Impact of Charge-Trapping Effects on Reliability Instability in Al_xGa_1−xN/GaN High-Electron-Mobility Transistors with Various Al Compositions

Abstract

In this study, we present a detailed analysis of trapping characteristics at the Al_xGa_1−xN/GaN interface of Al_xGa_1−xN/GaN high-electron-mobility transistors (HEMTs) with reliability assessments, demonstrating how the composition of the Al in the Al_xGa_1−xN barrier impacts the performance of the device. Reliability instability assessment in two different Al_xGa_1−xN/GaN HEMTs [x = 0.25, 0.45] using a single-pulse I_D–V_D characterization technique revealed higher drain-current degradation (∆I_D) with pulse time for Al_0.45Ga_0.55N/GaN devices which correlates to the fast-transient charge-trapping in the defect sites near the interface of Al_xGa_1−xN/GaN. Constant voltage stress (CVS) measurement was used to analyze the charge-trapping phenomena of the channel carriers for long-term reliability testing. Al_0.45Ga_0.55N/GaN devices exhibited higher-threshold voltage shifting (∆V_T) caused by stress electric fields, verifying the interfacial deterioration phenomenon. Defect sites near the interface of the AlGaN barrier responded to the stress electric fields and captured channel electrons—resulting in these charging effects that could be partially reversed using recovery voltages. The quantitative extraction of volume trap density (N_t) using 1/f low-frequency noise characterizations unveiled a 40% reduced N_t for the Al_0.25Ga_0.75N/GaN device, further verifying the higher trapping phenomena in the Al_0.45Ga_0.55N barrier caused by the rougher Al_0.45Ga_0.55N/GaN interface.

1. Introduction

High-electron-mobility transistors (HEMTs) based on III-V materials have been the next generation of high-power, high-radio-frequency, and high-temperature devices because of their high carrier concentration, high carrier mobility, and high breakdown voltage [1,2,3]. GaN-based HEMTs have recently attracted much attention because of their remarkable material properties and device performances, notably in high-power and RF applications up to the sub-terahertz regime [4,5,6]. These advantageous properties and performances are caused primarily by the excellent quality of the epitaxial layer consisting of the Al_xGa_1−xN barrier and the GaN channel layer, resulting from the fundamental electronic properties of two-dimensional electron gas (2DEG) on top of Si, Sapphire, and silicon carbide (SiC) substrates [7,8,9]. Because high-density 2DEG accumulates at the Al_xGa_1−xN/GaN interface, those electronic properties would reflect the quality of the interface via scattering procedures caused by dislocations [10]. During device operation, the interface quality of the Al_xGa_1−xN/GaN is essential for improving carrier transport in the channel [10].

Reliability concerns of Al_xGa_1−xN/GaN HEMTs have been caused by trap effects related to a drain, gate lag, and current collapse with various types of degradation [11,12,13]. The AlGaN layer, which is usually the surface layer and has an interface with the GaN channel, is a source of reliability instability, including trapping in the gate-to-drain access region, deep-level, and Al_xGa_1−xN/GaN interface [14]. Although most surface traps can be passivated with different kinds of passivation layers (e.g., SiN_x, SiO₂, and Al₂O₃), optimization of the traps inside the Al_xGa_1−xN layer and the Al_xGa_1−xN/GaN interface is still an ongoing investigation. The reliability instability issues of the Al_xGa_1−xN/GaN HEMTs are worse than that of conventional Si-based devices because of the interface quality of the Al_xGa_1−xN/GaN [15,16,17].

The defect sites in the AlGaN barrier layer and the interface Al_xGa_1−xN/GaN are the predominant cause of the transient-charging effects (Figure 1) [18,19]. The transient-charging effects follow two different processes, fast and slow transient charging. Channel carriers are easily injected into shallow defects (fast-transient charging) in the Al_xGa_1−xN barrier layer and the interface of Al_xGa_1−xN/GaN. Then, trapped charges in the shallow trap site follow thermally activated electron migration via trap-to-trap conduction (slow transient charging). The fast-transient-charging effect is responsible for mobility degradation and threshold voltage (V_T) instability in AlGaN/GaN HEMTs, while the slow transient charging causes long-term stress V_T instability. All of these are major concerns for implementing GaN-based HEMTs in future applications. Improving the reliability instability of Al_xGa_1−xN/GaN HEMTs requires thoroughly analyzing the trapping effects because the channel carriers can easily tunnel into the pre-existing defect sites in the Al_xGa_1−xN barrier layer and the interface Al_xGa_1−xN/GaN.

In this study, a comprehensive analysis of the trapping effects in Al_xGa_1−xN/GaN HEMTs with varying Al compositions was performed to optimize the device structure to obtain improved performance. We performed a reliability instability assessment in two different Al_xGa_1−xN/GaN HEMTs [x = 0.25, 0.45] using a single-pulse I_D–V_D technique that could be demonstrated by fast-transient-charging effects. For long-term reliability testing, we performed constant voltage stress (CVS) measurements under high-drain bias conditions to analyze the charge-trapping phenomena of the channel carriers. V_T shifting during constant voltage stress was compared between the devices to verify the interfacial degradation phenomena. Furthermore, the flicker noise characteristics were analyzed to gain knowledge of the dominant defect locations of the two structures. Finally, to verify the quantitative analysis of the trapping effects, the trap density (N_t) of both samples was calculated using the carrier mobility fluctuation (CMF) model [20].

2. Experimental Details

Figure 2 represents the fabrication process flow and the cross-sectional illustration of the Al_xGa_1−xN/GaN HEMTs used in this study. The epitaxial layers were grown on a semi-insulating SiC substrate using metal-organic chemical vapor deposition (MOCVD). Each layer was grown in the following order: ~270 nm of an AlN buffer layer, ~400 nm of GaN channel, and ~20 nm of an Al_xGa_1−xN [x = 0.25, 0.45] barrier layer. Cl₂-based inductively-coupled plasma (ICP) etching was used to isolate the devices for the mesa isolation procedure. The substrate was then cleaned for 30 s with a 1:5 solution of HCl and deionized water to remove any native oxide. The ohmic metallization of the source and drain was performed by an e-beam evaporator with a metal scheme of Ti/Al/Ni/Au (25/160/40/100 nm). Rapid thermal annealing was used to alloy the ohmic contacts at 830 °C and under N₂ ambient for 30 s. An additional padding layer of Ti/Au (20/300 nm) was deposited by an e-beam evaporator to ensure proper probe contact during device characterization. The contact resistance (R_C) and sheet resistance (R_SH) from TLM measurements were 0.25 Ω·mm and 380 Ω/□ for the Al = 25% sample and 0.28 Ω·mm and 420 Ω/□ for the Al = 45% sample, respectively. Finally, the gate pattern was defined using e-beam lithography, and a T-shaped Ni/Au (20/400 nm) short-channel gate was deposited. Gate-source and gate-drain distances were kept symmetrical, and the drain-to-source distance was fixed at 2 µm. All electrical characteristics were analyzed using the Keysight B1500A semiconductor parameter analyzer. The fast-transient-charging effect characterization was conducted using the single-pulse I_D–V_D with a pair of B1530A waveform generator modules. For the 1/f low-frequency flicker noise measurements, we used a dynamic signal analyzer HP 35670A and a current preamplifier SR570.

3. Results and Discussion

3.1. Charge-Trapping Analysis with Pulsed I–V

Figure 3a shows the DC transfer characteristics comparison of the devices with respect to the gate overdrive voltage (V_GS–V_T). Although the device characteristics are quite similar in DC measurements, the Al = 25% sample showed slightly higher drain current I_D (at high V_GS–V_T) and transconductance G_m. Figure 3b illustrates single-pulse I_D−V_D characteristics with different Al compositions in the barrier layer. The output characteristics of a single-pulse I_D–V_D technique with the rise (t_r) and fall time (t_f) of 50 ns were measured with a V_D sweep. Rise and fall times were kept small to achieve trap-free I_D–V_D characteristics [21]. A short pulse width of the gate and drain was applied during the measurement, reducing fast-transient trapping/de-trapping effects.

A significant reduction in the drain-current (∆I_D) is observed during the fall-down trace for the Al = 45% sample compared with the Al = 25% sample, related to the filling of the resonant traps during the rise time and pulse width through the fast-transient charging process. DC measurements cause a significant degradation because of higher integration time (~5 ms) [22,23].

Figure 3c depicts the fast degradation in the drain-current with respect to time when the gate pulse is V_GS–V_T = 2 V and drain bias is V_DS = 5 V, corresponding to the pulsed I_D−V_D characteristics. Channel carriers are trapped in the trap states near the interface of the Al_xGa_1−xN barrier layer and the interface Al_xGa_1−xN/GaN [23,24]. I_D degradation for Al_0.25Ga_0.75N/GaN device during 500 ns pulse width is ~20 mA/mm, while Al_0.45Ga_0.55N/GaN device illustrates I_D degradation in ~67 mA/mm. A significantly higher I_D degradation for the Al = 45% sample corresponds to a rougher interface between the barrier and GaN channel caused by higher lattice mismatching.

Drain-current degradation with respect to pulsed time is related to charge-trapping in the defect sites, which can be explained by the model of charging processes [25]. Channel carriers can be tunneled into the shallow defect sites in the AlGaN barrier layer and can occur to thermally activated electron migration between the defect sites with temperature dependency. The location of these defect sites is below the conduction band, as illustrated in Figure 1. Because of the extremely low trap energy of these shallow traps and the high density of states (DOE) from the GaN conduction band, the charging process will have a fast charging time. Slow transient charging can be attributed to the capture of secondary electrons induced from the trapped charges from the fast charging process.

3.2. Charge-Trapping Analysis with Constant Voltage Stress Condition

A long-term reliability evaluation was performed under high electric field conditions to verify the interfacial degradation from charge-trapping. Figure 4a illustrates the charge-trapping and de-trapping characteristics of two samples during a complete cycle of constant voltage stress at both gate and drain and relaxation cycle. Applied stress conditions were V_GS = 2 V and V_DS = 5 V. Threshold voltage shifting (∆V_T) from trapping in the interface states was evident. Channel carriers are trapped in the defect sites of the Al_xGa_1−xN barrier via the interface caused by a high electric field and thin barrier layer [24,26]. The degradation in V_T is consistent with the electron trapping at the Al_xGa_1−xN barrier layer defect locations from the GaN channel layer. This trapping phenomenon can be partially recovered by applying recovery voltages of V_GS and V_DS = 0 V.

The fast-transient trapping effect, which is active during a short (<1 ms), is accountable for the substantial change in the initial V_T (1 s). This effect is caused by the tunneling of channel carriers in the pre-existing defect sites inside the Al_xGa_1−xN barrier. The ∆V_T characteristics at Al = 45% had a higher initial ∆V_T and more degradation than at Al = 25%.

The time dependence of the V_T was investigated to quantify the charge-trapping phenomenon (Figure 4b). The fast-transient charge-trapping component, which is supposed to saturate fully after 1 s of stress, may be eliminated, and the power-law equation can be used to describe the time dependence ∆V_T~tⁿ of the ∆V_T (∆V_T − ∆V_T.initial (1 s)) [23,27,28]. Both devices degrade according to the power-law kinetics. Time exponent, n, is in the range of 0.17–0.21, a similar but somewhat lower range than for the Al = 25% device, corresponding to a lower interfacial degradation [29]. Regardless of the value of n, the ∆V_T values of Al = 45% devices are much higher than the Al = 25% device associated with higher trap states in the Al_0.45Ga_0.55N barrier.

3.3. Quantitative Analysis of Trap Density with 1/f Low-Frequency Noise

Low-frequency noise (LFN) is an effective tool for analyzing the interface states in a semiconductor device—predominantly responsible for performance degradation. A flicker noise(1/f noise) is usually generated from the following two causes: Carrier Number Fluctuation (CNF) and Carrier Mobility Fluctuation (CMF) [20]. Both are related to charge-trapping from the channel to the gate dielectric or barrier layer. Carrier interaction between the channel and the near-interface dielectric/barrier traps causes CNF noise. These charging effects also cause fluctuation in carrier mobility and result in correlated mobility fluctuations [30,31]. Both CNF and CMF should be considered for the quantitative trap extraction to evaluate accurate charge-trapping phenomena. With the CMF model proposed in the literature [20], it is possible to gain knowledge on the 1/f noise in all the operation regions (from linear to saturation and weak to strong inversion).

The 1/f noise measurements were performed from 1 Hz to 10 kHz at a fixed drain bias of V_DS = 0.5 V from off-state to accumulation, including the linear region. Figure 5a illustrates the normalized power spectral density (S_ID/I_D²) with respect to the frequency at V_GS = V_T condition. The power-law equation (1/f^ℽ function) is used to explain the frequency dependency of power spectral density (PSD) [18]. The 1/f^ℽ function was fitted with the measured data over the frequency range of 1 Hz to 10 kHz to extract the value of the frequency component (ℽ) [19]. Based on

Table 1. Comparison of the key reliability parameters of the AlGaN/GaN HEMTs.

Sample	∆ID [mA/mm]	n	ℽ	Nt [cm−3·eV−1]
Al0.25Ga0.75N/GaN	20	0.17	1	1.8 × 1019
Al0.45Ga0.55N/GaN	67	0.21	1.3	3 × 1019

, the value of ℽ is in the range of 1–1.3 (near 1), indicating that the defects/traps had uniform depth and energy [32]. Al = 45% devices had a ℽ value of 1.3 (Over 1), indicating that the most dominant trap locations are close to the interface of the Al_0.45Ga_0.55N barrier and GaN channel.

For the quantitative analysis of the trap states, the trap density (N_t) was extracted using the CMF model, as represented by [20,31]:

$$\frac{S_{ID}}{I_D^2}={\left(\frac{G_m}{I_D}\right)}^2{\left(1+{\alpha }_{SC}{\mu }_{eff}{C}_B\frac{I_D}{G_m}\right)}^2{S}_{Vfb}$$

(1)

where the variable α_sc is the coefficient for Coulomb scattering, µ_eff is the effective mobility of the carriers, and C_B is the capacitance per unit area of the AlGaN barrier. S_Vfb is the flat-band voltage and can be defined as follows [33,34,35]:

$$S_{Vfb}=\frac{q^{2}kT\lambda N_t}{WL{C}_B^{2}f}$$

(2)

where q, kT, and N_t are symbols used to represent elemental charge, thermal energy, and trap density, respectively. λ = [4π(2m*Φ_B)^1/2/h]⁻¹ represents the attenuation tunneling distance (Φ_B denotes the barrier height) [36]. Figure 5b illustrates a good fitting between the normalized drain-current power spectral density (S_ID/I_D²) and the right side of Equation (1), which prevails in the CMF model. Using Equation (2) and the S_Vfb extracted from the fitting, N_t for both devices was extracted. Al = 45% devices had a 40% higher N_t value of 3 × 10¹⁹ cm⁻³·eV⁻¹ compared with 1.8 × 10¹⁹ cm⁻³·eV⁻¹ for the Al = 25% devices. The reason for these noise characteristics is attributed to the fact that the Al_0.45Ga_0.55N/GaN interface creates higher defect sites near the interface, which increases the probability of the channel electron tunneling into the AlGaN barrier layer.

4. Conclusions

We demonstrated an in-depth trapping characteristic analysis of the Al_xGa_1−xN/GaN interface of AlGaN/GaN HEMTs based on the Al composition in the Al_xGa_1−xN barrier and how it affects device performance. Higher I_D degradation for the Al_0.45Ga_0.55N/GaN devices during the pulsed I_D−V_D characterization was attributed to the higher fast-transient trapping in the Al_0.45Ga_0.55N/GaN interface and reliability instability. During constant voltage stress conditions, the Al_0.45Ga_0.55N/GaN device had a higher V_T shift corresponding to higher trapping in the Al_0.45Ga_0.55N barrier. A larger time exponent n in the Al_0.45Ga_0.55N/GaN device indicated higher interfacial degradation. During quantitative extraction of volume trap density, the Al_0.45Ga_0.55N/GaN device had a 40% higher N_t, further verifying the higher trapping phenomena in the Al_0.45Ga_0.55N barrier caused by the rougher Al_0.45Ga_0.55N/GaN interface. These results demonstrate that trapping effects, which impact device performance considerably, are influenced primarily by the quality of the interface between the AlGaN and GaN layers. Future applications of the GaN HEMT devices can benefit from enhanced device properties by lowering the Al content to reduce lattice mismatches.