this section.

This study was aimed at identifying the best die attach material for aggressive SiC power module designs. The novel high-density HT-4000 power module from Wolfspeed was used as a vehicle for this test. A detailed procedure for the evaluation of high temperature die attach materials was presented in this paper. The candidate material under evaluation showed outstanding initial attach quality for both the die attach and substrate attach layers, in a single-step attach process. Process optimization showed a marked improvement in the attach quality. The optimized processes yielded very high initial bonding strengths for both die and contacts. The bonding strength was found to far exceed the recommended standards even at elevated temperatures. The attach layers were demonstrated to successfully pass rigorous reliability testing which stressed the die and substrate attach layers extensively.

The authors would like to acknowledge the assistance of the Regulatory and Compliance Engineering Group at Cree, Inc. for the contact pull tests presented in this study.