Improved V_th Stability and Gate Reliability of GaN-Based MIS-HEMTs by Employing Alternating O₂ Plasma Treatment

Abstract

The V_th stability and gate reliability of AlGaN/GaN metal–insulator–semiconductor high-electron-mobility transistors (MIS-HEMTs) with alternating O₂ plasma treatment were systematically investigated in this article. It was found that the conduction band offset at the Al₂O₃/AlGaN interface was elevated to 2.4 eV, which contributed to the suppressed gate leakage current. The time-dependent dielectric breakdown (TDDB) test results showed that the ALD-Al₂O₃ with the alternating O₂ plasma treatment had better quality and reliability. The AlGaN/GaN MIS-HEMT with the alternating O₂ plasma treatment demonstrated remarkable advantages in higher V_th stability under high-temperature and long-term gate bias stress.

1. Introduction

AlGaN/GaN metal–insulator–semiconductor high electron-mobility transistors (MIS-HEMTs) have superior properties, including suppressed gate leakage current, large forward gate swing range [1,2], which is required by power switches in high-efficiency, high-speed power systems [3,4,5]. Different insulators (e.g., Al₂O₃, HfO₂, SiO₂, AlN and SiN_x) [6,7,8] have been used as AlGaN/GaN MIS-HEMTs gate dielectric. The atomic layer-deposited (ALD) Al₂O₃ is more preferred because of its larger conduction band offset, high dielectric constant and high breakdown field values [9,10,11]. However, it has been reported that there is a large amount of hydroxyl (-OH) groups in ALD-Al₂O₃ [12,13] that use trimethylaluminum (TMA) and water as precursors. These -OH groups act as trap states and cause the AlGaN/GaN MIS-HEMTs to suffer from serious gate reliability and threshold voltage (V_th) instability challenge [14,15].

It is suggested that using O₃ as an oxidant during the deposition process of Al₂O₃ can improve device performance [16], but the carbon impurity in Al₂O₃ film increases [17]. It has been reported that there is less trap state density in the O₂ plasma-assisted ALD-Al₂O₃ film [18,19]. It has been reported that adding O₂ plasma in each ALD cycle can improve Al₂O₃ film quality [20]. However, the AlGaN surface can be damaged by O₂ plasma at the initial stage of Al₂O₃ film deposition, since the O₂ plasma can introduce deep-level traps at the AlGaN surface, which leads to device performance degradation and current collapse [21]. Meanwhile, there is little research on the threshold stability and gate reliability of the ALD-Al₂O₃ gate dielectric. We have already characterized the trap states and performance of the device with alternating O₂ plasma treatment in our previous articles [22]. In this work, the V_th stability and gate reliability characteristics of the AlGaN/GaN MIS-HEMTs with the alternating O₂ plasma treatment were investigated.

2. Device Structure and Fabrication Process

The AlGaN/GaN MIS-HEMTs were fabricated on the AlGaN/GaN heterostructure epitaxial sample, which was grown by metal–organic chemical vapor deposition (MOCVD). It consists of a 20 nm undoped Al_0.23Ga_0.77N barrier layer, 180 nm unintentionally doped GaN channel layer and 5.1 μm C-doped GaN buffer layer grown using MOCVD on a 6-inch Si (111) substrate. Figure 1a shows the schematic cross-sectional illustration of the AlGaN/GaN MIS-HEMTs. The AlGaN/GaN MIS-HEMTs process began with AlN/SiN_x passivation layer deposition. The device active region was isolated by mesa etching using BCl₃/Ar. Then, a Ti/Al/Ni/Au metal stack with a thickness of 20/160/50/50 nm was deposited by Electron Beam Evaporation (EBE) on the source/drain region, and ohmic contact was achieved by rapid thermal process (RTP) at 780 °C for 30 s in N₂ ambient. The transfer length method (TLM) test results show that the contact resistance was 1Ω·mm. The ALD-Al₂O₃ gate dielectrics with and without the alternating O₂ plasma treatment were deposited and denoted as devices A and B, respectively. Finally, Ni/Au metal stack was deposited for the gate electrode.

The schematic process flow of depositing ALD-Al₂O₃ film with the alternating O₂ plasma treatment is shown in Figure 1b. The entire depositing process was carried out using a Sentech SI ALD system. The deposition process consisted of a cycle of two sub-processes. Sub-process one: 4 nm ALD-Al₂O₃ was deposited with TMA and H₂O as precursors. Sub-process two: The film deposited in sub-process one was treated with in situ O₂ plasma for 2 min. The O₂ gas flow was 100 sccm, gas pressure was 15 Pa, and the plasma power was 100 W. Throughout the process, the substrate temperature was maintained at 300 °C. Sub-process one and sub-process two were repeated five times. Finally, a 20 nm ALD-Al₂O₃ film with alternating O₂ plasma treatment was obtained. It is worth noting that the deposited 4 nm ALD-Al₂O₃ film could serve as a protective layer on the AlGaN surface to prevent the O₂ plasma damage [23].

3. Results and Discussion

Figure 2 exhibits the atomic force microscopy (AFM) image of the Al₂O₃ film surface with an area of 2 µm × 2 µm. For the Al₂O₃ film with and without the alternating O₂ plasma treatment, the root mean square (RMS) of surface roughness is 0.094 nm and 0.096 nm, respectively. This indicates that the alternating O₂ plasma treatment will not have adverse effects on the surface morphology of the Al₂O₃ film.

The gate leakage current density of the device is illustrated in Figure 3a. The gate leakage density of device A significantly decreased compared with that of device B. The breakdown voltage of device A also improved. In order to explore the reasons for the reduction of gate leakage in device A, the gate leakage mechanism was analyzed. Considering that ALD-Al₂O₃ has good quality, Fowler–Nordheim (FN) tunneling was believed to be the dominant gate leakage mechanism [24]. The effective barrier width of the dielectric narrowed under the forward gate voltage, and driven by the electric field in the gate dielectric, electrons at the Al₂O₃/AlGaN interface could directly tunnel through the gate dielectric. Leakage current by FN tunneling is illustrated in Figure 3b, which can be expressed as

$$J_{FN}=\frac{q^2}{16\pi {\hslash \phi }_{ox}}{E}_{ox}^{2}exp(-\frac{4\sqrt{2m*{\left(q{\phi }_{ox}\right)}^3}}{3\hslash q{E}_{ox}})$$

(1)

where q is the charge of electrons, $\hslash$ is the Planck’s constant, ${\phi }_{ox}$ is the conduction band offset at Al₂O₃/AlGaN interface, $E_{ox}$ is the electric field strength in Al₂O₃ gate dielectric, $m^{*}$ is the effective electron mass in Al₂O₃, and 0.23$m_0$ of an electron mass was used for the Al₂O₃ film [19]. The FN plots of log $(J/E_{ox}^2)$ versus $1/E_{ox}$ were straight lines, as shown in Figure 3b, indicating that FN tunneling was the dominant gate leakage mechanism under a high electric field. The linear slope was used to extract the conduction band offset at the Al₂O₃/AlGaN interface, which were 2.40 and 1.87 eV, respectively, for devices A and B. The lower gate leakage current density of device A was attributed to the higher conduction band offset at the Al₂O₃/AlGaN interface. The conduction band offset for device A was larger than the value of 2.2 eV in Ref. [25].

Time-dependent dielectric breakdown (TDDB) is one of the most common characterization methods for evaluating gate dielectric reliability [26]. The testing process of TDDB involves applying a constant bias stress on the gate dielectric for a long time, and monitoring the variation in leakage current passing through the dielectric layer. The quality of the gate dielectric can be evaluated using the magnitude of leakage and the time to breakdown (t_BD) under the same gate bias stress. The reasons for leakage current and breakdown of the gate dielectric are as follows. There are defects inside the gate dielectric at the initial state, and these defects are mainly bulk defects formed during the sedimentation process. Applying electrical stress to the gate dielectric can induce random defects within the gate dielectric, causing leakage current. In addition, when electrons accelerate through the gate dielectric, it can also cause damage to the gate dielectric and form new defects. When the defects form a continuous seepage path inside the gate dielectric, the leakage current rapidly increases and the gate dielectric layer undergoes breakdown. High electrical stress will accelerate the generation of defects, generate higher leakage current, and thus accelerate the breakdown process of the gate dielectric. Due to the different breakdown voltages for device A and device B, two sets of gate bias were used to stress the devices A and B, respectively. The time-dependent gate breakdown characteristics are shown in Figure 4a,b. The t_BD for gate dielectric at different gate voltages statistically obey the Weibull distribution, which can be described by [27]

$$F\left(t\right)=1-exp[-{\left(\frac{t}{\eta }\right)}^{\beta }]$$

(2)

where t is the gate voltage application time, $\beta$ is the Weibull slope, and $\eta$ is the characteristic lifespan or scale factor.

The Weibull failure distribution is linearly simplified as follows:

$$ln[-ln(1-F\left(t\right)\left)\right]=\beta ln\left(t\right)-\beta ln\left(\eta \right)$$

(3)

A larger $\beta$ indicates a more concentrated distribution of t_BD in the breakdown characteristic [28]. Figure 4c,d shows the Weibull plots of the t_BD distribution for devices A and B. Weibull slope $\beta$ was extracted and found to be 5 and 4.5 for devices A and B, which indicated that ALD-Al₂O₃ with alternating O₂ plasma treatment has better quality and reliability. These results were larger than the value of 4.45 in Ref. [29], although Al₂O₃ had a thicker thickness (25 nm).

The V_th instability induced by high-temperature operation and long-term gate stress limits the commercial application of AlGaN/GaN MIS-HEMTs. To investigate the thermal stability of V_th, the transfer characteristic curves of device A and device B at various temperatures from 30 °C to 150 °C in steps of 30 °C were measured, as is shown in Figure 5. The OFF-state drain current increased by about two orders of magnitude as a result of increased buffer leakage current [30]. The ON-state I_DS decreased slightly due to the lower carrier mobility at higher temperatures [31].

Figure 6 shows the temperature-dependence V_th shift (ΔV_th) for devices A and B. The device A demonstrated a better V_th thermal stability and the maximum ΔV_th of 0.24 V was achieved at 150 °C at the I_DS level of 1 μA/mm, less than that of 0.55 V for device B at 150 °C. However, ΔV_th in Ref. [30] is larger than 1V at the same test temperature.

To assess the V_th stability of the device under long-time gate bias stress, the forward gate bias stress (V_{G_stress}) of 2 V was applied to the gate with source and drain grounded. A quick I_D-V_GS test was conducted after certain interval times (1, 5, 10, 20, 40, 60, 80, 100, 200, 400, 500, 600, 800, 1000, 2000, and 3000 s). Figure 7 shows the multiple I_D-V_GS curves throughout the entire testing process. The I_D-V_GS curves positively shift under the forward gate bias stress, which corresponded to electrons in the channel being trapped [32]. During the forward gate bias stress application process, the electric field in the AlGaN barrier layer is very high, especially at the edge of the gate. A strong electric field can cause electrons to tunnel from the defects in the AlGaN barrier layer to the valence band, which is known as Zener trapping. Electrons in 2DEG are then emitted into the defects, causing a decrease in electron concentration in the channel and a positive shift in the V_th.

As shown in Figure 8, the extracted ΔV_th after the 3000 s gate bias stress of 2 V were 0.55 V and 0.88 V for devices A and B, respectively. Device A showed a relatively small ΔV_th compared to device B. This indicated that the trap state density in the dielectric was reduced by the alternating O₂ plasma treatment [15]. Furthermore, subthreshold slope (SS) did not show any significant changes after long-time gate bias stress for both devices.

4. Conclusions

The V_th stability and gate reliability of the AlGaN/GaN MIS-HEMTs with alternating O₂ plasma treatment were investigated in this article. The conduction band offset at the Al₂O₃/AlGaN interface was elevated to 2.4 eV after the alternating O₂ plasma treatment, and hence resulted in lower gate leakage current density. The gate dielectric reliability was also improved, which was characterized by the TDDB test. The device with the alternating O₂ plasma treatment also showed improved thermal stability of V_th and long-time gate bias induced V_th instability. The proposed O₂ plasma alternating treatment technique was found to exhibit superior performance, which is highly desirable in high-performance and reliable power devices.