Total Ionizing Dose Effects on the Threshold Voltage of GaN Cascode Devices

GaN devices are nowadays attracting global attention due to their outstanding performance in high voltage, high frequency, and anti-radiation ability. Research on total ionizing dose and annealing effects on E-mode GaN Cascode devices has been carried out. The Cascode device consists of a low-voltage MOSFET and a high-voltage depletion-mode GaN MISHEMT. Cascode devices of both conventional processed MOSFET and radiation-hardened MOSFET devices are fabricated to observe the TID effects. Experiment results indicate that, for the Cascode device with conventional processed MOSFET, the VTH shifts to negative values at 100 krad(Si). For the Cascode device with radiation-hardened MOSFET, the VTH shifts by −0.5 V at 100 krad(Si), while shifts to negative values are 500 krad(Si). The annealing process, after the TID experiment, shows that it can release trapped charges and help VTH recover. On one hand, the VTH shift and recover trends are similar to those of a single MOSFET device, suggesting that the MOSFET is the vulnerable part in the Cascode which determines the anti-TID ability of the device. On the other hand, the VTH shift amount of the Cascode device is much larger than that of a previously reported p-GaN HEMT device, indicating that GaN material shows a better anti-TID ability than Si.


Introduction
For decades of development, silicon-based MOSFET devices have reached their ability limits [1].With the rapidly increased demands of power electronics, novel materials, such as GaN and SiC, are nowadays becoming more and more popular due to their capabilities of high power, temperature, frequency, and radiation tolerance [2][3][4].GaN devices are widely considered to be excellent candidates for microwave use [5][6][7][8][9][10][11][12][13][14].Recently, GaN materials have also become promising candidates for power use due to their advantages of high voltage and rapid switching time [15].Nowadays, enhancement-mode (E-mode), or the normally-off p-GaN gate HEMTs, have become the most commercial type as its threshold voltage is above 0 V, where the safety of the circuit is guaranteed.However, low-gate breakdown voltage [16], low-threshold voltage [17], and current-collapse effect [18] are still restricting the direct replacement of MOSFET.Especially, the gate-to-source breakdown voltage of p-GaN gate HEMTs is normally 6 V, while the conventional output of the driver chip in the power system is usually 0~10 V.At present, Cascode devices can be the temporary substitution for its ability of high-gate breakdown voltage and high-threshold voltage, despite the fact that the switching time is slower due to the existence of Si MOSFET.
In this study, we fabricated high-voltage Cascode devices combined with depletionmode (D-mode) GaN MISHEMT (Metal-Insulator-Semiconductor High Electron Mobility Transistor) and conventional processed MOSFET or radiation-hardened MOSFET.The total ionizing dose (TID) effect of the Cascode device is analyzed, and comparisons between both types of devices are made.As monitoring the negative shift of the threshold voltage of the MOSFET is well known and the standard process under TID-irradiation, we found out that Cascode devices share a similar phenomenon.

Device Structure and Test Results
A Cascode device combines a high-voltage D-mode GaN MISHEMT device and a lowvoltage E-mode Si MOSFET.The high-voltage GaN MISHEMT brings high drain-to-source breakdown voltage.While the low-voltage MOSFET brings high gate breakdown voltage, due to the existence of SiO 2 , which, on the other hand, leads to more severe TID effects.The schematic diagram and the package of the Cascode device are shown in Figure 1.
threshold voltage, despite the fact that the switching time is slower due to the existence of Si MOSFET.
In this study, we fabricated high-voltage Cascode devices combined with depletionmode (D-mode) GaN MISHEMT (Metal-Insulator-Semiconductor High Electron Mobility Transistor) and conventional processed MOSFET or radiation-hardened MOSFET.The total ionizing dose (TID) effect of the Cascode device is analyzed, and comparisons between both types of devices are made.As monitoring the negative shift of the threshold voltage of the MOSFET is well known and the standard process under TID-irradiation, we found out that Cascode devices share a similar phenomenon.As shown in Figure 1a, as a whole, the source of the MOSFET is the source of the Cascode device.The gate of the MOSFET is the gate of the Cascode device, and the drain of the GaN HEMT device is the drain of the Cascode device.Inside the Cascode device, as shown in Figure 1b, the back of the MOSFET is the drain, which is soldered to the bottom of the packaging case.The source of the MOSFET is connected to the gate of the GaN MISHEMT device.The source of the GaN MISHEMT device is connected to the bottom of the device, which is also connected to the drain of the MOSFET.The MOSFET device chip size is 3.4 mm × 3.3 mm, and the GaN MISHEMT device size is 2.3 mm × 1.8 mm.

A
To better understand the influences of the TID effects, two types of Cascode devices are prepared.Type A device consists of a conventional processed (not radiation-hard- As shown in Figure 1a, as a whole, the source of the MOSFET is the source of the Cascode device.The gate of the MOSFET is the gate of the Cascode device, and the drain of the GaN HEMT device is the drain of the Cascode device.Inside the Cascode device, as shown in Figure 1b, the back of the MOSFET is the drain, which is soldered to the bottom of the packaging case.The source of the MOSFET is connected to the gate of the GaN MISHEMT device.The source of the GaN MISHEMT device is connected to the bottom of the device, which is also connected to the drain of the MOSFET.The MOSFET device chip size is 3.4 mm × 3.3 mm, and the GaN MISHEMT device size is 2.3 mm × 1.8 mm. To better understand the influences of the TID effects, two types of Cascode devices are prepared.Type A device consists of a conventional processed (not radiation-hardened) MOSFET device and a D-mode GaN MISHEMT device, while Type B device consists of a radiation-hardened MOSFET devices and a D-mode GaN MISHEMT device in the same chip size.
For the radiation-hardened MOSFET devices, there are several special processes to enhance the anti-TID ability, as we reported before [19].
(1) Gate oxide growth atmosphere control.The charged defects of the gate oxide layer of MOSFETs include two parts: charges introduced by manufacturing processes and charges introduced by TID irradiation, as shown in Figure 2. To obtain a better quality dielectric layer, we arrange the gate oxide growth process after the high-temperature drivein process to prevent the possible impurities in the silicon from diffusing into the gate oxide layer during the high-temperature drive-in process, which might bring more defects.Also, we add a proper amount of impurities, such as chlorine gas, during the oxidation process to change the stress distribution of the local valence bond at the interface, reducing the local stress of the gate dielectric layer.We also use Ar instead of N 2 as the annealing gas during the oxide growth annealing process to reduce the concentration of Si-H and Si-OH bonds.
the same chip size.
For the radiation-hardened MOSFET devices, there are several special processes to enhance the anti-TID ability, as we reported before [19].
(1) Gate oxide growth atmosphere control.The charged defects of the gate oxide layer of MOSFETs include two parts: charges introduced by manufacturing processes and charges introduced by TID irradiation, as shown in Figure 2. To obtain a better quality dielectric layer, we arrange the gate oxide growth process after the high-temperature drive-in process to prevent the possible impurities in the silicon from diffusing into the gate oxide layer during the high-temperature drive-in process, which might bring more defects.Also, we add a proper amount of impurities, such as chlorine gas, during the oxidation process to change the stress distribution of the local valence bond at the interface, reducing the local stress of the gate dielectric layer.We also use Ar instead of N2 as the annealing gas during the oxide growth annealing process to reduce the concentration of Si-H and Si-OH bonds.(2) Optimization of body region implantation.The TID-induced positive charges near the Si/SiO2 interface can easily cause gate leakage current.Therefore, the body region implantation conditions are optimized to reduce this leakage.On one hand, traditional onetime field implantation is changed to multiple implantations, thus, the ion concentration of the body region edge is improved, depressing the influence of the interface positive charges caused by TID radiation.On the other hand, the implantation dose is optimized.
(3) Thin gate oxide thickness.The TID effect is strongly correlated with the gate oxide layer thickness.Generally, the thinner gate oxide thickness obtains smaller charges.Therefore, the gate oxide thickness, which is 100 Å~200 Å thinner than the conventional process MOSFET, is designed.
For the fabricated D-mode GaN HEMT devices, a metal-insulator-semiconductor (MIS) structure is designed to depress the gate current leak and, thus, improve the device's reliability [20].
Static parameters of the breakdown voltage, threshold voltage, and on-state resistance are tested for both types of devices.The breakdown voltage (BV) of Type A devices is 710 V, the threshold voltage (VTH) is +3.0 V, and the on-state resistance (RON) is 0.13 Ω.The BV of Type B devices is 710 V, the VTH is +3.6 V, and the RON is 0.13 Ω.The microscopic views of the fabricated devices are shown in Figure 3, where Figure 3a is the trench gate MOSFET and Figure 3b is the GaN MISHEMT.(2) Optimization of body region implantation.The TID-induced positive charges near the Si/SiO 2 interface can easily cause gate leakage current.Therefore, the body region implantation conditions are optimized to reduce this leakage.On one hand, traditional onetime field implantation is changed to multiple implantations, thus, the ion concentration of the body region edge is improved, depressing the influence of the interface positive charges caused by TID radiation.On the other hand, the implantation dose is optimized.
(3) Thin gate oxide thickness.The TID effect is strongly correlated with the gate oxide layer thickness.Generally, the thinner gate oxide thickness obtains smaller charges.Therefore, the gate oxide thickness, which is 100 Å~200 Å thinner than the conventional process MOSFET, is designed.
For the fabricated D-mode GaN HEMT devices, a metal-insulator-semiconductor (MIS) structure is designed to depress the gate current leak and, thus, improve the device's reliability [20].
Static parameters of the breakdown voltage, threshold voltage, and on-state resistance are tested for both types of devices.The breakdown voltage (BV) of Type A devices is 710 V, the threshold voltage (V TH ) is +3.0 V, and the on-state resistance (R ON ) is 0.13 Ω.The BV of Type B devices is 710 V, the V TH is +3.6 V, and the R ON is 0.13 Ω.The microscopic views of the fabricated devices are shown in Figure 3, where Figure 3a is the trench gate MOSFET and Figure 3b is the GaN MISHEMT.

TID Experiments
We prepared 24 devices, of which 12 were Type A and 12 were Type B. Each six devices was connected parallelly as a group, the drain and source electrodes were shorted and jointly led out, and the gate electrodes were led out separately, as shown in Figure 4a.The irradiation configurations of the individual device are shown in Figure 4b.We took the TID experiments at the Institute of Nuclear Physics and Chemistry, Chinese Academy of Engineering Physics.Both types of devices underwent TID experiments.The devices were irradiated at a 50 rad(Si)/s dose rate and a total dose of 500 krad(Si) under Co-60 conditions.After every 50 krad(Si) dose, the devices were tested once.In this research, "bias mode" is used during the irradiation process, and the gate bias voltage applied to the devices was set to be +12 V, which is the common work bias condition.We used a BR3500 semiconductor discrete device tester to test the device's parameters before and after the TID and annealing processes, as shown in Figure 4c.BV is tested at ID = 1.0 mA, VGS = 0 V, Vth is tested at ID = 1.0 mA, VDS = VGS, and RON is tested at VGS = 10 V, ID = 5 A. All the parameters are tested at TA = 25 °C.At higher temperatures, we used NTH64-70A to provide a stable high-temperature annealing environment.The presented data are calculated by the average test results.

TID Experiments
We prepared 24 devices, of which 12 were Type A and 12 were Type B. Each six devices was connected parallelly as a group, the drain and source electrodes were shorted and jointly led out, and the gate electrodes were led out separately, as shown in Figure 4a.The irradiation configurations of the individual device are shown in Figure 4b.We took the TID experiments at the Institute of Nuclear Physics and Chemistry, Chinese Academy of Engineering Physics.Both types of devices underwent TID experiments.The devices were irradiated at a 50 rad(Si)/s dose rate and a total dose of 500 krad(Si) under Co-60 conditions.After every 50 krad(Si) dose, the devices were tested once.In this research, "bias mode" is used during the irradiation process, and the gate bias voltage applied to the devices was set to be +12 V, which is the common work bias condition.We used a BR3500 semiconductor discrete device tester to test the device's parameters before and after the TID and annealing processes, as shown in Figure 4c.BV is tested at

Type A Devices
Since no TID-hardened measures were taken, the threshold voltage of the Type A devices shifted to negative values at 100 krad(Si).Although other static characteristics of the devices, such as BV and RON, were still normal (measured under the condition of VGS < 0 V), the devices were regarded as failed because they were no longer controllable and would seriously threaten the safety of the system.The pre-and post-TID results of the Type A devices are shown in Table 1.The tested results show that the VTH of the device shifted from +3.0 V to +1.2 V at 50 krad(Si), and the shift amount was −1.8 V.At 100 krad(Si), the VTH shifted to below 0 V, and the shift amount exceeded −3 V.This trend is consistent with the VTH shift trend of a single MOSFET device.For the BV of the device, there was almost no change, indicating that the total dose effect had a very small impact on the depletion-mode GaN MISHEMT device within 100 krad(Si).Since the device had failed at 100 krad(Si), the experiment stopped and no higher doses were added.
Table 1.Pre-and post-TID results of Type A (average from the tested results).

Type B Devices
For the devices with radiation-hardened measures, the VTH of the Type B devices also shifted negatively (TID dose, which shifted to +3.1 V at 100 krad(Si), and the shift amount was −0.5 V).By adding higher doses, it was found that the VTH of the device shifted to negative values at 500 krad(Si).The experimental results are presented in Table 2, where for BV and RON, the calculated average results are shown, while for VTH, all the tested individual results are shown, and the standard deviations under each dose amount and the linear fitted lines in Figure 5 are calculated.The standard deviation of each group of devices ranges from 0.072 to 0.214, which shows a relatively good consistency for our experiment.The threshold voltage shift trend of the device was similar to that of the radiationhardened MOSFET device as we reported before [19], indicating that the MOSFET device is the key factor limiting the total dose capability of the Cascode devices.The BV of the device slightly decreases and the RON slightly increases at higher doses, due to the high

Type A Devices
Since no TID-hardened measures were taken, the threshold voltage of the Type A devices shifted to negative values at 100 krad(Si).Although other static characteristics of the devices, such as BV and R ON, were still normal (measured under the condition of V GS < 0 V), the devices were regarded as failed because they were no longer controllable and would seriously threaten the safety of the system.The pre-and post-TID results of the Type A devices are shown in Table 1.The tested results show that the V TH of the device shifted from +3.0 V to +1.2 V at 50 krad(Si), and the shift amount was −1.8 V.At 100 krad(Si), the V TH shifted to below 0 V, and the shift amount exceeded −3 V.This trend is consistent with the V TH shift trend of a single MOSFET device.For the BV of the device, there was almost no change, indicating that the total dose effect had a very small impact on the depletion-mode GaN MISHEMT device within 100 krad(Si).Since the device had failed at 100 krad(Si), the experiment stopped and no higher doses were added.
Table 1.Pre-and post-TID results of Type A (average from the tested results).

Type B Devices
For the devices with radiation-hardened measures, the V TH of the Type B devices also shifted negatively (TID dose, which shifted to +3.1 V at 100 krad(Si), and the shift amount was −0.5 V).By adding higher doses, it was found that the V TH of the device shifted to negative values at 500 krad(Si).The experimental results are presented in Table 2, where for BV and R ON , the calculated average results are shown, while for V TH , all the tested individual results are shown, and the standard deviations under each dose amount and the linear fitted lines in Figure 5 are calculated.The standard deviation of each group of devices ranges from 0.072 to 0.214, which shows a relatively good consistency for our experiment.The threshold voltage shift trend of the device was similar to that of the radiation-hardened MOSFET device as we reported before [19], indicating that the MOSFET device is the key factor limiting the total dose capability of the Cascode devices.The BV of the device slightly decreases and the R ON slightly increases at higher doses, due to the high dose rate which leads to the decrease of the two-dimensional electron gas concentration in the device, which is consistent with the research results of X.Sun [21] and Y. S. Puzyrev [22].The threshold voltages of Type A and Type B devices at different doses are shown in Figure 5. Note: "std" stands for "standard deviation", and "ave" stands for "average".
Micromachines 2023, 14, x FOR PEER REVIEW 6 of 9 dose rate which leads to the decrease of the two-dimensional electron gas concentration in the device, which is consistent with the research results of X.Sun [21] and Y. S. Puzyrev [22].The threshold voltages of Type A and Type B devices at different doses are shown in Figure 5.
Table 2. Pre-and post-TID results of Type B. (BV and RON are the averages from the tested results).
Note: "std" stands for "standard deviation", and "ave" stands for "average".From the tested results above, it is observed that, compared with D-mode GaN MISHEMT devices, Si MOSFET devices are easier to fail in TID experiments, regardless of whether TID-radiation hardening measures are taken or not.This phenomenon indicates that the anti-TID ability of the GaN MISHEMT device is stronger than that of the MOSFET device, even when the MOSFET device has been radiation hardened.Overall, since the total ionizing dose level of 100 krad(Si) is sufficient for utilization in aerospace, Cascode devices with radiation-hardened measures can be a potential candidate in general aerospace systems.
Previously, we reported the TID and annealing effects of p-GaN HEMT devices [3].The largest threshold voltage shift amount is −0.2 V at 500 krad(Si), which is significantly smaller than the Cascode devices with/without radiation-hardened MOSFET in this research.This result indicates that p-GaN devices show better anti-TID ability than Cascode devices.The unstable V TH of the power device may cause difficulties in the design of the front-stage driver or controller.

Annealing Process
After the TID experiments, the annealing processes were taken to observe the recovery of V TH .The received TID-radiation doses of the Type A and Type B devices are 100 krad(Si) and 500 krad(Si), respectively.A 168 h high-temperature annealing (HTA, T A = 100 • C) and a 168 h room temperature annealing (RTA, T A = 25 • C) were separately taken, and the recovery of V TH was observed.During the annealing process, the V TH of the device was tested every 12 h.The experiment results are shown in Figure 6.

Conclusions
In this work, we designed Cascode devices combined with low-voltage MOSFET and high-voltage D-mode GaN MISHEMT.Both the conventional processed MOSFET device and the radiation-hardened MOSFET device were fabricated to analyze the TID effects.Experiment results show that the threshold voltage of the Cascode device shifts negatively, of which the shift trench is similar to a single MOSFET device, indicating that the MOSFET device is the key limiting the anti-radiation ability of the Cascode device, and annealing processes after the TID process help the threshold voltage recover.Compared with the p-GaN HEMT device, the Cascode device shows a more sensitive response to TID.While the advantages of high-gate breakdown voltage and high threshold voltage make Cascode a better device for directly substituting traditional high-voltage MOSFET.We believe that p-GaN HEMT devices and Cascode devices will share great properties in separate fields of power applications.
Author Contributions: Conceptualization, H.W.; methodology, X.F. and J.L.; formal analysis, M.Y.As annealing has been well considered a popular method for repairing TID-induced SiO 2 damages [23], tested results show that for both types of devices, the V TH gradually recovered from negative to positive, indicating that the TID effect did not cause complete damage to the device, and the V TH is partially recoverable.The annealing process releases the trapped charges, leading to the recombination of electrons and holes, and, thus, reduces the amount of trap charges both at the interface of Si/SiO 2 and inside SiO 2 of the MOSFET device.Therefore, the V TH is recovered, where the V TH of Type A under RTA returns to an average value of 1.32 V, the V TH of Type A under HTA returns to 2.65 V, the V TH of Type B under RTA returns to 2.32 V, and the V TH of Type B under HTA returns to 3.45 V.It is observed that both Type A and Type B devices show faster recovery trends of V TH under HTA, as high temperature can accelerate the release process of trapped charges.However,

Figure 1 .
Figure 1.The schematic diagram (a) and the package (b) of the Cascode device.

Figure 1 .
Figure 1.The schematic diagram (a) and the package (b) of the Cascode device.

Figure 3 .
Figure 3. Microscopic views of the fabricated devices, where (a) is the trench gate MOSFET and (b) is the GaN MISHEMT.

Figure 3 .
Figure 3. Microscopic views of the fabricated devices, where (a) is the trench gate MOSFET and (b) is the GaN MISHEMT.
and R ON is tested at V GS = 10 V, I D = 5 A. All the parameters are tested at T A = 25 • C. At higher temperatures, we used NTH64-70A to provide a stable high-temperature annealing environment.The presented data are calculated by the average test results.

Figure 4 .
Figure 4. TID experiment board.Each six devices is connected parallelly as a group (a).Irradiation configurations of individual devices (b).BR3500 semiconductor discrete device tester to test the device's parameters (c).

Figure 4 .
Figure 4. TID experiment board.Each six devices is connected parallelly as a group (a).Irradiation configurations of individual devices (b).BR3500 semiconductor discrete device tester to test the device's parameters (c).

Figure 5 .
Figure 5.The threshold voltages of Type A and Type B devices at different total doses.The small dots are the tested values, while the large dots are the calculated average values.The red line is the linear fitting for Type B, and the blue line is the linear fitting for Type A. The calculated slope is

Figure 5 .
Figure5.The threshold voltages of Type A and Type B devices at different total doses.The small dots are the tested values, while the large dots are the calculated average values.The red line is the linear fitting for Type B, and the blue line is the linear fitting for Type A. The calculated slope is −0.0076 mV/krad(Si).The goodness of fits (R 2 ) is 0.9946, which is higher than 0.99, showing good linear fitting.

9 Figure 6 .
Figure 6.Recovery of VTH of both types.The red solid dots are the tested values of HTA for Type B. The red hollow dots are the tested values of RTA for Type B. The blue solid dots are the tested values of HTA for Type A. The blue hollow dots are the tested values of RTA for Type A. The solid and dashed lines are the calculated average values.The formulas on the right side are the fitted recovery trends.

Figure 6 .
Figure 6.Recovery of V TH of both types.The red solid dots are the tested values of HTA for Type B. The red hollow dots are the tested values of RTA for Type B. The blue solid dots are the tested values of HTA for Type A. The blue hollow dots are the tested values of RTA for Type A. The solid and dashed lines are the calculated average values.The formulas on the right side are the fitted recovery trends.

Table 2 .
Pre-and post-TID results of Type B (BV and R ON are the averages from the tested results).