GaN-based high electron mobility transistors (HEMTs) have attracted much attention due to their high power density, low on-resistance, and high electron mobility^[1−3]. Besides, due to their high breakdown electric field, GaN-based devices have strong resistance to radiation and have been widely applied in the aerospace fields^[4−6]. The effects of ⁶⁰Co γ-ray radiation^[7−9] and proton radiation^{[10, 11]} on the performance of GaN HEMTs have been widely reported, and there are relatively few studies on the electrical properties of GaN HEMTs under electron radiation.

The electron radiation is one of the most common types of radiation in the earth's low orbit. The earth's radiation belt located above the equator is mainly composed of high-energy protons and electrons^[12], which is the main radiation source of semiconductor devices in the space radiation environment^[13]. Among them, the outer radiation zone located between 1 × 10⁴−6 × 10⁴ km above the earth's equator is mainly composed of electrons with an energy of 0.4−1 MeV, and extends about 60° to both sides of the earth's equator^[14]. The spacecraft orbiting the earth can be affected by electron radiation, and thus more attention should be paid to the impact on the semiconductor devices including GaN HEMTs^{[12, 15]}.

At present, some literature has reported the changes in the properties of GaN materials and devices after electron radiation. Sasaki et al. observed the degradation of carrier density and mobility in the GaN HEMTs under 2 MeV electron radiation, and attributed it to the defects in the GaN layer^[16]. Oh et al. reported that the gate leakage current decreased and drain−source current (I_DS) increased when the E-beam radiation power increased, which were caused by nitrogen vacancy neutralization and oxygen impurity removal^[17]. Hwang et al. studied the changes in electrical properties under 10 MeV electron radiation, suggesting that the radiation reduced the two-dimensional electron gas (2DEG) mobility and led to a decrease in the threshold voltage (V_TH)^[18]. Besides, Luo et al. reported an increase in gate current of GaN HEMTs at 0.8 and 1.2 MeV, and attributed it to the electron traps generated in the AlGaN barrier layer^[13]. Nevertheless, the degradation mechanism of GaN-based devices still required further investigation after electron radiation. In addition, the effect of radiation on device performance is related to various factors, such as device structure, fabrication process, radiation energy and dose^[16]. The results reported in previous literature may be different, so it is necessary to conduct in-depth research on the change in device performance under different radiation conditions.

In this paper, the electron radiation was performed on the D-mode GaN HEMTs with different doses. The degradation of electrical properties was mainly characterized, and the change in electrical parameters was discussed to explore the physical mechanism. In addition, the emission microscopy (EMMI) and scanning electron microscopy (SEM) techniques were used to locate the device damage. The effects of radiation dose on device performance were evaluated, which provided a basis for the device reliability in the radiation circumstance.

The devices used in this study were the depletion-type GaN HEMTs, and the structure was shown in Fig. 1. A 2 μm thick GaN buffer layer was grown on a SiC substrate by metal−organic chemical vapor deposition (MOCVD), followed by an epitaxial growth of 18 nm thick AlGaN on the buffer layer. The ohmic contact metal is Ti/Al/Pt/Au, and the Schottky contact metal is Ni/Au. The device is a ten-finger gate structure with a total gate width of 1.25 mm, in which the gate−source, gate−drain and gate length distances are 0.8, 2.85, and 0.25 μm, respectively. The device was fabricated on the AlGaN/GaN heterojunction, which can form 2DEG at the heterojunction interface by piezoelectric and spontaneous polarization^[19].

The 1 MeV electron radiation was carried out with the dose rate fixed at 1 × 10¹¹ cm⁻²·s⁻¹. A total of six unpackaged samples were used in the experiments, which were placed without biases during the radiation. The electrical properties were measured by the semiconductor parameter analyzer (Keithley 4200A) after each radiation. Fig. 2 presented the flow chart of the radiation experiments. Three samples were firstly radiated with a high dose of 5 × 10¹⁴ cm⁻², and the changes in electrical characteristics were compared before and after radiation. The results suggested that the device cannot be turned off at a lower gate voltage, and the three samples presented the similar failures.

Therefore, the radiation experiment was then performed on the remaining three samples with the dose reduced to 5 × 10¹² cm⁻², and the device did not fail under such conditions. Subsequently, the dose was gradually increased to 1 × 10¹³, 5 × 10¹³, and 1 × 10¹⁴ cm⁻². The total dose of the four radiation experiments was 1.65 × 10¹⁴ cm⁻², and the I_DS gradually became larger. In this study, the Sample A and B were used to illustrate the experimental results under high- and low-dose radiation, respectively.

Fig. 3(a) showed the transfer curves of Sample A before and after radiation at a dose of 5 × 10¹⁴ cm⁻² with the drain−source voltage (V_DS) of 1 V. It can be noticed that a part of I_DS (~116 mA) cannot be depleted even when the gate−source voltage (V_GS) was much lower than the V_TH (|V_GS| > |V_TH|). Besides, when the V_GS was larger than the V_TH, the I_DS increased and finally coincided with the results before radiation. It suggested that the gate still has a certain control ability to the channel current and did not fail completely^[20]. Besides, Fig. 3(b) showed the transconductance (g_m) before and after radiation, of which decreased from 271.3 to 232.2 mS, and its degradation was about 14.4%. It suggested that the gate control ability decreased significantly after high-dose radiation^[20], indicating that the gate may be damaged.

In addition, Fig. 4(a) presented the gate characteristics in logarithmic coordinates. It reflected that the forward gate current changed little after radiation, whereas the reverse leakage current increased from 8.20 to 27.5 μA at V_GS of −10 V, as the inset shown in linear coordinates. Besides, the output curves before and after radiation were shown in Fig. 4(b), and the increase in I_DS can be significant after radiation. The above results also suggested that the electron radiation has a great impact on the control ability of the gate, and the detailed discussions would be presented in the subsequent content.

To further investigate the gate damage after radiation, we performed the EMMI test on Sample A. The EMMI measurement is mainly used to determine the damage location of the device based on the photons generated by the electron−hole recombination under a certain bias voltage^[9], which is commonly used for failure point location and analysis in semiconductor devices. In this study, a V_GS of −7 V was applied and the undamaged GaN HEMT should be turned off without exciting a strong photon signal. The results in Fig. 5(a) showed that there was a significant luminous spot located at the gate position, which may lead to the increase in gate leakage current after high-dose radiation.

Subsequently, we used the SEM to further determine the gate damage. It can be seen in Fig. 5(b) that one of the gate fingers was significantly damaged after radiation, which was consistent with the EMMI results in Fig. 5(a) and may result in the degradation of electrical properties. Besides, the channel current cannot be controlled by the gate finger even when the lower gate voltage was applied on the radiated device, and thus we analyzed that the location of the radiation damage may be ascribed to the Schottky gate. The similar phenomenon of uncontrolled channel current of GaN HEMT after radiation has also been reported in previous study and can be attributed to the failure of gate contact^[20]. Besides, the leakage current increased significantly after electron radiation with higher dose in this study. The similar results of increased leakage current after radiation has also been reported^{[20, 21]}, and can be ascribed to the Schottky gate degradation after radiation.

When the applied V_GS was much lower than the threshold voltage (|V_GS| > |V_TH|), there was still current passing through the channel below the defect location, resulting in the uncontrollable current in Fig. 3(a) and increased leakage current in Fig. 4(a). When the device was in the semi-on state, the current under the damaged gate finger cannot be controlled, and it can be much larger than the current under the undamaged gate finger before radiation. Besides, the other gate fingers still maintained the ability to control the channel current, and thus the I_DS of the radiated device increased in Fig. 4(b). Compared with the unaged sample, the control ability of the gate was reduced and the peak value of g_m decreased in Fig. 3(b). When the device was turned on, the channel was fully opened and the saturation current was basically the same as that of the unaged device. Therefore, the I_DS was almost unchanged under the positive gate voltage in Fig. 3(a).

To further study the changes in electrical characteristics, the radiation dose was reduced to 5 × 10¹² cm⁻² and gradually increased in step to ensure that the Sample B did not fail after each radiation. Fig. 6(a) showed the transfer characteristics under different doses, suggesting that the I_DS in the linear region increased, and the inset showed that the V_TH was negatively shifted. It can be inferred that the positive charges in the AlGaN layer increased after radiation, and can be related to the total ionizing dose effect caused by electron radiation^{[12, 15]}. In addition, the device remained in the off state at a lower V_GS, which was different from the uncontrollable current under high-dose radiation in Fig. 3(a). The gate can maintain the control ability under low-dose radiation, which also proved that the device performance was closely related to the radiation dose. Fig. 6(b) showed the gate characterization at different radiation doses. It can be seen that both the forward and reverse gate current remained basically unchanged. It has been reported that the increase in gate current can be related to the electron traps created by the displacement damage effect^[13], and thus it can be inferred that the displacement damage effect may not be dominated in this study. It suggested that the gate of Sample B was almost unaffected by the low-dose radiation, which was quite different from the increase in leakage current after high-dose radiation in Fig. 3(a).

The output curves at different radiation doses were presented in Fig. 6(c), where the results at V_GS of −2 V were shown in the inset and the I_DS in the saturation region gradually increased. Besides, the OFF-state leakage current of I_D and I_G under high V_DS before and after electron radiation with lower dose was presented in Fig. 6(d). The V_GS was set as −6 V to ensure that the device was in the OFF-state and the leakage current was obtained with the V_DS increasing from 0 to 70 V. The similar measurements of OFF-state leakage current have also been conducted in previous studies^{[22, 23]}, and the variation trend and order of I_D and I_G can be credible. It can be noticed that both the I_D and I_G did not change after radiation with lower dose, which was also consistent with the unchanged forward and reverse gate current in Fig. 6(b). The above results suggested that the gate was almost unaffected by the low-dose radiation.

Unlike the gate damage under high-dose radiation, the above results presented the improvement of electrical characteristics under low-dose radiation from 5 × 10¹² to 1 × 10¹⁴ cm⁻², and the degradation and improvement of the GaN HEMT with the same order of radiation dose has also been reported in previous study^[20]. Besides, the result of improvement in device performance has also been reported in other literature, which is often attributed to the decrease in the density of traps in the AlGaN surface and the increase in 2DEG density (n_s) and mobility (μ)^{[15, 17]}. To further analyze the related mechanism, the n_s and μ of Sample B were calculated based on the above electrical properties under different radiation doses. The AlGaN/GaN HEMT can be described using the charge-control model^{[10, 24]}, and the n_s and μ of the 2DEG can be estimated using the following relation:

1ns=ϵ(VGS−VTH)qt,

2VDSIDS=RS+RD+LtWμϵ(VGS−VTH),

where ϵ is the dielectric constant of AlGaN, VTH is the threshold voltage, q is the electronic charge, and t is the thickness of the AlGaN barrier layer, R_S and R_D are the source and drain access resistances, and W and L are the gate width and length. The Eq. (2) can be regarded as V_DS/I_DS versus 1/(V_GS− V_TH), and the V_DS/I_DS is calculated in the linear region of output curves. The slope is related to the sheet carrier mobility, and the intercept is related to the source and drain access resistance R_S and R_D^[10]. The detailed calculation processes of n_s and μ with this method can be found in previous studies^{[10, 24]}.

With the above formulas, the calculated electrical parameters were listed in Table 1, indicating that both the n_s and μ increased after radiation. Specifically, the slight increase in n_s can be ascribed to the negative shift of V_TH after radiation. The increase in n_s was small (Δn_s = 0.14 × 10¹² cm⁻²) after the radiation dose of 1 × 10¹⁴ cm⁻², and the similar variation has been reported in previous studies. For example, Khanal et al. reported that the Δn_s was about 0.12 × 10¹² cm⁻² after proton radiation^[10]; Hu et al. reported that the Δn_s was about 0.3 × 10¹² cm⁻² after gamma-radiation^[24]. The calculated values of n_s and μ were close to the results reported in previous studies with the similar calculation methods^{[10, 24]}, indicating that the n_s and μ obtained in this study can be reliable. The above results confirmed our analysis that the mobility and 2DEG density were improved after radiation, which can be related to the improvement of electrical performance after low-dose radiation^{[12, 15, 25]}. Besides, the trap evolution after electron radiation may also lead to an improvement in its performance. During radiation, the high-energy electrons may interact with the device and the charges around the defects can be redistributed. It leads to an increase in the structure ordering and the decrease in trap density, and thus the electrical properties can be improved^{[5, 8, 26]}.

In summary, the experimental results suggested that the electron radiation may play a certain role in improving the performance of the device, and may provide an effective approach for device manufacturing and application. In particular, it should be noted that the low-dose electron radiation was carried out cumulatively on the same sample. The total dose was 1.65 × 10¹⁴ cm⁻² and the electrical characteristics were improved with increasing radiation dose. However, the gate leakage current became larger at a higher radiation dose of 5 × 10¹⁴ cm⁻². It can be noted that not all of the gate fingers were damaged with the above radiation dose, and thus the damage at this gate position may be random. Nevertheless, the device degradation with this radiation dose can be obvious and more efforts should be devoted to this issue. In addition, the above results showed that the GaN-based devices can be sensitive to electron radiation dose, and it is necessary to further study the changes of electrical characteristics with the dose between 1.65 × 10¹⁴ and 5 × 10¹⁴ cm⁻², and further investigate the turning point of device performance between this scope. Besides, the change in device performance with lower energy level may be even more significant than the results with higher energy level according to the results in previous reports^[13], and thus more efforts should be devoted to the analysis of the effects of electron radiation with different energy levels in further study.

In this paper, the 1 MeV electron radiation experiments of different doses were performed on the depletion GaN HEMTs, and the changes in electrical characteristics were studied. The experimental results showed that the device can be severely damaged at the highest dose of 5 × 10¹⁴ cm⁻² with the large gate leakage current. Combined with the SEM and EMMI tests, it is concluded that the failure was mainly due to the damage of gate finger under high-dose radiation, which lost the ability to control the channel current under the gate. In addition, during the radiation ranging from 5 × 10¹² to 1 × 10¹⁴ cm⁻², the I_DS gradually increased and the V_TH was negatively shifted. It is concluded that the improvement in device performance is mainly caused by the reduction of traps and the increase in electron mobility and 2DEG density. In this paper, the device performance and related mechanisms of GaN HEMTs were studied under different radiation doses, which provided a theoretical basis for the practical application of devices in extreme radiation environments. In addition, more experiments should be performed to analyze the effects of electron radiation on more kinds of GaN HEMTs with different radiation doses and energy levels in further study.