Simulation Study on the Structure Design of p-GaN/AlGaN/GaN HEMT-Based Ultraviolet Phototransistors

This work investigates the impacts of structural parameters on the performances of p-GaN/AlGaN/GaN HEMT-based ultraviolet (UV) phototransistors (PTs) using Silvaco Atlas. The simulation results show that a larger Al content or greater thickness for the AlGaN barrier layer can induce a higher two-dimensional electron gas (2DEG) density and produce a larger photocurrent. However, they may also lead to a larger dark current due to the incomplete depletion of the GaN channel layer. The depletion conditions with various Al contents and thicknesses of the AlGaN layer are investigated in detail, and a borderline between full depletion and incomplete depletion was drawn. An optimized structure with an Al content of 0.23 and a thickness of 14 nm is achieved for UV-PT, which exhibits a high photocurrent density of 92.11 mA/mm, a low dark current density of 7.68 × 10−10 mA/mm, and a large photo-to-dark-current ratio of over 1011 at a drain voltage of 5 V. In addition, the effects of other structural parameters, such as the thickness and hole concentration of the p-GaN layer as well as the thickness of the GaN channel layer, on the performances of the UV-PTs are also studied in this work.


Introduction
Ultraviolet (UV) photodetectors based on III-nitride semiconductors have numerous applications in military and civilian fields [1]. Various types of GaN photodetectors, such as the Schottky barrier [2,3], metal-semiconductor-metal [4,5], and p-i-n [6,7], have been widely explored. However, these usually do not have optical gains, which limits the scope of their applications.
In addition to the conventional photodetector configurations, phototransistors (PTs) based on AlGaN/GaN HEMT structures [8] have also been presented. Three methods of gate control have been explored, including recessed barrier gate [9][10][11], Schottky gate [12][13][14], and p-GaN gate [15][16][17]. Among these approaches, PT with a p-GaN optical gate has shown excellent comprehensive performance. PT with a p-GaN optical gate was first reported by Iwaya et al. in 2009 [15]. The device showed a large optical gain and a relatively low dark current. In 2020, Lyu et al. analyzed the working mode of PTs and reported a device with a much lower leakage current as well as a large photo-to-dark current ratio [16]. Recently, our group demonstrated a device with a fast response time to the microsecond level while maintaining excellent electrical performances [17]. The above progress suggests that p-GaN/AGaN/GaN HEMT-based UV PT exhibits both high responsivity and fast response time and shows great potential for high-performance UV detection.
In this work, the impacts of structural parameters on the performances of p-GaN/AlGaN/GaN HEMT-based UV PTs are systematically investigated using Silvaco software. The electron distribution, polarization charge density, polarization-induced twodimensional electron gas (2DEG) density, and conduction band diagrams are calculated to reveal the influence of hole concentration, layer thickness, and alloy composition on dark current and photocurrent. The variations of depletion conditions with the Al content and thickness of the AlGaN barrier are investigated in detail, and a borderline for full depletion and incomplete depletion was drawn as a reference of device design for subsequent researchers. Figure 1 plots the schematic structure of a typical PT, which consists of a 100-nm-thick p-GaN layer (p~1 × 10 18 cm −3 ), a 15-nm-thick Al 0.2 Ga 0.8 N barrier layer (n~1 × 10 15 cm −3 ), and a 300-nm-thick GaN channel layer (n~1 × 10 15 cm −3 ) from top to bottom. The gate length (L g ) and gate width (W g ) of the device are 4 µm and 100 µm, respectively. The gate-source distance (L gs ) and gate-drain distance (L gd ) are 2 µm. The operating principle of the device was illustrated in our previous work [17].

Device Structure and Simulation Models
Micromachines 2022, 13, 2210 2 of 9 In this work, the impacts of structural parameters on the performances of p-GaN/Al-GaN/GaN HEMT-based UV PTs are systematically investigated using Silvaco software. The electron distribution, polarization charge density, polarization-induced two-dimensional electron gas (2DEG) density, and conduction band diagrams are calculated to reveal the influence of hole concentration, layer thickness, and alloy composition on dark current and photocurrent. The variations of depletion conditions with the Al content and thickness of the AlGaN barrier are investigated in detail, and a borderline for full depletion and incomplete depletion was drawn as a reference of device design for subsequent researchers. Figure 1 plots the schematic structure of a typical PT, which consists of a 100-nmthick p-GaN layer (p~1 × 10 18 cm −3 ), a 15-nm-thick Al0.2Ga0.8N barrier layer (n~1 × 10 15 cm −3 ), and a 300-nm-thick GaN channel layer (n~1 × 10 15 cm −3 ) from top to bottom. The gate length (Lg) and gate width (Wg) of the device are 4 µm and 100 µm, respectively. The gatesource distance (Lgs) and gate-drain distance (Lgd) are 2 µm. The operating principle of the device was illustrated in our previous work [17]. Steady-state 2-D numerical simulations based on Silvaco TCAD Atlas software are performed. The definition of fundamental equations and physical models can be found in our previous work [18]. The spaces between the gate source and gate drain are filled with SiO2. The source and drain electrodes are defined as ohmic contacts. The wavelength and intensity of the incident light are set to 360 nm and 1 mW/cm 2 , respectively.

Hole Concentration of the p-GaN Layer
The simulation results show that a higher hole concentration in the p-GaN layer leads to a deeper depletion region in the GaN channel layer, as shown in Figure 2. For a 300 nm GaN layer that can completely absorb the incident light, a hole concentration of 1 × 10 18 cm −3 in the p-GaN layer is essential to fully deplete the GaN layer and suppress the leakage current. Because it is difficult to further increase the hole concentration of the p-GaN layer in MOCVD growth [19], the value is determined to be 1 × 10 18 cm −3 in later simulations. Steady-state 2-D numerical simulations based on Silvaco TCAD Atlas software are performed. The definition of fundamental equations and physical models can be found in our previous work [18]. The spaces between the gate source and gate drain are filled with SiO 2 . The source and drain electrodes are defined as ohmic contacts. The wavelength and intensity of the incident light are set to 360 nm and 1 mW/cm 2 , respectively.

Hole Concentration of the p-GaN Layer
The simulation results show that a higher hole concentration in the p-GaN layer leads to a deeper depletion region in the GaN channel layer, as shown in Figure 2. For a 300 nm GaN layer that can completely absorb the incident light, a hole concentration of 1 × 10 18 cm −3 in the p-GaN layer is essential to fully deplete the GaN layer and suppress the leakage current. Because it is difficult to further increase the hole concentration of the p-GaN layer in MOCVD growth [19], the value is determined to be 1 × 10 18 cm −3 in later simulations.

Thickness of the p-GaN Layer
With the increase in the p-GaN layer thickness from 50 to 200 nm, the photocurrent density between the source and drain decreases monotonically from 72.61 to 69.84 mA/mm, as presented in Figure 3. The reduction of photocurrent is due to the absorption loss of the incident light in the p-GaN layer. As a result, a thinner p-GaN is preferred to maintain a high photocurrent. However, a thickness of approximately 50 nm is essential to ensure the material quality and doping stability of p-GaN during MOCVD growth [20]. As a consequence, a trade-off should be made, and the thickness of the p-GaN layer is selected to be 50 nm in later simulations.

Thickness of the p-GaN Layer
With the increase in the p-GaN layer thickness from 50 to 200 nm, the pho density between the source and drain decreases monotonically from 72.61 mA/mm, as presented in Figure 3. The reduction of photocurrent is due to the ab loss of the incident light in the p-GaN layer. As a result, a thinner p-GaN is pre maintain a high photocurrent. However, a thickness of approximately 50 nm is to ensure the material quality and doping stability of p-GaN during MOCVD gro As a consequence, a trade-off should be made, and the thickness of the p-GaN selected to be 50 nm in later simulations.

Thickness of the p-GaN Layer
With the increase in the p-GaN layer thickness from 50 to 200 nm, the density between the source and drain decreases monotonically from mA/mm, as presented in Figure 3. The reduction of photocurrent is due to loss of the incident light in the p-GaN layer. As a result, a thinner p-GaN maintain a high photocurrent. However, a thickness of approximately 50 n to ensure the material quality and doping stability of p-GaN during MOCV As a consequence, a trade-off should be made, and the thickness of the p selected to be 50 nm in later simulations.

Thickness of the GaN Channel Layer
With the increase in the GaN channel layer thickness from 100 to 400 n current density between the source and drain rises slightly from 71.53 to 7 as presented in Figure 4. The increase of photocurrent is due to the extend depth of the incident light in the GaN channel layer. As a result, a thicker G preferred to maintain a high photocurrent.

Thickness of the GaN Channel Layer
With the increase in the GaN channel layer thickness from 100 to 400 nm, the photocurrent density between the source and drain rises slightly from 71.53 to 73.07 mA/mm, as presented in Figure 4. The increase of photocurrent is due to the extended absorption depth of the incident light in the GaN channel layer. As a result, a thicker GaN channel is preferred to maintain a high photocurrent.  However, as the thickness exceeds 300 nm, the GaN layer cannot be f as displayed in Figure 5d. This will result in increased leakage current and device performance, as discussed before. As a consequence, a trade-off s made, and the thickness of the GaN channel layer is selected to be 300 nm in tions.  However, as the thickness exceeds 300 nm, the GaN layer cannot be fully depleted, as displayed in Figure 5d. This will result in increased leakage current and degenerated device performance, as discussed before. As a consequence, a trade-off should also be made, and the thickness of the GaN channel layer is selected to be 300 nm in later simulations.  However, as the thickness exceeds 300 nm, the GaN layer cannot be fully as displayed in Figure 5d. This will result in increased leakage current and de device performance, as discussed before. As a consequence, a trade-off shou made, and the thickness of the GaN channel layer is selected to be 300 nm in lat tions.  Figure 6a demonstrates the polarization charge density at the AlGaN/Ga junction interface, which is induced by spontaneous and piezoelectric polari dark conditions, the polarization-induced 2DEG is depleted by the p-GaN lay sented in Figure 6b. Additionally, the conduction band of the GaN channel is fla illustrated in Figure 6c. Under UV illumination, no obvious difference was obs the polarization charge density, while the 2DEG was restored, as shown in Figu   Figure 6b. Additionally, the conduction band of the GaN channel is flattened, as illustrated in Figure 6c. Under UV illumination, no obvious difference was observed for the polarization charge density, while the 2DEG was restored, as shown in Figures 6a and 6b, respectively. The conduction band sinks at the heterojunction interface, as illustrated in Figure 6d. The Al content and thickness of the AlGaN barrier layer are two critical paramete adjusting the 2DEG density, which determines the magnitude of the photocurrent. Wi an increase in the Al content from 0.15 to 0.25, both the spontaneous polarization an piezoelectric polarization are continuously enhanced. As a result, the increased Al conte of the AlGaN barrier layer leads to a larger polarization charge density, a higher 2DE density, and a bigger conduction band offset, as presented in Figure 6. With the increase in the AlGaN layer thickness from 10 to 20 nm, the polarizati charge density remains the same, as presented in Figure 7a. However, a higher 2DE density is observed, both in the dark and under illumination, as presented in Figure 7 The increased thickness of the AlGaN layer also leads to a bigger conduction band offs as presented in Figure 7c,d. As a result, a larger photocurrent can be achieved with thicker AlGaN layer. The Al content and thickness of the AlGaN barrier layer are two critical parameters adjusting the 2DEG density, which determines the magnitude of the photocurrent. With an increase in the Al content from 0.15 to 0.25, both the spontaneous polarization and piezoelectric polarization are continuously enhanced. As a result, the increased Al content of the AlGaN barrier layer leads to a larger polarization charge density, a higher 2DEG density, and a bigger conduction band offset, as presented in Figure 6.

Al Content and Thickness of the AlGaN Barrier Layer
With the increase in the AlGaN layer thickness from 10 to 20 nm, the polarization charge density remains the same, as presented in Figure 7a. However, a higher 2DEG density is observed, both in the dark and under illumination, as presented in Figure 7b. The increased thickness of the AlGaN layer also leads to a bigger conduction band offset, as presented in Figure 7c,d. As a result, a larger photocurrent can be achieved with a thicker AlGaN layer.
It is worth noting that despite the improved photocurrent, a larger Al content or greater thickness for the AlGaN layer also leads to an increased leakage current. With a larger Al content of 0.25, the 300 nm GaN channel layer could not be completely depleted, and the leakage current density rose significantly, as displayed in Figures 8a and 8b  It is worth noting that despite the improved photocurrent, a larger Al content or greater thickness for the AlGaN layer also leads to an increased leakage current. With a larger Al content of 0.25, the 300 nm GaN channel layer could not be completely depleted, and the leakage current density rose significantly, as displayed in Figure 8a and    It is worth noting that despite the improved photocurrent, a larger Al content or greater thickness for the AlGaN layer also leads to an increased leakage current. With a larger Al content of 0.25, the 300 nm GaN channel layer could not be completely depleted, and the leakage current density rose significantly, as displayed in Figure 8a and    maintaining a low dark current density of 7.68 × 10 mA/mm and a large photo-to-dar current ratio of over 10 11 at a drain voltage of 5 V, as presented in Figure 9b. As a cons quence, the optimized structure is determined, as shown in the insert of Figure 9b. Com pared with our previous work [17], the optimized structure demonstrates a similar ph tocurrent density and photo-to-dark-current ratio with a light intensity that is one magn tude lower.

Conclusions
In summary, a comprehensive simulation via Silvaco Atlas was conducted to reve the impact of the structural parameters on the performances of p-GaN/AlGaN/Ga HEMT-based UV PTs. The hole concentration and thickness of the p-GaN layer as well the thickness of the GaN channel layer are studied and optimized. The depletion cond tions with various Al contents and thicknesses of the AlGaN barrier layer are investigat in detail, and a borderline between full depletion and incomplete depletion was draw Finally, an optimized structure with an Al content of 0.23 and a thickness of 14 nm achieved for UV-PT, which exhibits the highest photocurrent density of 92.11 mA/mm low dark current density of 7.68 × 10 −10 mA/mm, and a large photo-to-dark-current rat of over 10 11 at a drain voltage of 5 V. We believe that the results have drawn a clear phy ical map of HEMT-based PTs and could be a useful guide for device design for subseque researchers.   Under this premise, more detailed simulations are conducted with Al content in 0.01 steps and thickness in 1 nm steps. The results reveal that the device with an Al content of 0.23 and a thickness of 14 nm has the highest photocurrent density of 92.11 mA/mm while maintaining a low dark current density of 7.68 × 10 −10 mA/mm and a large phototo-dark-current ratio of over 10 11 at a drain voltage of 5 V, as presented in Figure 9b. As a consequence, the optimized structure is determined, as shown in the insert of Figure 9b. Compared with our previous work [17], the optimized structure demonstrates a similar photocurrent density and photo-to-dark-current ratio with a light intensity that is one magnitude lower.

Conclusions
In summary, a comprehensive simulation via Silvaco Atlas was conducted to reveal the impact of the structural parameters on the performances of p-GaN/AlGaN/GaN HEMTbased UV PTs. The hole concentration and thickness of the p-GaN layer as well as the thickness of the GaN channel layer are studied and optimized. The depletion conditions with various Al contents and thicknesses of the AlGaN barrier layer are investigated in detail, and a borderline between full depletion and incomplete depletion was drawn. Finally, an optimized structure with an Al content of 0.23 and a thickness of 14 nm is achieved for UV-PT, which exhibits the highest photocurrent density of 92.11 mA/mm, a low dark current density of 7.68 × 10 −10 mA/mm, and a large photo-to-dark-current ratio of over 10 11 at a drain voltage of 5 V. We believe that the results have drawn a clear physical map of HEMT-based PTs and could be a useful guide for device design for subsequent researchers.

Conflicts of Interest:
The authors declare no conflict of interest.