Development of Catalytic-CVD SiN_x Passivation Process for AlGaN/GaN-on-Si HEMTs

Abstract

We optimized a silicon nitride (SiN_x) passivation process using a catalytic-chemical vapor deposition (Cat-CVD) system to suppress the current collapse phenomenon of AlGaN/GaN-on-Si high electron mobility transistors (HEMTs). The optimized Cat-CVD SiN_x film exhibited a high film density of 2.7 g/cm³ with a low wet etch rate (buffered oxide etchant (BOE) 10:1) of 2 nm/min and a breakdown field of 8.2 MV/cm. The AlGaN/GaN-on-Si HEMT fabricated by the optimized Cat-CVD SiN_x passivation process, which had a gate length of 1.5 μm and a source-to-drain distance of 6 μm, exhibited the maximum drain current density of 670 mA/mm and the maximum transconductance of 162 mS/mm with negligible hysteresis. We found that the optimized SiN_x film had positive charges, which were responsible for suppressing the current collapse phenomenon.

1. Introduction

AlGaN/GaN high electron mobility transistors (HEMTs) are promising candidates for microwave power amplification and power switching applications owing to their excellent characteristics, such as wide energy bandgap, as well as high breakdown field, mobility, and saturation velocity [1,2]. Although significant progress has been achieved in the past decades, surface trapping and its related current collapse phenomena remain critical for device performance and reliability. The current collapse phenomenon that degrades the dynamic characteristics of the device contributes significantly to performance instability and reliability of the device. The current collapse phenomenon can be mitigated by surface passivation along with a field plate where the dielectric passivation process is carefully optimized [3]. Among the various dielectric materials, including SiO₂, silicon nitride (SiN_x), SiON, Al₂O₃, and AlN [4,5,6,7], SiN_x films have been used frequently for the passivation layer of GaN FETs and have been reported to be effective for suppressing the current collapse phenomenon [8,9]. Various deposition systems have been utilized for SiN_x film deposition, including plasma-enhanced chemical vapor deposition (PECVD), inductively coupled plasma CVD (ICP-CVD), and low-pressure CVD (LPCVD). These systems employ high-power plasma and/or a high deposition temperature causing plasma damage at the surface and thermal budget problems. Therefore, a catalytic CVD (Cat-CVD) process was proposed to solve these problems. The molecules of the source gases are decomposed in a vacuum chamber by catalytic cracking reaction using heated catalyzers without plasma reaction and then the cracked species are transported to the substrates to form a film [10,11,12,13].

In this study, we determined high-quality SiN_x deposition process conditions using a Cat-CVD system. It was found that the optimized SiN_x passivation process was very effective in reducing the interface trap density and, thus, successfully suppressed the current collapse phenomenon, thereby improving the device performance.

2. Experimental Details

2.1. Optimization of Cat-CVD SiN_x Deposition Process

The Cat-CVD system used in this work is illustrated in Figure 1 where a tungsten wire with a diameter of 0.7 mm was used as a catalyzer located at a distance of ~70 mm apart from the substrate. Catalyzers were placed in an octagon between the gas injection and the substrate, and the full length was ~2.9 m. The catalyzer temperature was monitored by the resistance of the tungsten wire.

To optimize the SiN_x deposition process, the films were deposited on Si wafers using a Cat-CVD system with SiH₄ and NH₃ gas mixtures, during which the gas pressure, flow rate, and catalyzer temperature were varied as the process parameters. The deposition rate, refractive index, and buffered oxide etchant (BOE) etch rate of each film were measured to evaluate the film quality. Detailed process conditions and measurement results are summarized in

Table 1. Catalytic-chemical vapor deposition (Cat-CVD) silicon nitride (SiN_x) properties as function of deposition conditions.

Chuck Temp [°C]	Wire Temp [°C]	Pressure [Pa]	NH3/SiH4	Dep Rate [A/s]	R.I.	BOE (10:1) Etch Rate [nm/min]
400	1730	4	5	0.90	1.90	7.1
400	1730	4	50	0.88	1.94	4.80
350	1730	4	50	0.86	1.96	3.72
250	1730	4	45	0.80	1.97	2.22
150	1730	4	45	0.81	1.99	3.57
250	1600	4	19	0.56	2.83	-
250	1730	4	20	0.77	2.31	5.14
250	1730	4	25	0.75	2.03	2.15
250	1730	4	33	0.60	1.94	5.28
250	1650	2	50	0.30	1.90	8.02
250	1650	2	38.5	0.36	2.00	2.77
250	1650	2	33.3	0.38	2.12	2.39
250	1650	4	25	0.61	2.02	2.3
250	1650	4	29	0.60	2.12	2.0
250	1650	4	30	0.70	2.43	3.5
250	1650	8	25	0.91	1.97	5.37
250	1650	8	20	1.08	2.01	3.17
250	1650	8	16.7	1.20	2.12	3.78

.

The deposition rate and refractive index of SiN_x films obtained under different process conditions are plotted in Figure 2a,b, respectively. The deposition rate for the gas pressure of 2 Pa is <0.4 Å/s, which takes a significant period to deposit conventional passivation layers whose thicknesses are typically at least tens of nanometers. The sample with a refractive index of 1.9 was considered as a reference sample. The wet etch rate is a simple indicator to evaluate the dielectric film quality. It was reported that the wet etch rate had a strong function of the hydrogen content that is known to degrade the film quality and stability characteristics [14,15,16]. The wet etching tests were carried out at room temperature using a 10:1 buffered oxide etchant (BOE), and the results are plotted in Figure 3. The film with a refractive index of 2.12 exhibited the lowest wet etch rate of 2 nm/min, whereas the reference film exhibited a wet etch rate of 7.1 nm/min.

For a detailed analysis of the films, four different films were selected, which had a refractive index of 1.90, 2.02, 2.12, and 2.43. The surface roughness and breakdown field of each sample were measured. Moreover, X-ray reflection (XRR) and Fourier-transform infrared (FT-IR) measurements were carried out to evaluate the density and hydrogen content of the films, respectively.

Table 2. Cat-CVD SiN_x characteristics with refractive indices of 1.90, 2.02, 2.12, and 2.43.

Refractive Index of SiNx Film	Surface Roughness [nm]	Breakdown Field [MV/cm]	Density [g/cm3]	H Contents [%]
1.90	0.41	9.0	2.4	13.6
2.02	0.30	8.7	2.6	7.2
2.12	0.28	8.2	2.7	6.0
2.42	0.36	5.8	2.4	9.8

presents the surface roughness, breakdown field, film density, and hydrogen content of the selected films.

The surface roughness of the film with a refractive index of 2.12 was 0.28 nm, which is significantly lower than that of the reference film with a refractive index of 1.90. The film density of the sample with a refractive index of 2.12 was 2.7 g/cm³, which was the highest among the samples investigated. The high film density suggested that the film contained relatively less impurities. Indeed, the film density had the same tendency as the BOE wet etch rate.

Figure 4 shows the FT-IR absorption spectra of the selected films. The IR absorption spectra exhibited peaks at 830 cm⁻¹, 2100 cm⁻¹, and 3350 cm⁻¹, corresponding to Si–N, Si–H, and N–H bonding, respectively [15]. The relative hydrogen concentration in atomic % was determined from the bond concentrations calculated from the FT-IR spectra. The hydrogen content of the film with a refractive index of 2.12 was 6%, which was significantly lower than that of the reference sample.

In order to investigate the chemical composition of the film, X-ray photoelectron spectroscopy (XPS) analysis was carried on the reference film with a refractive index of 1.90 and a film with a refractive index of 2.12. Spectra of the core levels of Si, N, O, and C were collected and analyzed. Figure 5 shows the chemical compositions of the two films. Si 2p and N 1s are the main species in the SiN_x films. From Figure 5f, the oxygen impurity concentration for the film with a refractive index of 2.12 was significantly lower than that for the reference film. The oxygen impurities were reduced from 11.5% to 4.7%. Furthermore, the carbon impurity concentration decreased from 1.5% to 0.9%. As indicated in Figure 5e, the N 1s peak for the film with a refractive index of 2.12 is ascribed to the Si–N, N–O, and Si–O–N bonds whose respective binding energies are 397.7 eV, 398.1 eV, and 397.6 eV. The atomic concentrations of Si–N, N–O, and Si–O bonds in the film with a refractive index of 2.12 were 88.63%, 5.8%, and 5.58% [17], whereas those for the reference film were 70.9%, 14.13%, and 14.97%, respectively. The O 1s and Si 2p peaks were in agreement with the peaks reported in References [17,18,19]. The total amounts of Si and N peak intensity of the optimized film were 46.74% and 47.51%, respectively, while those of the reference film were 41.99% and 45%, respectively.

Based on film analysis, the deposition conditions for the film with a refractive index of 2.12 were determined to be optimum; a NH₃/SiH₄ flow ratio of 29, a gas pressure of 4 Pa, and a chuck temperature of 250 °C with a catalyzer temperature of 1650 °C. The deposition rate of the optimized SiN_x passivation process was ~0.6 Å/s with a thickness variation of 1.16% across a 4-inch wafer. The optimized film exhibited a film density of 2.7 g/cm³, a BOE etch rate of ~2 nm/min, and a breakdown field of 8.2 MV/cm.

As shown in Figure 6, a Rutherford Backscattering Spectrometry (RBS) analysis was carried out to investigate the stoichiometry of the film with a refractive index of 2.12 where a 115 nm thick film was used to obtain reliable analysis data. The stoichiometry of SiN_x film defined by x = [N]/[Si] was 1.2, which was lower than the standard stoichiometric Si₃N₄ film of 1.33 [20,21]. The smaller x value indicates a Si-rich thin film compared to the stoichiometric Si₃N₄ film, which is associated with positive charges in the film.

2.2. Device Fabrication

The epitaxial layers consisted of a 2 nm GaN capping layer, 20 nm AlGaN barrier layer, 300 nm undoped GaN layer, and 1.6 μm GaN buffer layer grown on a Si (111) substrate. After solvent cleaning, 40 nm thick SiN_x film was deposited using a Cat-CVD system to protect the GaN surface during high-temperature ohmic annealing. Ohmic contacts with a source-to-drain distance of 6 μm were formed by low-damage recess etching using a BCl₃/Cl₂-based plasma process prior to Si/Ti/Al/Mo/Au (=5/20/80/35/50 nm) metallization [22]. The ohmic contacts were annealed by a rapid thermal annealing process at 800 °C for 1 min in N₂ ambient. Mesa isolation was then performed by a BCl₃/Cl₂ plasma etching process with a relatively higher plasma power. The SiN_x protection layer was removed by an SF₆-based low-damage plasma etching process in order to eliminate any surface damage during ohmic annealing. A new SiN_x film with a thickness of 30 nm was deposited again as a passivation layer using the same Cat-CVD system. SiN_x films with three different refractive index values (1.9, 2.12, and 2.43) were deposited for comparison. The deposition conditions for each refractive index value were identical to those shown in Table 1. Post-deposition annealing was conducted at 500 °C for 10 min in N₂ ambient. Finally, a gate foot length of 1.5 μm was defined by photolithography, under which the SiN_x layer was etched by an SF₆ plasma etching process. A Ni/Mo/Au (=40/15/400 nm) metal stack was used for the Schottky gate contact with a gate overhang length of 0.5 μm toward the source and drain contacts. Metal-insulator-semiconductor (MIS) gate devices with the same gate process on the SiN_x passivation layer were fabricated without etching for analyzing the interface quality. The cross-sectional schematics of the Schottky and MIS gate HEMTs are presented in Figure 7.

3. Results

It was observed that the maximum drain current and transconductance were enhanced along with the off-state drain current reduction as the refractive index of the passivation layer increased. As revealed in Figure 8, the maximum drain current density and transconductance of the device for the optimized passivation process condition with a refractive index of 2.12 are 670 mA/mm and 162 mS/mm, respectively.

The transfer and sub-threshold slope characteristics of MIS AlGaN/GaN-on-Si HEMTs with 30 nm SiN_x layers with different refractive indices are compared in Figure 9. It should be noted that the device with a refractive index of 2.12 exhibited significant negative shift in threshold voltage with the lowest leakage current density, which is 1.3 × 10⁻⁹ A/mm, along with a small sub-threshold slope of 105 mV/dec. The negative shift in threshold voltage can be explained by positive charges in the SiN_x film in conjunction with the low H content. As compared in Table 2, the film with a refractive index of 2.12 had the lowest H content and the films with refractive indices of 1.90 and 2.43 had relatively higher H contents. The interface fixed charges and bulk charges of the film with a refractive index of 2.12 are discussed later with capacitance-voltage (C-V) measurements.

In order to investigate the current collapse phenomenon, the devices were measured under pulse mode operation with a pulse width of 200 ns and a pulse period of 1 ms where the quiescent bias conditions were varied by V_GSQ = 0/−6 V and V_DSQ = 0/10/20/30/40 V. We observed that the pulsed current-voltage characteristics were significantly improved by employing the optimized passivation process. The current collapse phenomenon was defined by the reduction of drain current density from a drain voltage of 5 V with V_DSQ = 0 V to a drain voltage of 10 V with V_DSQ = 40 V [8,9]. The current collapse reduction of the Schottky gate HEMT with the passivation film with a refractive index of 2.12 exhibited superior characteristics to others in Figure 10. The current collapse reduction was only 6.5%.

Trapping effects at the surface states are responsible for the current collapse phenomenon of AlGaN/GaN HEMTs. In order to suppress the electron trapping effects, the passivation film must have positive charges to attract the trapped electrons [23,24,25,26,27]. The net charges of the SiN_x film can be modified under the deposition conditions that alter the stoichiometry of SiN_x. To investigate the fixed charges of the passivation films, the frequency-dependent C-V characteristics of MIS gate devices were measured from 1 kHz to 1 MHz, from which the flat-band voltage and hysteresis characteristics could be determined. The MIS gate device had a circular MIS gate with a diameter of 100 μm, which was surrounded by an ohmic electrode. In Figure 10a,c, the threshold voltage is negatively shifted with increasing refractive index [24,28,29], which indicates that a high-refractive-index film has more positive fixed charges in the MIS gate than the films with lower refractive indices. In comparison with the reference film with a refractive index of 1.90, the flat-band voltage of the device with the film having a refractive index of 2.12 was negatively shifted from −7.7 V to −13 V with smaller flat band voltage hysteresis (Figure 11d). The flat-band voltage hysteresis was reduced from 440 mV for a refractive index of 1.90 to 280 mV for a refractive index of 2.12.

In order to extract the fixed charge density of the optimized film, four samples were prepared with different SiN_x thicknesses: 20, 30, 35, and 40 nm. It should be noted that the AlGaN barrier layer used for the experiments was 9 nm, which was relatively thinner than that used for device fabrication. The density of fixed charges at the SiN_x on the surface was extracted by V_th-thickness dependency. A second derivative method [29] was used to obtain a flat-band voltage from the C-V characteristics as a function of SiN_x thickness. The flat-band voltage was plotted against the normalized total cap thickness (t_n = t_n (SiN_x) + t_n (AlGaN) = t_sinx/ε_sinx + t_AlGaN/ε_AlGaN) [30]. The flat-band voltages extracted for SiN_x thicknesses of 20, 30, 35, and 40 nm were −4.4, −6.8, −8.4, and −10.6 V, respectively. Figure 12 and Figure 13 present the extracted flat-band voltages of the AlGaN/GaN MIS FETs with SiN_x dielectric films as a function of film thickness. The interface fixed charge density and bulk charge density can be extracted by [31]

$$V_{FB}=W_{MS}-\frac{Q_{int}}{{\epsilon }_0{\epsilon }_r}{t}_{ox}-\frac{Q_{bulk}}{{\epsilon }_0{\epsilon }_r}{t}^2{}_{ox,}{Q}_{total}=Q_{int}+Q_{bulk}·t_{ox}$$

(1)

where V_FB is the flat-band voltage, W_MS is the work function difference between the metal and semiconductor, Q_int is the total charge, Q_bulk is the bulk charge, and t_ox is the dielectric thickness. The total charge densities as a function of the SiN_x film thickness are indicated in Figure 14.

Assuming that the AlGaN polarization charge for an Al mole fraction of 0.2 and a barrier thickness of 9 nm was 1.5 × 10¹² cm⁻² [32], the extracted interface fixed charge density and the bulk charge density were estimated to be 6.63 × 10¹² cm⁻² and 3.3 × 10¹⁸ cm⁻³, respectively. It is suggested that the large amounts of positive interface and bulk charges of the SiN_x passivation film were responsible for the significant negative shift of V_th of the MIS HEMT as shown in Figure 9. The positive charges in the MIS gate acted as a virtual gate that required more negative voltage to deplete the 2DEG channel.

The interface trap density (D_it) for MIS gates was investigated using a conductance method [33]. The extracted D_it characteristics are plotted in Figure 15. The lowest D_it level was achieved by a refractive index of 2.12, which was 2 × 10¹² cm⁻² eV⁻¹ at a trap energy level of 0.43 eV. The low D_it level can further explain the improved current collapse phenomenon observed in the Schottky device. The current collapse characteristics observed in this work are compared with those reported in previous studies [33,34,35,36,37,38] in

Table 3. Comparison of the current collapse characteristics.

	Dielectric	Thickness [nm]	VDSQ [V]	Current Collapse Reduction [%]
This work	Cat-CVD SiNx	30	40	6.5
[34]	PECVD SiNx	20	40	10.8
[35]	Cat-CVD SiNx	60	25	30
[36]	PEALD AlN/Al2O3/PECVD SiNx	2/20/150	40	22.1
[37]	In-situ SiNx/PECVD SiO2	50/7	40	8
[38]	PECVD SiNx	40	30	14

. We suggest that the Cat-CVD SiN_x passivation process developed in this work is very promising for suppressing the current collapse phenomenon of GaN devices.

4. Conclusions

We have developed a high-quality SiN_x film deposition process using a Cat-CVD system for the passivation layer of AlGaN/GaN HEMTs. A Cat-CVD SiN_x dielectric layer has been optimized to reduce the trapping phenomenon originating from the surface states. Modification of the SiN_x film stoichiometry can result in positive net charges that can effectively suppress the trapping phenomenon. The optimized passivation process conditions were a NH₃/SiH₄ gas flow ratio of 29, gas pressure of 4 Pa, and chuck temperature of 250 °C with a catalyzer temperature of 1650 °C, which resulted in a refractive index of 2.1. The device fabricated under the optimized passivation process conditions exhibited excellent electrical characteristics. The high film density and low interface trap density of the optimized passivation film were responsible for the improved current–voltage characteristics and caused the suppression of the current collapse phenomenon. Therefore, the Cat-CVD SiN_x film is a promising passivation layer for AlGaN/GaN HEMTs.