AlGaN/GaN high electron mobility transistors (HEMTs) are group III–V compound semiconductors that exhibit wide band gap (3.4 eV), high breakdown field (3.3 MV/cm), and good thermal conductivity (160 W/mK) characteristics [1–4]. Furthermore, the polarization at the junction between AlGaN and GaN naturally forms a two-dimensional electron gas (2-DEG) channel, enabling high electron mobility. As a result of these properties, compared to conventional Si and GaAs devices, AlGaN/GaN HEMTs are considered promising next-generation semiconductor devices for use in high-frequency and high-power applications, such as radio frequency (RF) switches and high-power RF transistors [5]. However, during high-voltage operations, a strong electric field forms along the 2-DEG channel, leading to the acceleration of electrons, which in turn obtain high kinetic energy. Consequently, the electrons that manage to acquire the energy required to overcome the energy barrier are trapped in the gate-drain access region, potentially degrading device performance. Moreover, in short-channel AlGaN/GaN HEMT devices, shortening the gate length to achieve improved RF characteristics leads to an intensified concentration of the electric field, which is likely to exacerbate the degradation caused by hot carriers [6, 7]. Considering this context, it is obvious that analyzing device degradation is crucial to ensure the reliability of short-channel GaN HEMTs.

To conduct this analysis, we induced hot carrier stress conditions in 0.15 μm short-channel AlGaN/GaN HEMTs using electroluminescence (EL) spectroscopy. Additionally, constant voltage stress was applied while maintaining the same drain voltage and power consumption conditions as hot carrier stress conditions to comparatively analyze the degradation of device characteristics under stress.

The devices analyzed in this study were short-channel HEMTs with a T-gate structure and a gate footprint length of 0.15 μm. The 0.15-μm T-gate, featuring two gate fingers and a gate width of 50 μm, was constructed using E-beam lithography by implementing a metal lift-off process. As shown in Fig. 1, the epi-wafer consisted of a 5-μm 4H-SiC substrate layer, a 0.2-μm GaN nucleation layer, a 2-μm GaN buffer layer, a 25-nm Al_25.5Ga_74.5N barrier layer, and SiN passivation. The gate-source and gate-drain distances were 1 μm and 2 μm, respectively.

EL signals are closely related to the generation of hot carriers owing to the recombination of electron-hole pairs [8–10]. For this study, EL spectroscopy was conducted at the Korea Electrotechnology Research Institute. As shown in Fig. 2(a), EL spectroscopy was performed to extract the voltage conditions for accelerating the generation of hot carriers. When different combinations of the gate and drain voltages were applied to the device, changes in the EL signal were observed due to changes in the electric field.

As shown in Fig. 2(b), a bell-shaped EL intensity-V_GS curve was attained, with peak EL intensity attained at V_D = 40 V and V_G = −3 V. To investigate the effects of hot carrier injection, constant voltage stress was applied for 7,240 seconds using Keithley’s 2651A and 2410 source meters, after which device degradation over time was analyzed by measuring DC characteristics during the application of stress.

Fig. 3 presents the DC characteristics before and after applying constant voltage stress at V_D = 40 V and V_G = −3 V points at which peak EL intensity was observed. The transfer characteristics before and after stress were measured at V_D = 1 V. Furthermore, the threshold voltage extracted through linear extrapolation from the fresh device was −3.2 V, while the maximum transconductance (g_m.max) value was 293.3 mS/mm. The output characteristics were measured at V_G = 0 V, with the maximum drain current (I_DS.max) being 897.9 mA/mm (V_D = 5 V). After the stress test, the threshold voltage achieved a positive shift of 0.36 V, while g_m.max decreased by 9.6% and I_DS.max reduced by 15%. The degradation of these device parameters resulted from the hot carrier injection, considering that peak EL intensity is a clear signature of hot carrier generation.

Fig. 4 illustrates the results of the device degradation caused by stress conditions under different EL intensities. These parameter changes were compared to those related to the fresh state of the devices. Each constant voltage stress was conducted on devices with the same structure that underwent the same batch process. When stressed at V_G = −4.5 V, the threshold voltage underwent a positive shift of 0.18 V from −3.1 V, while I_DS.max decreased by 11.5% from 873.8 mA/mm. Under stress conditions, a gradual change in the device parameters was observed over time. Furthermore, the positive shift in threshold voltage and the decrease in I_DS.max were exacerbated by the stress voltage at a stronger EL intensity. In fact, a significant degradation in g_m.max was observed under peak EL stress conditions.

Notably, to generate hot electrons, two requirements must be met: a high electric field and a sufficient number of accelerated electrons. In this context, although the electric field between the gate and the drain was higher at V_G = −4.5 V compared to V_G = −3.5 V, hot electron generation was hardly probable because a deeply pinched-off channel cannot provide enough electrons to be hot.

However, self-heating may also affect degradation characteristics, since the channel current in the two conditions considered in the previous tests were not identical. Therefore, to identify the dominant degradation mechanism, on-state stress was applied to the device, with the gate and drain voltages inducing the same power consumption as the stress condition at peak EL intensity.

Fig. 5 shows the changes in device parameters under on-state stress (V_G = 0 V, V_D = 8 V) at the same power consumption (6.4 W/mm) as the voltage condition at peak EL intensity (V_G = −3 V, V_D = 40 V). As V_GS increased from the deeply pinched-off condition, electron density in the channel increased and the electrons accelerated to be hot, leading to an increase in EL intensity. However, after surpassing the peak, the intensity decreased due to a reduction in the electric field between the gate and the drain [11]. After applying on-state constant voltage stress, no degradation in the device parameters was observed despite the sufficient presence of carriers. Therefore, the difference in the degradation observed in the two stress tests can be attributed to the difference in EL intensity, which relates to whether hot electrons exist at the same channel temperature.

The results of the electric field distribution under stress conditions, obtained using Silvaco ATLAS Technology Computer-Aided Design (TCAD) software, are shown in Fig. 6. When comparing the V_G = −4.5 V and V_G = −3 V conditions at V_D = 40 V, the former exhibited a relatively larger electric field. However, device degradation was less in the V_G = −4.5 V condition due to the almost complete absence of carriers compared to the V_G = −3 V condition. Additionally, when considering the V_G = 0 V and V_D = 8 V condition, which consumed the same power as the V_G = −3 V and V_D = 40 V condition, under on-state stress, although the channel was formed and carriers were sufficient, a significant difference in the electric field of V_GD was observed, which was crucial for the generation of hot carriers. This suggests that the generation of hot carriers under peak EL intensity conditions led to the degradation of device performance, further indicating that hot carrier generation necessitates the simultaneous presence of carriers and a strong electric field. As shown in Fig. 7, when a high voltage was applied to the drain, a strong electric field was formed in the channel, leading to the acceleration of electrons, which in turn caused impact ionization and resulted in the generation of electron-hole pairs. Subsequently, the energetic electrons were injected into the AlGaN barrier and trapped in the GaN buffer layer. Such trapped hot carriers induce channel depletion, leading to degradation of device performance, such as a decrease in g_m.max [12, 13].

In this work, constant voltage stress tests were conducted based on varying EL signal intensity and power consumption to investigate the degradation caused by hot carrier injection in short-channel AlGaN/GaN HEMTs. The results showed that the most pronounced degradation in device performance occurred under conditions of peak EL intensity, evidenced by a positive shift in the threshold voltage, a decrease in I_DS.max, and a reduction in g_m.max. This implies a close correlation between EL intensity and the degradation caused by hot carriers.