AlGaN/GaN high electron mobility transistors (HEMTs) have lots of advantages, such as large critical electrical field, high two-dimensional electron gas (2DEG) density and high saturation velocity. Therefore, AlGaN/GaN HEMTs are suitable for realizing high-power and high-efficiency amplifiers for the next generation wireless communication, satellite communication and radar systems.^[1–4] However, excessive gate leakage current in the power AlGaN/GaN HEMT remains a critical challenge. It is commonly believed that the high reverse gate leakage current is tunneling current caused by high-density traps.^[5–10] Some of the researchers reported that the reverse gate leakage current is dominated by Frenkel–Poole emission (FPE) under high temperatures (> 250 K) and carrier transport via trapped state near the gate metal/semiconductor interface is a dominant source.^[7–9]

In this paper, the gate leakage currents of our passivated devices are found not to accord with those from the FPE model. Thus, a modified model of FPE is developed by implementing an additional leakage current path besides the gate, which is caused by passivation/semiconductor surface traps. In order to separate the additional leakage current from the total current and obtain the modified current expression, two samples of AlGaN/GaN HEMTs are fabricated. One is with ALD-AlN/PECVD-SiN_x mixed passivation, and the other is only with PECVD SiN_x passivation. It is found that the measured data and the results from the modified model are in excellent agreement with each other at the temperatures ranging from 295 K to 475 K.

The schematic cross-sections of the AlGaN/GaN HEMTs are shown in Fig. 1. Two samples of AlGaN/GaN HEMTs were fabricated by using the same material and device processing, but with different passivation. Sample A was passivated by PEALD AlN/PECVD SiN_x, and sample B was only passivated by PECVD SiN_x. The AlGaN/GaN heterostructure was grown by metal organic chemical vapor deposition (MOCVD) on a 2-inch (1 inch = 2.54 cm) SiC substrate. A 2-μm undoped GaN layer, a 1-nm AlN spacer, and a 12-nm Al_0.3Ga_0.7N barrier layer were epitaxed from the bottom to top. Room-temperature Hall measurements show the 2DEG density of 1.08 × 10¹³ cm⁻² and electron mobility of 2079 cm²/(V·s), resulting in a sheet resistance of 278 Ω/□.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. Cross sections of AlGaN/GaN HEMTs: (a) Sample A passivated by PEALD AlN/PECVD SiNx, (b) Sample B passivated by PECVD SiNx.

The device isolation was formed by Cl₂/BCl₃ plasma dry etching. Then, the Ti/Al/Ni/Au metal stack deposition was annealed at 850 °C for 30 s in nitrogen atmosphere to realize ohmic contact (0.45 Ω·mm). The source–drain spacing was 8 μm. After that, a 2-μm Schottky gate with Ni/Au (50/200 nm) metal stack was deposited by electron-beam evaporation, located in the middle of the source–drain spacing. Sample A was passivated by 5-nm ALD-AlN and 50-nm PECVD-SiN_x at 300 °C, while sample B was passivated only by 50-nm PECVD SiN_x at 300 °C.

The temperature-dependent gate current–gate voltage (I_g–V_g) curves of Samples A and B are shown in Fig. 2. The reverse gate current densities increase with increasing reverse gate bias and temperature.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. Ig–Vg characteristics of (a) sample A, (b) sample B, and ln(Jg/Er) − characteristics of (c) sample A and (d) sample B.

For the high temperatures (> 250 K), the gate leakage current is dominated by the emission of electrons via traps, which is successfully explained by the FPE model. The current density associated with FPE is given by^[8]

From Eqs. (2)–(4), it can be found that ln(J_FPE/E_r) should be a linear function of . However, the values of gate current density J_g of samples A and sample B do not accord with those from the FPE model as shown in Figs. 2(c) and 2(d) respectively. This is because there should be additional gate leakage path (Region II) besides the leakage currents in Region I for FPE as shown in Fig. 3.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. Tracks of electrons at high reverse gate voltage.

Based on the Poisson equation, the width of the electrons emitting from gate to the surface trap can be expressed as

The vertical electric field in region II (as shown in Fig. 3) can be expressed as

The gate leakage current in region II can be expressed as

Substituting Eqs. (5), (6), and (12) into Eq. (13), I_II can be written as

Based on Eqs. (11), (14), and (15), it can be concluded that the plot of I_II versus (ϕ_B − V_g)^0.5 should be a straight line.

In order to obtain the trap states at the gate metal/semiconductor interface by C_t–f characteristics, large area devices are fabricated. The gate areas of sample A′ passivated by AlN/SiN_x and sample B′ passivated by SiN_x are both 80 μm×100 μm. Figure 4 shows that the C_t−f characteristics of samples A′ and B′ are almost the same, and C_t can be expressed as^[13]

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. Ct−f characteristics of (a) sample A′ and (b) sample B′.

Based on Eq. (16) and the measured C_t−f curve, it can be concluded that the trap states at the gate metal/semiconductor interface are the same for both samples A′ and B′ (N_t and τ are 3.39 × 10¹² cm⁻²· eV⁻¹ and 4.4 μs as shown in Fig. 4). The passivation will not affect the gate metal/semiconductor interface state, but it will affect the surface state besides the gate. Therefore, only I_II of sample A and sample B are not equal, and ΔI_g can be expressed as follows:

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. Plots of ΔIg versus (ϕB − Vg)0.5.

The modified model of FPE can be given by

When |V_g| < V_th (|V_g| values for samples A and B are both nearly 3.3 V), the electric field in the semiconductor under the gate metal can be approximately expressed as

Substituting Eqs. (14), (19), and (20) into Eq. (18), I_g can be written as

Fig. 6.

	Figure Option ViewDownloadNew Window
	Fig. 6. Comparison of the results from FPE model (red dashed lines), and the modified FPE model (black solid lines) with the measured reverse Ig–Vg data for different temperatures of sample B.

The modified FPE model is only suitable for the single channel, but unsuitable for the AlGaN/GaN double-channel HEMTs.^[14–17] The capacitance in region II (C_II) is comprised of C_II1 and C_II2 (as shown in Fig. 7), where C_II2 is related to V_g. Combining with Eqs. (6), it can be concluded that the vertical electric field in region II (E_{II − y}) is related to V_g. Then, combining with Eqs. (5), (6), (12)–(14), it is clear that I_II versus (ϕ_B − V_g)^0.5 will not be a straight line any more. Therefore, the modified FPE model is unsuitable for the AlGaN/GaN double-channel HEMTs.

Fig. 7.

	Figure Option ViewDownloadNew Window
	Fig. 7. AlGaN/GaN double-channel HEMT.

By considering the non-ideal interface states, a modified model of gate leakage current in AlGaN/GaN HEMT is proposed. Comparing with the FPE model, the modified model introduces an additional leakage current path bypassing the gate (I_II). In order to separate the gate leakage currents, two samples with different passivation are fabricated. I_II is a linear function of (ϕ_B − V_g)^0.5, which is proved by the plots of temperature-dependent ΔI_g−(ϕ_B − V_g)^0.5 of samples A and B. Moreover, the modified gate leakage current model is verified by comparing with the measured temperature-dependent I_g–V_g of sample B, showing excellent agreement. In addition, it should be pointed out that the modified FPE model is only suitable for the single channel, but unsuitable for the AlGaN/GaN double-channel HEMTs.