Introduction

Gallium nitride (GaN) high electron mobility transistors (HEMTs) have been widely utilized in fields such as power electronics and radio frequency, including power amplifiers (PA), low noise amplifiers (LNA), and switches, due to their high breakdown fields, along with high channel carrier densities and high drift velocities1,2,3,4,5. The unique performance of GaN HEMTs also drives their applications in radar6, electronic warfare systems7, fifth-generation (5G) wireless communication8,9, and more. On the other hand, it should be noted that with the continuous advancement of pulse power technology and the extensive application of high-power radars, communication transmitters10, and high-power microwave equipment11,12, the current electromagnetic environment is not only becoming increasingly complex but also seeing a constant increase in power density. Under such circumstances, electronic systems including GaN HEMTs will undoubtedly be seriously interfered with and even damaged13,14,15,16.

There are already numerous literary works on the reliability and failure mechanism of GaN HEMTs. However, most studies still focus on their intrinsic reliability, such as current collapse17,18, high field degradation effect19,20,21, metal thermal stability22, and self-heating effect23,24,25. Few pieces of literature have reported on the failure mechanisms of GaN HEMTs caused by high-power microwave pulses, especially under high-power microwave injection at the drain terminal. At present, the only some relevant reports also focus on the case of microwave pulse injection at the gate terminal, which mostly occurs in the usage scenario of low noise amplifiers. Thermal breakdown model14,26,27, mechanical damage model16, and even thermo-electro multi-physics coupling failure model15 are used to explain the damage mechanism of GaN HEMTs caused by high-power microwave injection at the gate terminal. However, in applications like transmit/receive (T/R) modules and radio frequency modules within communication base station systems, once external high-power microwaves are coupled via antennas, there is a likelihood that they will be injected in reverse into the radio frequency output end of the power amplifier. In such a scenario, microwave signals will be applied to the drain electrode of the GaN HEMT device rather than the gate electrode. Up to now, there remain relatively few studies focusing on the failure mechanism of GaN HEMT under these circumstances.

In this work, high-power microwave pulses were injected inversely into GaN-HEMT power amplifiers through the output port (i.e., the drain terminal of the HEMT). A new failure mode of our GaN HEMTs has been observed, which is quite different from the failure modes of the aforementioned mechanisms. To explore this new failure mode and its underlying mechanism, we conducted research on failure analysis and numerical simulation.

Experiment

The integrated circuit of the power amplifier under test (EUT), namely TGF2023-2-0128, was fabricated by Qorvo using 0.25 μm high-power GaN/SiC HEMT technology. The structure of this power amplifier was identical to that described in Ref.13. The impedance of the GaN HEMT was matched to 50 Ω through a multistage L-C low-pass matching network. Typically, the power amplifier provided a saturated output power of 6 watts in the S-band of the electromagnetic spectrum. In the experimental study, the gate-to-source voltage was set at -5 V and the drain-to-source voltage was set at + 28 V.

The experimental methods for high-power microwave effects generally include irradiation and injection methods. The injection method involves directly injecting microwave pulses into the target through a transmission cable. This method is often utilized to study the effect mechanism and effect law as well as to calculate the effect threshold of the entire system29. An experimental platform for the reverse injection effect of high-power microwave, with real-time monitoring of the output waveform of the GaN microwave power amplifier, was built by using a circulator. The experiment system consists of a high-power microwave generation system, the equipment under test (EUT), an accompanying test system, and an effect monitoring system. The schematic of this setup is shown in Fig. 1.

Fig. 1
figure 1

Schematic diagram of high-power microwave reverse injection experiment system.

The testing process was carried out in two steps. In the first step, it was ensured that the GaN HEMT power amplifier was in a normal state by adjusting the output of the signal source (Agilent E8257D). The power amplifier was in the pulse output mode, with the period set at 2 µs, the pulse width set at 500 ns, and the output power approximately 6 watts. The output signal was fed to the oscilloscope (LeCroy WavePro 640Zi) through the circulator (from Port 1 to Port 2) and an attenuator to monitor the waveform. The change in waveform could be used to evaluate the health status of the GaN HEMT power amplifier. In the second step, high-power microwave pulses were received by the receiving antenna, then adjusted by a high-power adjustable attenuator and a dual directional coupler, and finally injected inversely into the GaN HEMT power amplifier through the circulator (from port 3 to port 1). The high-power adjustable attenuator was used to adjust the peak power of the injected microwave pulses, and the dual directional coupler, combined with a power meter (Keysight N1912A), was used to monitor the power of the input and reflected signals.

A self-made time-domain synchronization control system and a signal source were employed to control the pulse width, repetition frequency and pulse number of the microwave pulses. A typical demodulated waveform injected into port 3 of the circulator is shown in Fig. 2. The actual waveform can be regarded as a sine wave that lasts for 100 ns in the time domain.

Fig. 2
figure 2

Typical waveform reversely injected into the GaN HEMT power amplifier via the circulator.

Results and discussion

In the initial stage of the experiment, the high-power microwave pulses had a pulse width of 100 ns, a repetition rate of 20 Hz, and 100 pulses per pulse train. Regarding the peak power of the high-power microwave pulses, it gradually increased from 30dBm (about 1 W), with each pulse train increasing by about 1dB. When the peak power reached approximately 40dBm (about 10 watts), the output signal of the GaN HEMT power amplifier began to be interfered with. The interference time and maximum amplitude grew as the peak power of the high-power microwave pulses increased, which was entirely consistent with what was described in Ref.13. The previous research conducted by our group (Reference13) has laid a solid foundation for the current study. Specifically, that research explored the interference effects induced by microwave pulses on the GaN-HEMT power amplifier prior to its being damaged, and this has contributed to a better understanding of the effect mechanism in our present study. When the peak power was continuously increased to 53dBm (about 200 watts), the output signal on the oscilloscope vanished, and the DC power supply used for drain-source bias reported an error, indicating that it could no longer apply a + 28 V voltage to the device. It was preliminarily determined that the GaN power amplifier had burned out.

To confirm the health status of the GaN power amplifier, two GaN power amplifiers, one without testing and the other after testing, were selected for decapsulation and optical microscope analysis. Figure 3 (a) presents the actual circuit diagram of the GaN HEMT power amplifier used in the experiment. Figure 3 (b) shows the GaN HEMT together with its peripheral impedance matching network. The structures of the GaN HEMTs are shown in Fig. 3 (c) and (d). As shown in Fig. 3 (d), the GaN HEMT was almost completely damaged after the reverse injection of high-power microwave pulses with a peak power of 200 watts.

Fig. 3
figure 3

(a) Topography of GaN HEMT power amplifier. (b) GaN HEMT and its peripheral impedance matching network. (c) GaN HEMT in normal condition before testing. (d) GaN HEMT in damaged state after testing.

In such a serious case of burnout, a lot of useful information will be concealed, and it is difficult to analyze the physical process of the damage. Based on this, we hope to obtain GaN HEMTs with relatively light damage by adjusting the pulse parameters of high-power microwave, such as the number and width of the pulses. In the first step, the number of reverse injection pulses was reduced from 100 to 50, then to 20, then to 10, then to 5, and finally to 1 in turn. Under the condition that other pulse parameters remained unchanged, the GaN HEMTs were basically damaged, and the damage morphologies were similar to that shown in Fig. 3 (d), without any sign of damage attenuation. In the second step, with the pulse number maintained at 1, the pulse width of high-power microwave was successively reduced from 100 ns to 50 ns and then to 20 ns. Once again, the same phenomenon occurred, and the GaN HEMTs were still completely damaged. The above damage process clearly demonstrates an electric field breakdown mechanism that is entirely different from the reported damage models caused by high-power microwave as mentioned in Ref.14,15,16,26,27,30,31,32,33. It appears that the damage energy of the GaN HEMT does not primarily originate from high-power microwave pulses.

To verify the physical damage process, more detailed experiments and physical analyses with a narrower pulse width were carried out. Firstly, a high-power microwave pulse with a full width at half maximum (FWHM) of approximately 6.5 ns was obtained through the waveguide breakdown structure shown in Fig. 4(a), and its waveform is presented in Fig. 4(b). By adjusting the attenuation value of the adjustable attenuator, the peak power of the high-power microwave pulse was adjusted back to 53 dBm. Subsequently, a high-power microwave pulse was injected in reverse into the output port of the GaN HEMT power amplifier. It was observed that the GaN HEMT remained damaged after the effect experiment, as depicted in Fig. 5(a).

Fig. 4
figure 4

(a) Appearance schematic of the waveguide breakdown structure. (b) Pulse waveform diagram of the high-power microwave following waveguide breakdown.

To conduct a more in-depth analysis of the physical damage process, the dual beam focus ion beam (FIB) cross-section analysis technique (FEI Helios 660) was employed to examine the failures of the GaN HEMT. The cross-sectional views depicting the damage morphology of the GaN HEMT are presented in Fig. 5(c) and 5(d). For comparison purposes, a partial tilted side view of the GaN HEMT prior to testing is shown in Fig. 5(b). By observing the damage morphologies of the GaN HEMT, it can be discerned that during the damage process, the electrode metal melted and diffused into the crystal material. Additionally, the AlGaN barrier layer was damaged, and even the drain electrode bonding line was burned out.

Fig. 5
figure 5

(a) Damage morphology of the GaN HEMT after testing. (b) Tilted side view of a new GaN HEMT prior to testing. (c) and (d) Cross-sectional views of the GaN HEMT at different locations.

Obviously, it is impossible to cause such serious damage to GaN HEMT solely by the energy of a single high-power microwave pulse. Hence, we infer that the high-power microwave pulse is merely the inducement for the damage process. The high voltage coupled to the drain electrode may stimulate the device to produce a local short circuit. Subsequently, the GaN HEMT power amplifier will be completely burnt out mainly by the energy provided by the external bias voltage of the device itself. This mechanism can be called the stimulation burn-out effect. High-voltage TCAD simulations were conducted using the JEMS-CDS Device31,34. We simulated an AlGaN/GaN HEMT loaded with high-power microwave coupling voltage. Theoretical and experimental studies show that the voltage signal caused by a narrow-band high-power microwave pulse entering a ribbon cable is approximately a sinusoidal voltage signal with low damping35. For general narrow-band high-power microwave pulses with low frequencies, the approximate calculation of high-power microwave pulses can be made as follows:

The incident power of a high-power microwave pulse can be approximately expressed as

$$\:{P}_{in}={\left(\frac{\frac{{V}_{m}}{2}}{\sqrt{2}}\right)}^{2}/{R}_{L}={{V}_{m}}^{2}/8{R}_{L}$$
(1)

Here, Vm is the amplitude of the coupling voltage, and RL is the matching impedance of the GaN power amplifier.

The simulations take into account strain piezoelectric polarization, impact ionization, Shockley–Read–Hall and Auger recombination, high-field saturation, carrier-dependent lifetimes, avalanche, and thermionic models. Regarding the impact ionization model, the default van Overstraeten–de Man model is chosen, and the model parameters for electron impact ionization coefficients that depend on mole fraction are adopted as published in36. Interface charges were calculated following the work of Ambacher et al.37. The material structure of the simulated device, arranged from bottom to top, comprises a 15 μm-thick SiC substrate, a 3 μm-thick GaN buffer layer and a 20 nm-thick AlGaN barrier layer. With regard to the device’s geometrical dimensions, the gate length is fixed at 0.25 μm, while the gate-source spacing is 1 μm and the gate-drain spacing is 3 μm. The simulated potential profile and electric field distribution for the AlGaN/GaN HEMT biased to the voltage induced by high-power microwave are presented in Fig. 6. The damage region of the AlGaN/GaN HEMT shown in Fig. 5(c) is in excellent agreement with the strong electric field region shown in Fig. 6(b), and a large abnormal region emerges below the drain electrode.

Fig. 6
figure 6

Simulated potential profile (a) and electric field distribution (b) for the AlGaN/GaN HEMT biased with the voltage induced by high-power microwave.

Based on the foregoing experimental and simulation outcomes, the damage physical process of the AlGaN/GaN HEMT induced by high-power microwave is depicted in Fig. 7. Firstly, the reverse injection of high-power microwave is coupled to the drain electrode of the device, which can be equivalent to a sinusoidal voltage signal with a large amplitude. As can be seen from the simulation results in Fig. 6(b), the high voltage induces the formation of a critical breakdown electric field region on the side of the gate biased towards the drain electrode and below the drain electrode. The carriers in the high-field region are accelerated to possess sufficient energy to collide with the lattice and generate electron-hole pairs. The newly generated electrons and holes will be accelerated, and their collision with the lattice will generate new electron-hole pairs. This will cause a sharp increase in the leakage of the device and lead to avalanche breakdown. In addition, due to the lattice mismatch between the substrate and buffer layer of the AlGaN/GaN HEMT device, there are a large number of defects in the material38, which is helpful for the formation of leakage channels. When the leakage of the buffer layer is excessive, the device may break down in advance, which is the reason for the formation of the burning morphology shown in Fig. 5(c). After the high-power microwave signal disappears, since the bias voltage of the device still exists and the gate-drain voltage is still as high as 33 V, the leakage channel between the gate and drain electrodes will still remain until the device is completely burned. This is the reason why a single high-power microwave pulse can completely burn the AlGaN/GaN HEMT power amplifier.

Fig. 7
figure 7

Schematic diagram of the damage process in the AlGaN/GaN HEMT power amplifier caused by high-power microwave.

Conclusion

In conclusion, we have conducted an investigation into the damage effects and mechanism of the AlGaN/GaN HEMT power amplifier induced by high-power microwave. It was discovered that a single high-power microwave pulse with a nanosecond-level pulse width could completely damage the GaN HEMT power amplifier. Numerical simulations revealed that the mechanism underlying this effect is that the high voltage generated by the coupling of high-power microwave pulses induces avalanche breakdown near the gate and below the drain of the device, thereby forming a leakage channel. Since the power amplifier is an active device, upon the disappearance of high-power microwave pulses, the device’s own bias voltage continues to utilize this leakage channel to extend the burned area of the device until the device is entirely damaged. This is a typical stimulation burnout effect. This discovery holds great significance for the protection and reinforcement design of RF devices as well as the application of high-power microwave.