Determining the Lifetime of Gallium Nitride Devices
This article provides information on the challenges in determining the lifetime of gallium nitride devices and the methods used to extract lifetime data. These methods include gate lifetime, switching circuit-based lifetime, and mission profile lifetime extraction.

The JEDEC JESD47 standard and other similar standards have been used for decades to establish criteria for lifetime extraction for Si devices. This method is founded on knowledge of the failure modes of silicon-based devices and statistical calculations to ensure accuracy with small samples.

However, these conventional qualification tests do not take into account the switching situations of power devices. Static, high-voltage biases, and high temperatures are the only conditions used in the vast majority of the JEDEC-mandated tests.


What are the challenges in determining the lifetime of gallium nitride?

The reliability of performing static stress testing to determine the device lifetime is an open subject for GaN devices because of the various failure types. Several recent publications have investigated how to calculate GaN device lifetime during application-use switching conditions.

This article details investigations on gate lifetime, drain switching lifetime, and mission-profile lifetime extraction as shown in Fig. 1. Lifetime determining methods for Gallium Nitride
Fig. 1. Lifetime determining method

Gate Lifetime

Si and SiC Devices

The MOS gate used in traditional Si and SiC power transistors has a limited lifetime because of the oxide's reliability. At high temperatures, gate oxides in Si and SiC devices have been shaped to exhibit sufficient intrinsic lifetime.

The high-temperature gate bias (HTGB) test and the time-dependent dielectric breakdown (TDDB) test are often used to measure the MOS's reliability and durability. Both of them follow the JEDEC JESD92 standard.


GaN MIS-HEMT

Except for the good HTGB qualification data from vendors, very few gate lifetime studies have been reported for commercial D-mode GaN MIS-HEMTs used in direct-drive and cascode devices. Research devices using nitride dielectrics or Al2O3 for the gate insulator have been the primary basis for most MIS-HEMT publications. The TDDB behavior may be seen in the gate failures of most of these MIS-HEMTs.


SP-HEMT

As the SP-HEMT has become a standard device platform in many foundries, it has become the primary focus of gate lifetime studies in recent years. The difference between the average drive voltage and the highest voltage an SP-HEMT can handle can be as low as 1 V. Therefore, it is very important that the design of the SP-HEMT driver prevents gate-to-source bias overshoot caused by parasitic gate-loop inductance or commutation crosstalk. In addition, a Schottky-type p-gate with a floating p-GaN layer is not completely insulating.


DC Stress Test in SP-HEMT

The gate lifespan of SP-HEMTs has been investigated using a number of different approaches. There has been extensive use of dc tests subjected to either a constant voltage stress or a step voltage stress. The comparatively long-term parametric alterations following an off-state dc stress are typically characterized by the dc stress test. The parametric alterations that recover during the post-stress measurement time, which is typically at least 10 s, cannot be captured. Therefore, the measured shifts are not an accurate representation of those that occur during transitory switching.


I-V Pulse Test in SP-HEMT

Shortly (down to 1 μs) following the switching stress, the pulse I-V test can detect dynamic parametric alterations. Recently, the breakdown and dependability of gates have been investigated using a square-wave pulse I-V test. Pulse I-V tests can provide data for lifetime extraction that is identical to that received from dc tests.

It has been found that the effective gate lifespan (the product of the duty cycle and total lifetime) grows with temperature and depends only weakly on frequency.

In contrast, as the frequency hits 1 MHz, the effective gate lifespan declines with both increased frequency and duty cycle. This can be understood due to the fact that high-frequency gate-to-source bias switching places a strain on the electrostatics in p-GaN.


Circuit Test in SP-HEMT

The presence of many dependencies indicates that the switching strategies have a significant influence on the gate lifespan of SP-HEMT. The potential consequence of this is that the gate lifetime determined by dc tests may not be applicable to the operational performance of devices in power converters.

A recent study has introduced a circuit-based approach for assessing the gate lifetime in the context of inductive load switching. Circuit tests make it possible to characterize how a device's parameters shift during application-use circumstances.

The difficulties in conducting circuit tests revolve around determining the parametric shift of interest in situ while causing the least amount of disruption to the converter's operation.The parametric shifts analyzed in single-event and steady-state switching should also be carefully considered because they may differ due to the different trapping states.

This method addresses a condition that cannot be effectively evaluated using traditional dc or pulse I-V methods. Because parasitic-induced gate overshoot has a resonance-like quality, this technique combines inductive switching in the drain-source loop with gate ringing that has a pulse width as short as 20 ns.

The MOS gates of silicon insulated-gate bipolar transistors (Si IGBTs) and silicon carbide metal-oxide-semiconductor field-effect transistors (SiC MOSFETs) do not have these dependencies. The use of this method to describe the switching lifespan in addition to the failure boundary of the gate is desirable.


Causes of Failure in SP-HEMT

Finally, the failure mechanism of SP-HEMT continues to be debated amidst conflicting claims of its time-dependent degradation. Increases in gate-leakage current are consistently reported as a cause of device failure, with various explanations offered for this phenomenon.

Some attribute it to the high electric field in the metal/p-GaN Schottky junction or AlGaN barrier layer, while others have pointed to the creation of a percolation pathway. In a commercial SP-HEMT, impact ionization in p-GaN and the subsequent breakdown of the Schottky contact on p-GaN can explain why the failure boundary has a positive temperature coefficient. This is different from the standard percolation-induced TDDB hypothesis.

It appears that the SP-HEMT gate stack architecture has longevity issues, especially with quick switching. Improved gate architectures at the device level and driver approaches at the circuit level are the subject of continuing research.


Switching Circuits Based Lifetime Determination

The gate and drain-source loops of a GaN HEMT both greatly benefit from the use of switching circuits in order to extract the HEMT's lifetime. Since voltage is a significant lifetime accelerator, HTRB tests at several drain biases are commonly employed for traditional Si transistor lifetime extraction.

There is some doubt about the practicality of HTRB-based lifetime extraction due to the unique and complex failure processes in GaN HEMTs. In recent research, the lifetime of GaN HEMTs has been looked at in application-use switching situations that include both hard-switching and soft-switching stresses during the converter's operation.

Some of these studies employ the CIS circuit with the HSW turn-on stresses rather than the turn-off stresses, which is similar to the overvoltage stress testing. To mimic the switching behavior observed in boost converter applications, the I-V switching locus makes use of the HSW turn-on event in a CIS circuit. A lifetime plot is generated by repeatedly cycling through these switching events until failure.

Comparisons are then made between these experimental data and the simulated lifespan metrics, which are derived from simulations in which switching energy is the key parameter. It is then used as the baseline for the lifetime of GaN for different applications.


Mission-Profile Lifetime

There have been the latest modifications to the mission profile-based methodology used to extract the lifetime of electronic systems. For instance, the onboard chargers for HEVs and EVs have a much higher lifetime requirement compared to conventional internal combustion engines.

Other pulsed power applications, such as medical devices, have power cycles on the order of milliseconds to charge and discharge the load. System-level solutions to extract the lifetime can be implemented once the failure mechanism is understood. The consideration of system-level variables is often required for lifetime extraction based on a mission profile. System-level variables often include things like packaging, system integration, harmonics, and electromagnetic interference (EMI) from a business perspective.

GaN device reliability is already a concern due to the wide range of packages needed to support different chip sizes. Si devices from many manufacturers now have a common footprint, making them easier for users and vendors. However, there is still no standardization on the footprint of GaN devices between manufacturers, which is inconvenient for the end user and makes for an unstable supply chain. This diversity in packaging necessitates further evaluation of each new package's reliability by suppliers.

Built-in protection for gate drivers is a useful feature that can help speed up the market adoption of modern GaN devices. A combination of overcurrent, overtemperature, and undervoltage cutoffs is ideal for providing system-level protection. System-level integration that emphasizes a low-inductance design is also crucial for reducing undesirable symptoms like ringing and voltage overshoot.

Finally, GaN devices have a significantly smaller gate charge than Si devices. New and effective EMI solutions are needed because of the increased dv/dt and di/dt in GaN device operations. It also leads to additional harmonics and EMI problems at the system level. It is currently unclear how EMI and harmonics affect the longevity and dependability of GaN devices.

Summarizing the Key Points

  • The reliability of performing static stress testing to determine device lifetime is an open subject for GaN devices due to various failure types.
  • The JEDEC JESD47 standard and similar standards have been used for lifetime extraction for Si devices, but these conventional qualification tests do not account for switching situations in power devices.
  • Investigations have been conducted on gate lifetime, drain switching lifetime, and mission-profile lifetime extraction.
  • High-temperature gate bias and time-dependent dielectric breakdown tests are used to measure MOS reliability and durability.

Reference

Kozak, Joseph Peter, Ruizhe Zhang, Matthew Porter, Qihao Song, Jingcun Liu, Bixuan Wang, Rudy Wang, Wataru Saito, and Yuhao Zhang. “Stability, Reliability, and Robustness of GaN Power Devices: A Review.” IEEE Transactions on Power Electronics 38, no. 7 (July 2023): 8442–71. https://doi.org/10.1109/tpel.2023.3266365.