Enhanced high electron mobility transistor featuring a lattice matched AlInGaN/GaN heterojunction and composite gate structure Open Access

GaN materials are widely recognized for their broad bandgap and high critical electric field, making them pivotal in high-power devices, particularly in radar and aerospace applications.^1–3 GaN HEMTs have gained considerable interest due to their high carrier density, large electron mobility, and high breakdown voltage.^4,5 However, conventional AlGaN/GaN HEMTs confront several challenges. The conventional AlGaN/GaN HEMT generates a significant electron population at the heterojunction interface, induced by polarization effects. These electrons are confined within a potential well at the interface, forming a two-dimensional electron gas (2DEG) that causes the device to conduct electricity even at zero gate voltage.^6,7 This normally open device complicates circuit design and increases switching power consumption. To address these issues, there is a need for designing enhanced HEMTs that can mitigate these challenges.^8–10 Enhancement-mode HEMTs reduce drive losses in applications such as Class D and Class E power amplifiers and significantly improve circuit safety and reliability in high-power switching applications. In addition, enhancement-mode HEMTs can reduce the complexity of circuit design for RF applications. Enhanced HEMT techniques encompass recessed gate technology,^11–14 P-type cap layer technology,^15–22 F-ion implantation technology,^23–25 and polarization doping technology.^26,27 Nonetheless, each of these techniques exhibits inherent limitations. For example, recessed gate technology may thin the barrier layer beneath the gate due to etching, culminating in increased gate leakage current and diminished drain saturation current.²⁸ P-type cap layer technology consumes 2DEG within the channel, resulting in a reduction in saturation current.^29–31 F-ion implantation technology imposes stringent temperature requirements, intricate process complexity, and elevated costs, curtailing its broad applicability.³² Consequently, achieving a higher threshold voltage while ensuring a larger saturation current remains is a pivotal challenge in enhanced HEMT design.

This study introduces an enhanced HEMT featuring a lattice matched AlInGaN/GaN heterojunction and a composite gate structure (CGS). The CGS, comprising a recessed gate and a P-cap layer gate, significantly boosts the device’s threshold voltage, achieving 6.3 V, with a transconductance of 177.51 mS/mm. In addition, leveraging AlInGaN enriches the concentration of 2DEG and carrier mobility, thereby elevating the device’s saturation current to 2321.15 mA/mm, marking a notable 30% enhancement over conventional HEMTs. In addition, surpassing an Al component of 0.63 effectively eliminates gate leakage current, significantly enhancing device safety.

Figure 1: shows the schematic diagrams of various HEMT structures. (a) Diagrams of the conventional AlGaN/GaN HEMT (sample I). (b) Diagrams of composite gate structure HEMT with AlGaN barrier layer (sample II). (c) Diagrams of composite gate structure HEMT with AlInGaN barrier layer (CGS-HEMT). Table I lists some of the details. The CGS-HEMT introduces a composite gate structure by amalgamating a recessed gate and a P-cap layer gate, while substituting the conventional AlGaN barrier layer with AlInGaN to form an AlInGaN/GaN heterojunction. In the traditional AlGaN/GaN HEMT, a significant amount of 2DEG formed at the heterojunction interface leads the device into a natural conducting state, thereby increasing its switching power consumption and posing safety risks. Consequently, this paper proposes an enhanced HEMT with a lattice-approximate-matched AlInGaN/GaN heterojunction and composite gate structure (CGS). On one hand, a larger Vth is achieved through recessed gate and P-type cap layer technologies to enhance device safety. On the other hand, utilizing AlInGaN with stronger spontaneous polarization intensity as the barrier layer can generate more 2DEG and higher carrier mobility, thereby increasing the drain saturation current. Optimization of certain structural parameters of the device is performed to achieve optimal performance.

This simulation study uses the Silvaco TCAD tool for two-dimensional device simulation. The simulation is completed by numerically solving basic semiconductor equations such as the Poisson equation, continuity equation, and transport equation.³³ The potential distribution is obtained by solving the Poisson equation, and important device parameters such as the width of the space charge region and the intensity of the electric field are derived. The continuity equation relates the carrier density and current density to determine the concentration distribution of electrons and holes, the current density distribution, and the current transport path. The transport equation describes the carrier motion in the semiconductor and is used to solve the total current density. The fitting work with the experimental data was carried out first to ensure the reliability of the simulation. The structure utilized during the fitting process closely resembles the experiment structure. In this simulation study, a comprehensive set of physical models was employed to ensure accurate device characterization, including the Shockley–Read–Hall (SRH) recombination model for carrier lifetime analysis, the polarization model for III-nitride heterostructures, the field-dependent mobility model for GaN, the low-field mobility model, and the Auger recombination model for high-injection effects. The low-field mobility model was specifically configured with a lattice temperature parameter of 300 K to reflect standard room-temperature operating conditions. In addition, the P-GaN cap layer is set to Schottky contact with the metal. The simulated data are calibrated against the experimental data.³⁴ As clearly demonstrated in Figs. 2(a) and 2(b), an excellent agreement is observed between the simulated and experimental results, with a deviation of less than 5% across the entire measurement range. This result shows that the physical models used are appropriate. This remarkable consistency not only validates the accuracy of the selected physical models but also confirms the appropriateness of the implemented numerical approaches and material parameters for the device under investigation.

To investigate the influence of the composite gate structure and AlInGaN barrier layer in HEMTs, this study compares sample I, sample II, and CGS-HEMT structural configurations. Figure 3 illustrates the transfer characteristics curves of sample I, sample II, and CGS-HEMT. It’s evident that the composite gate structure significantly enhances the device’s threshold voltage. Sample I exhibits a threshold voltage of -2 V, indicating a depletion-mode device. For sample II, the barrier layer is thinner because it is partially etched. This reduces the density of the two-dimensional electron gas (2DEG). Similarly, the P-type cap consumes 2DEG of the channel. Consequently, the V_th significantly increases to 6.2 V. Building upon sample II, CGS-HEMT replaces the barrier layer with AlInGaN. It is observed that CGS-HEMT exhibits a substantial increase in saturation current while sacrificing only a minimal threshold voltage. This is attributed to AlInGaN’s stronger spontaneous polarization, which generates more 2DEG. While these 2DEGs enhance the saturation current, they also contribute to a reduction in the threshold voltage. However, the 2DEG beneath the gate is consumed by the composite gate structure. Only a small portion of the 2DEG causes a slight decrease in the threshold voltage. To elucidate this mechanism clearly, Fig. 4 shows the conduction band distribution in the vertical direction of sample I, sample II, and CGS-HEMT.

As shown in Fig. 4, the conduction band of sample I is lower than the Fermi level, indicating that it is naturally conductive. However, the conduction band energy of sample II and CGS-HEMT lies above the Fermi level, requiring additional voltage to enable conduction. Consequently, the threshold voltage of sample II and CGS-HEMT is higher than that of sample I. However, for sample II and CGS-HEMT, the energy between the bottom of the conduction band and the Fermi level of sample II is greater than that of CGS-HEMT. Therefore, sample II requires a larger voltage than CGS-HEMT to operate the device. This accounts for the higher threshold voltage of sample II compared to CGS-HEMT.

Figure 5 shows the transconductance of sample I, sample II, and CGS-HEMT. The transconductance indicates gate control capability. The higher the transconductance, the stronger the gate control. It can be observed that the transconductance of sample II and CGS-HEMT is lower than that of sample I. This is because the introduction of the P-cap layer pushes the gate further away from the channel. As a result, the control of the gate over the channel is weakened, thus reducing the transconductance. However, CGS-HEMT outperforms sample II in transconductance. This superiority is attributed to CGS-HEMT’s barrier layer material being AlInGaN, which approximately matches the GaN lattice of the channel layer, resulting in reduced scattering in the channel layer. Transconductance is related to carrier mobility, and reduced scattering leads to increased carrier mobility, thereby boosting transconductance.

Figure 6 depicts the output characteristic curves of sample I, sample II, and CGS-HEMT at V_gs = 20 V. The saturation drain output current of CGS-HEMT is obviously higher than that of sample I and sample II. This is attributed to CGS-HEMT’s barrier layer being AlInGaN, which exhibits a stronger spontaneous polarization intensity compared to AlGaN. In AlInGaN/GaN heterojunctions, this polarization generates more polarization charges, thus leading to a higher density of 2DEG and consequently the highest maximum saturation source–drain current in CGS-HEMT. On the other hand, the saturation drain output current of sample II is smaller than that of sample I. This is mainly due to the introduction of a P-type GaN cap layer in sample II. Although the P-type GaN cap layer increases the conduction band, thus increasing the V_th of the device. At the same time, it also exhausts the carrier in the channel, resulting in a lower carrier concentration, which reduces the drain current. Therefore, the saturation drain output current of sample II is smaller than that of sample I.

The performance of CGS-HEMT is significantly influenced by the P-type cap layer, where the thickness and doping concentration play crucial roles as key parameters. In order to explore the influence of P-type cap layer thickness on the electrical properties of CGS-HEMT, while maintaining other parameters constant (P-type doping concentration at 1 × 10¹⁸ cm⁻³, a barrier layer of 20 nm Al_0.76In_0.17Ga_0.07N), transfer characteristics and transconductance curves were calculated for cap layer thicknesses ranging from 20 to 70 nm, as illustrated in Figs. 7 and 8. It can be observed that with increasing cap layer thickness, V_th also significantly increases, indicating that altering the cap layer thickness can adjust the device’s V_th. However, the increase in P-type cap thickness resulted in a significant reduction in transconductance. This effect occurs because the thicker cap layer pushes the gate further away from the channel, thus reducing the control of the gate over the channel. Figure 9 illustrates the threshold voltage and transconductance of CGS-HEMT at different P-cap thicknesses. A compromise suggests that an optimal cap layer thickness of 40 nm is advisable.

Following the determination of the optimal cap layer thickness, the subsequent investigation delves into the influence of P-type doping concentration on device performance. Figure 10 depicts transfer characteristic curves for varying doping concentrations within the P-type cap layer. Higher doping concentrations notably increase the V_th. This phenomenon stems from hole introduction, which depletes the two-dimensional electron gas beneath the gate, elevating the conduction band and necessitating a higher voltage for the device to conduct. Consequently, the device’s threshold voltage increases notably. Figure 11 illustrates the impact of doping concentration on device transconductance. It can be observed that variations in doping concentration do not affect the magnitude of transconductance. However, the maximum transconductance shifts to the right with increasing doping concentration. Nonetheless, we note that doping concentration significantly affects the device’s saturation output current. Therefore, Fig. 12 demonstrates the influence of different doping concentrations on device output characteristics. It is apparent that the drain current decreases significantly with increasing doping concentration. This is because the high doping concentration leads to the diffusion of holes into the GaN channel layer, which consumes 2DEG in the channel and reduces the carrier concentration. At the same time, holes also affect the carrier transport and reduce the mobility of carriers. Therefore, the drain current is significantly reduced. Figure 13 depicts the threshold voltage and saturation drain output current of CGS-HEMT at different P-type cap layer doping concentrations. To attain a larger threshold voltage, the drain current is not too small. We determined that the optimal doping concentration is 5 × 10¹⁸ cm⁻³.

Building upon the discussion, when the Al component is ∼4.47 times the In component, AlInGaN and GaN exhibit approximate lattice matching. Therefore, x = 4.47 y and z = 1-x-y = 1–5.57 y. In essence, Al_xIn_yGa_zN can be expressed as Al_4.47yIn_yGa_1-5.47yN. In this study, the value of y ranges from 0.11 to 0.17, incremented by 0.01, with corresponding adjustments in the Al and Ga components. Figure 14 illustrates the transfer characteristic curves of different component compositions of AlInGaN and Al_0.3Ga_0.7N, with a partial magnification inset. The data illustrate that increasing the Al and In components in AlInGaN lowers threshold voltage. Specifically, the threshold voltage of CGS-HEMT surpasses that of sample II only when the barrier layer is Al_0.49In_0.11Ga_0.40N. This phenomenon can be clarified through their band diagrams. Figure 15 shows the band diagrams corresponding to each component in Fig. 14, with a partial magnification inset. An increase in the Al and In components within AlInGaN results in a gradual lowering of the conduction band, thereby bringing the conduction band closer to the Fermi level. Consequently, only a small gate voltage needs to be applied to make the device conduct electricity, thereby reducing the threshold voltage. However, for sample II, its conduction band energy is only slightly lower than that of a specific AlInGaN composition. Hence, the threshold voltage of sample II is only lower than that of the CGS-HEMT with a barrier layer of Al_0.49In_0.11Ga_0.40N.

Figure 16 presents the output characteristics of CGS-HEMT with various compositions of Al_xIn_yGa_zN barrier layers and sample II with an Al_0.3Ga_0.7N barrier layer. It can be observed that sample II exhibits the minimum drain-source saturation current, with a value of I_sat = 1653.09 mA/mm. For CGS-HEMT, as the Al and In components increase, the corresponding drain-source saturation current gradually increases. Therefore, the CGS-HEMT with a barrier layer of Al_0.76In_0.17Ga_0.07N demonstrates the maximum drain-source saturation current, with a value of I_sat = 2337.96 mA/mm. The reason behind this phenomenon lies in the nearly lattice matched AlInGaN/GaN heterojunction. Even though the barrier layer loses its piezoelectric effect and the piezoelectric polarization charges disappear, the presence of the spontaneous polarization effect ensures sufficient spontaneous polarization surface charges at the interface. In addition, AlInGaN exhibits stronger spontaneous polarization intensity compared to AlGaN. This enhanced polarization effect induces a stronger internal electric field at the heterojunction, thereby amplifying the modulation of the energy band, deepening the potential well of the 2DEG, and expanding its capacity. The number of polarization charges generated by AlInGaN spontaneous polarization is greater than the total polarization charge in AlGaN. Therefore, sample II exhibits the lowest 2DEG concentration and consequently the minimum drain-source saturation current. For CGS-HEMT, under the condition of maintaining approximate lattice matching, a higher Al component results in a deeper quantum well structure and stronger restriction on electrons. Therefore, the density of 2DEG increases as the In component. Consequently, the CGS-HEMT with a barrier layer of Al_0.76In_0.17Ga_0.07N demonstrates the maximum drain-source saturation current.

Figure 17 illustrates the gate leakage current curves of the Al_xIn_yGa_zN barrier layer CGS-HEMT and sample Ⅱ. In CGS-HEMT with varying compositions, the gate leakage current decreases as the Al and In components increase, eventually reaching magnitudes as low as 10⁻⁵–0. This reduction can be attributed to the elevation of the AlInGaN barrier due to increased Al content, which impedes electron overflow from the channel to the gate, thus minimizing gate leakage current. Furthermore, in lattice matched AlInGaN/GaN heterostructures, the trap density is nearly zero, rendering trap effects negligible on gate leakage current. The gate leakage current of sample II is only smaller than that of the CGS-HEMT with a barrier layer of Al_0.49In_0.11Ga_0.40N, and the reason is shown in the band diagram in Fig. 15. The barrier height of sample II is only higher than that of the CGS-HEMT with an Al_0.49In_0.11Ga_0.40N barrier layer and lower than that of other CGS-HEMT compositions. When the Al component exceeds 0.63, there is no gate leakage current. At this point, regardless of the applied gate voltage, no current leaks from the gate, significantly enhancing device safety.

Regarding the influence of Al and In components on device performance, this study does not consider compromise relationships. With the increase of Al and In components, the slight decrease in threshold voltage compared to the significant increase in drain-source saturation current is within an acceptable range, and the device gate leakage current also decreases to 0, eliminating the safety hazard. Therefore, Al_0.76In_0.17Ga_0.07N is selected as the barrier layer for CGS-HEMT. Furthermore, Fig. 18 shows a comparison of the threshold voltage and saturated drain current of the CGS-HEMT (red star) with data from other GaN HEMTs.^1,7,39–43 It is observed that the proposed CGS-HEMT not only exhibits a higher threshold voltage but also demonstrates significantly improved output characteristics. These findings strongly suggest that this approach holds considerable promise for improving the performance of HEMT devices.

This paper proposes an enhanced high-electron-mobility transistor (CGS-HEMT) featuring a nearly lattice-matched AlInGaN/GaN heterojunction and composite gate structure. Simulation using Silvaco TCAD software demonstrates the capability of this structure to achieve enhanced HEMT performance. Furthermore, optimization of specific parameters such as the P-GaN cap layer, P-type doping concentration, and composition of the AlInGaN barrier layer is conducted. Results show that with a 40 nm thick P-GaN cap layer, a P-type doping concentration of 5 × 10¹⁸ cm⁻³, and an Al_0.76In_0.17Ga_0.07N barrier layer, the CGS-HEMT exhibits significant improvements in threshold voltage, transconductance, and drain-source saturation current. At V_ds = 10 V, the threshold voltage and transconductance are 7.3 V and 177.5 mS/mm, respectively. At V_gs = 20 V, the drain-source saturation current can reach 2337.96 mA/mm. Therefore, the CGS-HEMT proposed in this paper has great potential in GaN HEMT devices.

This work was supported by the Natural Science Foundation of Fujian Province (Grant Nos. 2023J011458 and 2022J011274) and the Industry-University-Research Project of Xiamen City (Grant No. 2023CXY0421).

The authors have no conflicts to disclose.

Kai Niu: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Software (lead); Writing – original draft (lead); Writing – review & editing (lead). Hao-Xiang Lin: Software (equal). Li-E. Cai: Funding acquisition (lead); Investigation (lead); Methodology (lead); Project administration (lead); Resources (lead); Software (equal); Supervision (lead). Zhi-Chao Chen: Investigation (equal); Software (equal). Zhi-Yu Ma: Software (equal). Yi-Fei Chen: Software (equal). Xiang-Yu Liu: Software (equal). Chuan-Tao Sun: Software (equal). Hai-Feng Lin: Supervision (equal). Zai-Jun Cheng: Supervision (equal).

The data that support the findings of this study are included within the article.