Technical Article

Understanding Failure Mechanisms in Test-to-Fail Methodology for eGaN Devices in Solar Applications

June 07, 2023 by Shengke Zhang

Modern solar panels are demanding increasingly higher power density and longer operating lifetimes. Solar applications, including power optimizers and panels with built-in microinverters, are popular with an increasing number of solar customers, where low-voltage GaN power devices are extensively used.

Greater than 25 years of reliable operation is a typical requirement for solar installations. The test-to-fail methodology stresses devices to fail quickly. By understanding the intrinsic underlying failure mechanisms, physics-based lifetime models can be developed to accurately predict the lifetime under all mission profiles [1-5]. In this report, we use these physical insights and apply them to the unique demands of solar applications.

 

Figure 1. EPC2212 time to failure vs. VGS at 25°C MTTF (and error bars) are shown for four different voltage legs. Image used courtesy of Bodo’s Power Systems [PDF]

 

Gate Stress

Representative discrete GaN devices (EPC2212) showed excellent long-term gate reliability. Failure analysis was conducted on multiple failures from the study, and a consistent failure mode was found between the gate metal and the metal field plate. By understanding the underlying failure mechanism, a first-principles model was developed to explain all observations. This model can be used to predict the lifetime under different gate biases, temperatures, and duty cycles. The lifetime equation is plotted against the measured accelerated data for EPC2212 in Figure 1. Figure 1 shows that EPC2212 has less than 1 ppm failure rate projected over more than 35 years of lifetime under continuous DC gate bias at the maximum rated gate voltage (VGS = 6 V). This projected result is also consistent with EPC’s field experience with gate failures.

 

Drain Stress

One common concern for GaN is dynamic on-resistance. This is a condition whereby the on-resistance of a transistor increases when the device is exposed to high drain-source voltage (VDS). By understanding the hot electrons trapping mechanism, a hard switching topology circuit was developed and implemented to accelerate this failure mechanism by providing more hot electrons at maximum rated VDS [2,6-8] and beyond. Using the characterization test results from this development, a physics-based lifetime model was developed to describe the dynamic RDS(on) effects in eGaN FETs under all bias and temperature stress conditions.

Flyback is one of the most used topologies for micro inverters in solar applications, where the EPC2059, a 170 V max VDS-rated product, is frequently selected by solar customers for such applications. Figure 2 shows an EPC2059 device that was operated under continuous hard switching at 136 V (80% of the max rated drain bias of 170 V) while the case temperature was modulated at 80°C, where 80°C is considered a nominal operation temperature for solar applications. The measured data and the corresponding model predict the RDS(on) increase due to continuous hard switching in 35 years is expected to be approximately 10%.

Another popular option for solar applications is to use a DC-DC converter in the primary stage (typically a full bridge) of a microinverter. This topology is frequently used in a power optimizer, which has been increasingly adopted by solar providers due to its superior efficiency. GaN devices such as 100 V-rated EPC2218, EPC2088, and EPC2302, among others, are a good fit for this application. Figure 3 shows the projected RDS(on) increase of an EPC2218 device is expected to be 10% in 35 years of continuous hard switching operation at 80 V, ambient temperature.

 

Figure 2. Projected RDS(ON) shift of EPC2059, a 170 V rated device, in 35 years of continuous hard-switching operation is expected to be approximately 10%. Image used courtesy of Bodo’s Power Systems [PDF]

 

Therefore, eGaN devices demonstrate good robustness in dynamic on-resistance with more than 25 years of life and beyond.

 

Figure 3. Projected RDS(ON) shift of EPC2218, a 100 V-rated device, in 35 years of continuous hard-switching operation is expected to be approximately 10%. Image used courtesy of Bodo’s Power Systems [PDF]

 

Thermo-Mechanical Stress

Thermo-mechanical reliability is another critical area of particular interest in solar applications. Solar panels are placed outside and experience significant ambient temperature change. A similar test-to-fail approach was used to study the board-level thermo-mechanical reliability of EPC2218A.

Three combinations of test conditions are studied, as shown below.

  • TC1 condition without underfill: −40°C to 125°C
  • TC2 condition without underfill: −40°C to 105°C.
  • TC1 condition with underfill: −40°C to 125°C, where the underfill manufacturer is HENKELS and the part number is ECCOBOND-UF 1173.

All parts were mounted on test coupons consisting of a 2-layer, 1.6 mm thick, FR4 board using SAC305 solder paste and water-soluble flux. A group of 88 devices was tested for each leg, and all three test legs used similar ramp rates and dwell times at the two temperature extremes. After every temperature cycling interval, electrical screening was performed, where exceeding datasheet limits was used to determine failures. Physical cross-sectioning and SEM inspection followed to further examine the electrical test failures. Solder joint cracking was found to be the single failure mode throughout all failures analyzed.

 

Figure 4. Weibull plots of temperature cycling results for EPC2218A. Image used courtesy of Bodo’s Power Systems [PDF]

 

Figure 4 shows Weibull failure distribution of the temperature cycling results. The failure distribution was analyzed using a 2-parameter Weibull distribution for each temperature cycling leg using maximum likelihood estimation (MLE) [9]. The fits are indicated by solid lines in the graph.

From TC1 (−40°C to 125°C) to TC2 (−40°C to 105°C) without underfill, a strong acceleration was found. Two primary failure mechanisms are responsible for the significant acceleration. First, the difference in ∆T of the two testing conditions leads to the acceleration of the solder fatigue failure mechanism [10,11]. However, this failure mechanism alone is insufficient to explain the acceleration observed. A second mechanism, the creep solder joint failure mechanism, is introduced. Creep is believed to be the main effect during the dwell period at the hot temperature extreme [11-16].

After 1800 cycles of TC1 (−40°C to 125°C) with underfill, no failures have been found to date. This shows that applying proper underfill material can significantly improve the thermo-mechanical capability of the chip-scale package devices. Based on the test results, a more general TC lifetime model was developed.

\[N=A\cdot f^{-a}\cdot\Delta T^{-\beta}\cdot exp{\Big(}\frac{Ea}{kT_{Max}}{\Big)}\,equation(1)\]

Where is the number of cycles to fail, ƒ is the cycling frequency, and α is the frequency exponent, at -1/3 [12-17]. This frequency term is to describe the frequency of usage. ΔT is the range of temperature change, and β is the temperature exponent. Since SAC305 solder is used, β is 2.0 [12-17]. The last variable is an Arrhenius term that models the creep failure mechanism, where Ea is the activation energy, k is the Boltzmann constant, and Tmax is the maximum temperature in Kelvin units (°K). By comparing the mean-time-to-fail between TC1 and TC2 without underfill, the Ea was calculated to be 0.2 eV.

In real-world applications, solar panels experience varying temperature profiles. As a result, a more general lifetime model is warranted to include all mission profiles. An empirical equation is therefore developed in equation 2.

\[\frac{1}{N_{Total}}=\frac{a}{N_{\Delta Ta}}+\frac{b}{N_{\Delta Tb}}+\cdot\cdot\cdot+\frac{i}{N_{\Delta Ti}}\,equation(2)\]

Where NTotal is the total calculated lifetime of the number of cycles, NΔTi corresponds to cycles-to-failure for the condition of ΔTi, and i is the fraction of time the device was operational under ΔTi.

Now let’s examine a real-world example to estimate the lifetime by applying different mission profiles. The first assumption is that the solar panels are installed in Phoenix, Arizona, where solar is well-suited for the climate that has long sun exposure but also demands stringent thermo-mechanical requirements. Using the year 2023 forecast as an example [18] and then adding 30 °C of device self-heating on top of each ambient mission profile, the projected lifetime of EPC2218A with underfill material at 0.1% failure rate is estimated to be approximately 42 years due to temperature cycling stress.

 

Test-to-Fail Methodology Summary

Making use of EPC’s 100 V-rated generation 5 product family with underfill for real-world solar applications vastly reduces thermal cycling reliability risk while giving excellent lifetimes that significantly exceed the expected 25 years.

 

References

1. Lidow, A., “GaN Transistors for Efficient Power Conversion, Third Edition”, John Wiley & Sons, 2019

2. Lidow, A., “GaN Power Devices and Applications”, 2021

3. Meneghini, M., et al. “GaN-based power devices: Physics, reliability, and perspectives”, J. Appl. Phys. 130, 181101, 2021

4. De Santi, C. et al., “Review on the degradation of GaN-based lateral power transistors”, Advances in Electrical Engineering, Electronics and Energy, Vol. 1, 100018, 2021

5. Stockman, A. et al., “Gate Conduction Mechanisms and Lifetime Modeling of p-Gate AlGaN/GaN High-Electron-Mobility Transistors”, IEEE Transactions on Electron Devices, PP(99):1-8, 2018

6. Lidow, A et al., “Intrinsic Failure Mechanisms in GaN-on-Si Power Transistors”, IEEE Power Electronics Magazine, vol. 7, no. 4, pp. 28-35, 2020

7. Zhang, S. et al., “GaN Reliability and Lifetime Projections”, CIPS 2022; 12th International Conference on Integrated Power Electronics Systems, pp. 1-7, 2022

8. Brazzini, T., et al., “Mechanism of hot electron electroluminescence in GaN-based transistors,” J. Phys. D: Appl. Phys. 49, 435101, 2016

9. Cramér, H., „Mathematical methods of statistics“, Princeton Univ. Press (1946)

10. JEDEC Standard, “Temperature Cycling”, Test Method JESD22- A104F, November 2020

11. Automotive Electronics Council, “FAILURE MECHANISM BASED STRESS TEST QUALIFICATION FOR DISCRETE SEMICONDUCTORS IN AUTOMOTIVE APPLICATIONS”, AEC-Q101-Rev E, March 2021

12. Norris, K. C., & Landzberg, A. H., “Reliability of Controlled Collapse Interconnections”, IBM Journal of Research and Development, 13(3), pp. 266–271, 1969

13. Vasudevan, V., and Fan, X., “An Acceleration Model for LeadFree (SAC) Solder Joint Reliability Under Thermal Cycling,” ECTC, pp. 139–145, 2008

14. Sun, F.Q., Liu, J.C., Cao, Z.Q. et al. “Modified Norris–Landzberg Model and Optimum Design of Temperature Cycling Alt.” Strength Mater 48, pp. 135–145, 2016

15. Lall, P., Shirgaokar, A., and Arunachalam, D. „Norris–Landzberg Acceleration Factors and Goldmann Constants for SAC305 Lead-Free Electronics.“ ASME.  J. Electron. Packag., 134(3), 031008, 2012

16. Deshpande, A., Jiang, Q., Dasgupta, A., and Becker, U., „Fatigue Life of Joint-Scale SAC305 Solder Specimens in Tensile and Shear Mode,“ 18th IEEE Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems (ITherm), Las Vegas, NV, USA, pp. 1026-1029, 2019

17. Cui, H., “Accelerated Temperature Cycle Test and Coffin-Manson Model for Electronic Packaging”, RAMS, pp. 556-560, 2005

18. “MSN weather”, https://www.msn.com/en-us/weather/monthlyforecast, January 2023

 

This article originally appeared in Bodo’s Power Systems [PDF] magazine

Featured image used courtesy of Adobe Stock