Weak-Light-Enhanced AlGaN/GaN UV Phototransistors with a Buried p-GaN Structure

Abstract

We propose a novel ultraviolet (UV) phototransistor (PT) architecture based on an AlGaN/GaN high electron mobility transistor (HEMT) with a buried p-GaN layer. In the dark, the polarization-induced two-dimensional electron gas (2DEG) at the AlGaN/GaN heterojunction interface is depleted by the buried p-GaN and the conduction channel is closed. Under UV illumination, the depletion region shrinks to just beneath the AlGaN/GaN interface and the 2DEG recovers. The retraction distance of the depletion region during device turn-on operation is comparable to the thickness of the AlGaN barrier layer, which is an order of magnitude smaller than that in the conventional p-GaN/AlGaN/GaN PT, whose retraction distance spans the entire GaN channel layer. Consequently, the proposed device demonstrates significantly enhanced weak-light detection capability and improved switching speed. Silvaco Atlas simulations reveal that under a weak UV intensity of 100 nW/cm², the proposed device achieves a photocurrent density of 1.68 × 10⁻³ mA/mm, responsivity of 8.41 × 10⁵ A/W, photo-to-dark-current ratio of 2.0 × 10⁸, UV-to-visible rejection ratio exceeding 10⁸, detectivity above 1 × 10¹⁹ cm·Hz^1/2/W, and response time of 0.41/0.41 ns. The electron concentration distributions, conduction band variations, and 2DEG recovery behaviors in both the conventional and novel structures under dark and weak UV illumination are investigated in depth via simulations.

1. Introduction

III-nitride semiconductors have enabled the development of ultraviolet (UV) photodetectors (PDs) for diverse applications, including medical imaging, environmental monitoring, and space exploration [1,2,3,4,5]. Gallium nitride (GaN) is considered one of the most promising materials for the development of visible-blind UV PDs, owing to its wide direct bandgap of 3.4 eV at room temperature in the wurtzite phase. Over the past two decades, extensive research has focused on the exploration of various GaN-based PD structures, such as Schottky barrier [6], metal–semiconductor–metal (MSM) [7,8], and p-i-n [9] devices. However, these PDs generally lack optical gain, necessitating external amplification circuits, which restricts their application scope.

In recent years, phototransistors (PTs) employing AlGaN/GaN high-electron-mobility transistor (HEMT) structures [10] have attracted significant attention beyond traditional PD configurations. Three types of gate control schemes have been explored, namely recessed barrier gates [11,12,13], Schottky gates [14,15,16,17,18], and p-GaN gates [19,20,21,22,23,24,25]. Among the explored strategies, PTs utilizing p-GaN optical gates have demonstrated remarkable potential for visible-blind UV detection. The earliest report of PTs incorporating p-GaN optical gates can be traced back to 2009, when Iwaya et al. introduced a device exhibiting high optical gain and relatively low dark current [19]. Subsequently, in 2020, Lyu et al. explored the operational mechanisms of such PTs and presented a design with significantly reduced leakage current and an enhanced photo-to-dark current ratio (PDCR) [20]. Building on these developments, our group, in 2022, demonstrated a PT capable of microsecond-level response speeds while retaining excellent electrical characteristics [21]. These advancements collectively establish the p-GaN/AlGaN/GaN HEMT-based UV PT as an emerging device architecture, distinguished by its low dark current, high internal gain, strong PDCR, and outstanding UV-to-visible rejection ratio (UVRR).

For PTs based on the conventional p-GaN/AlGaN/GaN HEMT structure, the optical switching mechanism follows the process described in our previous work [25]. However, these devices suffer from two fundamental limitations that hinder their performance, particularly under weak UV illumination. First, the top p-GaN layer absorbs part of the incident light, leading to optical loss and reduced external quantum efficiency. Second, the depletion region formed by the p-GaN gate typically spans the entire GaN channel (~300 nm). During illumination, restoring the two-dimensional electron gas (2DEG) requires the depletion region to retract over this full depth, which demands a strong photovoltage and thus a high threshold light intensity. This severely limits weak-light detection capability. Moreover, the long depletion retraction path introduces considerable carrier delay and slows down the modulation of channel conductivity, thereby degrading the temporal response of the device. These challenges—poor weak-light sensitivity and slow response speed due to deep depletion modulation—are precisely what the present work aims to overcome.

This work introduces a new UV phototransistor (PT) design utilizing an AlGaN/GaN HEMT structure with a buried p-GaN layer to enhance weak-light detection performance. In the dark, the buried p-GaN gate depletes the polarization-induced two-dimensional electron gas (2DEG) at the AlGaN/GaN interface, closing the conduction channel. When UV light is applied, the depletion region contracts to the AlGaN/GaN interface and enables the recovery of the 2DEG. When the device is activated, the depletion region contracts to a distance roughly equivalent to the AlGaN barrier thickness, which is significantly smaller—by an order of magnitude—compared to the retraction distance in traditional p-GaN/AlGaN/GaN PTs, where it matches the GaN channel thickness. As a result, the proposed device removes the incident light absorption loss from the top p-GaN layer and, crucially, allows the 2DEG to recover even under very low light intensity. The weak-light detection performance and switching speed of both the conventional and proposed structures are systematically analyzed using Silvaco TCAD Atlas. Simulations provide a detailed examination of the electron concentration distributions, conduction band profiles, and 2DEG recovery behaviors in both device types under dark and weak UV illumination conditions.

2. Device Structures and Operation Principles

The device structures and operating principles of the PTs are schematically illustrated in Figure 1. The gate length (L_g) and gate width are set to be 2 μm and 1 m, respectively. The distance of the source to gate (L_gs) and gate to drain (L_gd) are both defined as 2 μm. In the device layout, the regions between the gate and the source/drain are insulated with SiO₂, while the source and drain electrodes are implemented as ohmic contacts.

The conventional PT consists of a capped p-GaN gate layer, an AlGaN barrier layer, and a GaN channel layer from top to bottom. In the dark, the built-in electric field of the p-GaN/i-GaN junction drives the depletion region to extend across the entire GaN channel and reach the buffer layer. Under UV illumination, photogenerated electron–hole pairs within the depletion region are separated and drift in opposite directions, producing a photovoltaic effect that reduces the built-in electric field strength. Therefore, a relatively strong light intensity is required to shrink the depletion region from the bottom of the GaN channel layer to above the barrier layer, as shown in Figure 1a.

The novel PT consists of a buried p-GaN gate layer, a GaN channel layer, and an AlGaN barrier layer from bottom to top. The AlGaN barrier layer within the gate region is partially etched to reduce the polarization charge density at the AlGaN/GaN heterojunction, which ensures the full depletion of polarization-induced 2DEG in the dark. In the source and drain region, however, a high density of polarization charges as well as 2DEG remain at the AlGaN/GaN interface, which ensures the formation of good Ohmic contacts. Under UV illumination, the depletion region slightly shrinks for several nanometers beneath the AlGaN/GaN interface, and the 2DEG is restored. Therefore, a significantly weaker light intensity is required to turn on the novel device compared with the conventional one, as illustrated in Figure 1b.

3. Simulation Models and Parameter Optimizations

Two-dimensional steady-state simulations were performed using Silvaco TCAD Atlas (version 2022, Silvaco Santa Clara, CA, USA) to comprehensively evaluate the performance advantages of the proposed device structure. The simulation framework is based on the solution of Poisson’s equation, carrier continuity equations, and current transport equations. A range of physical models were employed and grouped into five categories: carrier statistics, mobility models, carrier generation and recombination, and band-to-band tunneling. Additionally, the polarization model was activated to account for spontaneous and piezoelectric polarization effects. Key material parameters and model coefficients were extracted from the published literature [26,27] and our previous work [28,29,30]. The wavelength and intensity of the incident light are fixed at 360 nm and 100 nW/cm², respectively, unless otherwise specified. The electron concentrations of the unintentionally doped layers are determined to be 1 × 10¹⁵ cm⁻³.

According to the operating principle of the proposed structure, the conduction channel should remain nearly fully depleted under dark conditions to ensure a low threshold light intensity and suppress leakage current. Therefore, the hole concentration of the buried p-GaN gate, the GaN channel thickness, and the composition and thickness of the AlGaN barrier layer must be synergistically optimized. To effectively deplete a high-density two-dimensional electron gas (2DEG) at the AlGaN/GaN heterojunction and enhance photocurrent during the turn-on process, the p-GaN gate should have a high hole concentration. Accordingly, the thickness and hole concentration of the buried p-GaN layer are set to 100 nm and 1 × 10¹⁸ cm⁻³ [31], respectively [29]. To facilitate the formation of low-resistance Ohmic contacts in the source and drain regions, the AlGaN barrier layer is designed with a thickness of 20 nm and an Al mole fraction of 0.25, ensuring a high 2DEG density. The residual barrier thickness in the gate region is co-optimized with the GaN channel thickness, as summarized in

Table 1. Structural optimization of the novel PT with varying GaN channel thickness.

Channel Thickness	50 nm	100 nm	150 nm
Remaining Barrier Thickness	11 nm	10 nm	9 nm
Polarization Charge Density	8.5 × 1012 cm−2	7.9 × 1012 cm−2	7.1 × 1012 cm−2
2DEG Density 1	5.2 × 108 cm−2	5.9 × 108 cm−2	4.3 × 108 cm−2
Photocurrent	1.49 × 10−3 mA/mm	1.68 × 10−3 mA/mm	1.21 × 10−3 mA/mm

¹ The 2DEG density is calculated under 360 nm illumination at an intensity of 100 nW/cm².

.

As the GaN channel thickness increases from 50 nm to 100 nm, the remaining AlGaN barrier thickness is reduced from 11 nm to 10 nm to ensure near-complete depletion by the buried p-GaN gate. Consequently, the polarization charge density at the AlGaN/GaN heterojunction decreases from 8.5 × 10¹² cm⁻² to 7.9 × 10¹² cm⁻². Despite this, the restored 2DEG density under 360 nm illumination slightly increases from 5.2 × 10⁸ cm⁻² to 5.9 × 10⁸ cm⁻², accompanied by a rise in photocurrent from 1.49 × 10⁻³ mA/mm to 1.68 × 10⁻³ mA/mm. The improvement in photoresponse is attributed to improved optical absorption in the thicker GaN layer, which facilitates the retraction of the depletion region and promotes 2DEG restoration. Further increasing the GaN channel thickness from 100 nm to 150 nm results in a reduction in the remaining barrier thickness from 10 nm to 9 nm to maintain near-complete depletion. Along with a continued decrease in polarization charge density, the restored 2DEG density declines from 5.9 × 10⁸ cm⁻² to 4.3 × 10⁸ cm⁻², leading to a corresponding drop in photocurrent from 1.68 × 10⁻³ mA/mm to 1.21 × 10⁻³ mA/mm. These results indicate that beyond a certain thickness, the gain in optical absorption becomes marginal—consistent with the Beer–Lambert law [32]—while the reduced polarization charge density impairs 2DEG recovery and degrades the device’s photoresponse. Therefore, an optimal trade-off between incident light absorption and polarization-induced charge density at the AlGaN/GaN interface is essential for performance enhancement. Based on this balance, a GaN channel thickness of 100 nm and a remaining AlGaN barrier thickness of 10 nm are identified as the optimal structural parameters for the proposed device.

The structural parameters of the conventional phototransistor were optimized in our previous work [25]. Specifically, the p-GaN gate layer was designed with a thickness of 100 nm and a hole concentration of 1 × 10¹⁸ cm⁻³. The AlGaN barrier layer was set to 15 nm in thickness with an Al composition of 0.15. The GaN channel layer was designed to be 300 nm thick.

4. Comparison of Conventional and Novel PTs

The electron concentration profiles of the PTs under dark and weak UV illumination conditions at zero bias are presented in Figure 2. For the conventional device, due to the presence of a p-n junction, the conduction channel beneath the capped p-GaN gate is fully depleted in the dark, as shown in Figure 2a. At this point, the dark current of the device is very low. Upon UV illumination, due to the photovoltaic effect at the p-n junction, the depletion region within the GaN channel shrinks from the bottom toward the top, as illustrated in Figure 2b. At this point, the device generates a photocurrent. However, due to the relatively weak light intensity and limited photovoltaic effect, the 2DEG beneath the p-GaN gate, which dominates the magnitude of the photocurrent, is not restored. In contrast, for the novel device, the conduction path above the buried p-GaN gate is also fully depleted in the dark, as shown in Figure 2c. Upon UV illumination, the depletion region contracts to a location slightly below the AlGaN/GaN heterojunction, leading to the restoration of a high-density 2DEG, as illustrated in Figure 2d. Consequently, the novel device is expected to exhibit a significantly higher photocurrent under weak light illumination while ensuring a low dark current, indicating a significantly enhanced weak-light detection capability.

The conduction band diagrams of the PTs under dark and weak UV illumination conditions at zero bias are presented in Figure 3. For the conventional device, the conduction band remains well above the Fermi level in the dark, as illustrated in Figure 3a. Under UV illumination, the conduction band shifts downward but remains above the Fermi level, suggesting that the 2DEG is not restored, as illustrated in Figure 3b. In contrast, for the novel device, the conduction band at the AlGaN/GaN heterojunction is much closer to the Fermi level in the dark, as shown in Figure 3c. Upon UV illumination, it shifts below the Fermi level, signifying a substantial restoration of the 2DEG, as illustrated in Figure 3d. It is worth noting that under the same light intensity, the higher density of recovered 2DEG in the novel device results in an increased photocurrent and enhanced sensitivity. These energy band simulations further validate the enhanced sensitivity of the novel structure under weak UV illumination.

The 2DEG densities at the AlGaN/GaN heterojunction interfaces of the PTs under dark and weak UV illumination conditions at zero bias are illustrated in Figure 4. For the conventional device, the 2DEG located beneath the capped p-GaN gate is completely depleted in the dark, as shown in Figure 4a. Upon UV illumination, a slight restoration of the 2DEG density is observed, however, it remains very low, corresponding to limited photocurrent generation and poor sensitivity. For the novel device, the 2DEG in the active region is also depleted by the buried p-GaN gate in the dark, as shown in Figure 4b. However, in the non-active region, where the AlGaN barrier is thicker, a considerable 2DEG density is preserved, ensuring that the buried p-GaN layer does not interfere with Ohmic contact formation. Upon UV illumination, the 2DEG density in the active region is restored to over 1 × 10⁸ cm⁻², which corresponds to a high photocurrent and enhanced sensitivity, indicating that the conductive channel of the novel device can be effectively reactivated by relatively weak UV illumination. These results confirm the superior weak-light detection capability of the novel structure.

The J–V characteristics of the conventional and novel PTs are presented in Figure 5a and Figure 5b, respectively. For both devices, the dark currents are suppressed to below 1 × 10⁻¹¹ mA/mm at a drain voltage of 5 V, indicating complete depletion of the conduction channels. Compared with the previously reported dark currents [25], which are limited by the detection limit of the measurement system, the simulated dark current is slightly lower. Under UV illumination, the novel device exhibits a high photocurrent exceeding 1 × 10⁻³ mA/mm and a large PDCR greater than 10⁸ at 5 V, which are three orders of magnitude higher than those of the conventional device. These findings are consistent with the simulated conduction band diagrams and 2DEG distributions shown in Figure 3 and Figure 4, further confirming the superior weak-light sensitivity and turn-on capability of the novel structure.

Responsivity (R) is a fundamental and standalone figure of merit that quantifies the strength of the photoresponse, defined as the ratio of the generated photocurrent to the incident optical power:

$$R={\displaystyle \frac{i_{ph}}{P_{\mathrm{o}\mathrm{p}\mathrm{t}}}}$$

(1)

where i_ph is the difference between the photocurrent and dark current, P_opt is the optical power of the incident light. The optical power can be calculated as follows:

$$P_{opt}=I_{opt}\times S$$

(2)

where I_opt represents the intensity of the UV illumination, and S is the total absorption area. Accordingly, the responsivity can be further expressed as follows:

$$R={\displaystyle \frac{i_{ph}}{I_{opt}S}}$$

(3)

At the illumination intensity of 100 nW/cm², the drain currents of the conventional and novel devices are calculated to be 1.26 μA and 1.68 mA, respectively. Consequently, the calculated responsivities are 6.31 × 10² A/W and 8.41 × 10⁵ A/W, respectively.

Figure 6 presents the photocurrent densities and responsivities under 360 nm UV illumination at a drain voltage of 5 V, with incident light intensities ranging from 100 nW/cm² to 100 μW/cm². For the conventional device, the photocurrent densities and responsivities increase sharply at low UV intensities, attributed to the improvement of the restored 2DEG density, as shown in Figure 6a. The simulated trend aligns well with previously reported experimental results [25], which confirm that a strong UV illumination is required for the turn-on operation of the conventional device. For the novel device, however, the photocurrent remains almost constant over the whole light intensity range, as shown in Figure 6b. As a consequence, the responsivity decreases approximately linearly as light intensity increases. These results indicate that the 2DEG in the active region is fully restored under very weak light intensity, further validating the superior weak-light detection capability of the novel structure.

The noise characteristics of the conventional and novel devices were analyzed using Silvaco Atlas simulations to evaluate their weak-light detection capability [33,34]. The noise power density spectra (S_n) over the frequency range of 1–10⁶ Hz in the dark at a drain voltage of 5 V are shown in Figure 7. The novel device exhibits a significantly lower noise level compared to the conventional device, which further confirms its superior capability for detecting weak ultraviolet signals.

Consistent with the experimental results previously reported [25], the simulated S_n spectra for both device structures decrease linearly on a logarithmic scale, as indicated by the red fitting curve in Figure 7. The observed linear dependence of S_n on frequency confirms that the dominant source of the low-frequency noise is trap-assisted 1/f (flicker) noise. The total noise power ⟨i_n⟩² can be calculated by integrating S_n over the frequency range as follows:

$${〈i_n〉}^2={\int }_0^{BW}{S}_n\left(f\right)df$$

(4)

where BW is the simulation bandwidth. The noise equivalent power (NEP), defined as the minimum optical power needed to generate a photocurrent equal to the total noise, is given as follows:

$$NEP={\displaystyle \frac{\sqrt{{〈i_n〉}^2}}{R\sqrt{BW}}}$$

(5)

where R is the responsivity of the devices. Detectivity (D) is defined as the reciprocal of NEP; thus, a higher detectivity indicates better detector performance. As detectivity depends on the device’s sensitive area and the electrical bandwidth, the specific detectivity (D*), which normalizes the signal-to-noise ratio, can be expressed as follows:

$$D^{*}={\displaystyle \frac{\sqrt{A}}{NEP}}={\displaystyle \frac{R\sqrt{A\times BW}}{\sqrt{{〈i_n〉}^2}}}$$

(6)

where A is the device area. The calculated D* values for the conventional and novel devices are 7.06 × 10¹⁴ cm·Hz^1/2/W and 4.07 × 10¹⁹ cm·Hz^1/2/W, respectively. The ultra-high D* exceeding 1 × 10¹⁹ cm·Hz^1/2/W confirms the superior weak-light detection capability of the novel structure.

Figure 8a and Figure 8b illustrate the variation of responsivity with incident light wavelength for the conventional and novel PTs at a drain voltage of 5 V, respectively. The conventional device exhibits a peak responsivity of 6.31 × 10² A/W at a wavelength of 360 nm. Compared with the spectral responsivity up to 10⁵ A/W previously reported [25], which is measured at a relatively stronger intensity of 30 μW/cm², the simulated responsivity here is limited by the low UV illumination intensity and the incomplete restoration of the 2DEG density. In contrast, the novel device achieves a maximum responsivity of 8.41 × 10⁵ A/W at 360 nm, which is three orders of magnitude higher than that of the conventional device. A sharp cutoff is observed around 365 nm, corresponding to the bandgap energy of the GaN layer. Despite the relatively weak UV illumination intensity, the UVRR, defined as the ratio of the responsivity at 360 nm to that at 400 nm, reaches as high as 3.0 × 10⁸. This remarkable enhancement in the responsivity underscores the novel structure’s superior capability for weak UV light detection.

Response time is another critical figure of merit for PDs as it determines the device’s suitability for specific applications. It is defined by the speed at which the device reacts to abrupt changes in the incident signal. The normalized transient responses of the PTs at a drain voltage of 5 V were simulated using Silvaco Atlas and are presented in Figure 9. For the conventional device, the rise time—measured as the time taken for the current to increase from 10% to 90% of its peak value—is 9.49 ns, as shown in Figure 9a. Similarly, the fall time—defined as the time required for the current to decrease from 90% to 10% of its peak value—is 11.17 ns. Compared with the previously measured response time on the microsecond scale [25], the simulation results neglect the impact of etching-induced damage. This, in turn, indicates that etching damage should be minimized as much as possible during device fabrication. In contrast, the novel device demonstrates significantly faster rise and fall times of 0.41 ns each, as illustrated in Figure 9b. This remarkable improvement is attributed to the much shorter retraction distance of the depletion region during switching operations. The enhanced temporal response substantially broadens the application prospects of the novel device in high-speed UV communication, real-time sensing, and advanced optoelectronic systems.

It is worth noting that, in addition to the conventional AlGaN/GaN heterostructure, the AlScN/GaN heterojunction also represents a promising candidate for UV detection [35]. AlScN is an emerging III-nitride semiconductor that can be epitaxially grown as high-quality single-crystalline films by both molecular beam epitaxy (MBE) [36,37] and metal-organic chemical vapor deposition (MOCVD) [38,39]. Notably, a Sc composition of approximately 18% enables near lattice matching with GaN [40], facilitating the formation of high-quality heterointerfaces. Moreover, AlScN/GaN heterostructures can support high-density 2DEG [41,42], which has already been widely exploited in power and radio-frequency (RF) devices [43]. Given these advantages, AlScN/GaN also holds great promise for phototransistor applications; their enhanced polarization effects and superior crystalline quality may lead to improved photoresponsivity and faster temporal response. Overall, the AlScN/GaN heterostructure represents a compelling alternative platform for high-performance UV PTs and warrants further investigation in future studies.

5. Conclusions

In summary, a novel UV PT architecture based on the AlGaN/GaN HEMT with a buried p-GaN layer has been proposed. Comprehensive simulations using Silvaco Atlas were conducted to elucidate the operating mechanisms and performance advantages of the novel design. Experimental results from previous studies were also referenced to validate the reliability of the simulation model and its conclusions. Benefiting from the significantly reduced retraction distance of the depletion region during device turn-on operation, the proposed structure exhibits markedly enhanced weak-light detection capability and faster switching speed compared to the conventional design. Under weak UV illumination of 100 nW/cm², the device achieves a high photocurrent density of 1.68 × 10⁻³ mA/mm, a peak responsivity of 8.41 × 10⁵ A/W, a large PDCR of 2.0 × 10⁸, a high UVRR exceeding 10⁸, a superior D* of 4.07 × 10¹⁹ cm·Hz^1/2/W, and a rapid response time of 0.41/0.41 ns. These results provide valuable insights into UV PT design and demonstrate the strong potential of the proposed structure for high-speed and low-light-level detection, enabling applications such as medical imaging and environmental monitoring. Future work will focus on epitaxial growth and device fabrication, with an emphasis on improving uniformity to support the development of UV PT arrays for imaging systems.