Influence of Gate Depletion Layer Width on Radiation Resistance of Silicon Carbide Junction Field-Effect Transistors

Abstract

Silicon carbide junction field-effect transistors (SiC JFETs) are promising candidates as devices applicable to radiation conditions, such as the decommissioning of nuclear facilities or the space environment. We investigate the origin of the threshold volage (V_th) shift and hysteresis of differently structured SiC JFETs. A large positive V_th shift and hysteresis are observed for a depletion-type JFET with a larger depletion layer width. With changing the sweep range of the gate voltage and depletion width, the V_th shift was positively proportional to the difference between the channel depth and depletion width (channel depth–gate depletion width). By illuminating the sub-band gap light, the V_th of the irradiated depletion JFETs recovers close to nonirradiated ones, while a smaller shift and hysteresis are observed for the enhancement type with a narrower width. It can be interpreted that positive charges generated in a gate depletion layer cause a positive V_th shift. When they are swept out from the depletion layer and trapped in the channel, this gives rise to a further V_th shift and hysteresis in gamma-irradiated SiC JFETs.

1. Introduction

A three-times-wider band gap of silicon carbide (SiC) than silicon (Si) can effectively suppress the number of electron–hole pairs generated in devices using gamma rays [1]. Due to a lower amount of displacement damage caused by Compton electrons generated through irradiation, SiC devices are attractive for long-term use for the decommissioning of nuclear facilities or in the space environment [2,3,4,5]. In particular, the device structure of SiC junction field-effect transistors (JFETs) is of interest for electronics working in high-temperature and/or high-radiation conditions [6,7]. The gate of SiC JFETs consists of a p-type-doped region instead of the gate oxide; thus, the threshold voltage (V_th) shift caused by a positive charge (holes) accumulated in the oxide is suppressed. Less than 1 V of the V_th shift after MGy irradiation is much smaller than that of SiC metal-oxide semiconductor (MOS) FETs [8,9,10,11,12,13]. Thus far, there have been a few reports regarding the radiation effect of SiC JFETs. Most of them showed no significant degradation in device characteristics after several MGy irradiations [8,9,10]. In contrast, a V_th shift due to a positively charged passivation oxide was reported [11]. We previously reported a positive shift in V_th with increasing doses [13]. Furthermore, V_th hysteresis appeared due to sweep direction and a range of gate voltages (V_Gs).

The V_th shift and its instability are some of the major reliability issues studied for SiC MOSFETs. A positive V_th shift gives rise to increases in conduction losses, whereas a negative shift causes an increased drain leakage current and power consumption [14,15]. When either shift occurs in a different inverter, it broadens the distribution of the inverter’s characteristics and contributes to the reduction in the static noise margin of the static random access memory (SRAM). V_th instability would also be a reliability concern for the development of circuits using SiC JFETs. Recently, SiC complementary JFETs (CJFETs) were developed through the ion implantation of impurities in a semi-insulating substrate as components of high-temperature circuits [16,17]. Another group showed a SiC amplifier based on a 4H-SiC MOSFET that could work well up to 500 degrees whilst improving the reliability of the metal/SiC contact [18]. In addition, aiming at installing it in robots working in nuclear facilities, a Si/4H-SiC hybrid pixel device was demonstrated [19].

However, the origin of the V_th shift and the instability of gamma-irradiated SiC JFETs are not understood. In this report, SiC JFETs were irradiated with gamma rays over 10 MGy (H₂O). Such a high dose of irradiation allowed for the radiation effects to be more clearly observable. By comparing the degradation behaviors between the two differently structured devices, how they occurred was investigated.

2. Materials and Methods

Cross-sectional views of n-channel enhancement-type (normally off) and depletion-type (normally on) JFETs fabricated in AIST are depicted in Figure 1. the Nominal channel length (L_ch) and depth (D_ch) were 2.2 and 0.6 μm, respectively. The channel width (W_ch) was 72 μm. The gate and base layer were formed through the implantation of aluminum ions into an n-type epitaxial layer. The implanted aluminum concentration profile was confirmed using secondary ion mass spectrometry (SIMS) and the simulated aluminum profile [13]. Every two or three steps of the aluminum implantation to attain a box-shaped profile were carried out for each depletion or enhancement type. The source and drain (n⁺) regions were formed through phosphorous ion implantation. Ohmic contacts were formed through the deposition of nickel/aluminum, followed by sintering. Afterwards, aluminum contact pads were deposited.

For the depletion-type JFETs, a bottom p-type layer was electrically connected with a source n⁺ layer at the source contact pad. The channel region was surrounded by a thermally grown SiO₂ layer (t_SiO2) 1 μm thick for electrical isolation. For the enhancement type, the bottom p-type base layer and gate were connected. Accordingly, the p⁺ basket-shaped gate layer formed around the channel region electrically isolated each device. The surface was also passivated with a 1 μm thick SiO₂ layer.

The Irradiation of the gamma rays was performed at room temperature under an argon (Ar) atmosphere to prevent the oxidation of the aluminum electrodes. The purity of the Ar was more than 99.9995 vol.%. The total dose was 17 Mgy (H₂O) for the depletion and 15 Mgy for the enhancement types. During irradiation, the SiC JFETs were put into an aluminum tube inside a stainless chamber in order to maintain an approximate charged particle equilibrium. No bias was applied and none of the electrodes were grounded during the irradiation. The dose rate at the position where the JFETs were irradiated was evaluated with a PMMA (poly-methylmethacrylate) dosimeter (Radia Industry Co., Ltd., Takasaki, Japan) [20,21]. The dose rate was 9–10 kGy/h (H₂O). In order to evaluate the effect of the surface-generated charges, the depletion-type JFETs were aged at room temperature in air after 17 of Mgy irradiation.

The drain current (I_D) gate voltage (V_G) and drain voltage (V_D) characteristics were measured at room temperature in the dark using Agilent 4156B or Keysight B1500A semiconductor parameter analyzers after each irradiation. During the measurement, the source and substrate were commonly grounded and a drain voltage (V_D) of 6 V was applied to the drain electrode. To investigate the changes in the electrical characteristics due to the irradiation, the V_G was swept with a 10 mV interval. The delay, which was defined as the time before the acquisition of I_D at each V_G step, was set to 0 s. The V_th was calculated through the linear extrapolation of the square root of the I_D–V_G curves in the saturation region. We irradiated 10 depletion and 4 enhancement JFETs with channel widths of 18, 36 and 72 μm. Each average V_th shift from that of a pristine device was 0.50 or 0.069 V. Each standard deviation was 0.1 or 0.02 V. The V_th hysteresis was investigated by changing the sweeping interval of the V_G from 10 to 300 mV. Moreover, at a fixed step interval of 300 mV, a measurement was carried out for different delay times and sweep ranges.

The energy levels of the traps (defects) formed in the devices were characterized by an illuminating sub-bandgap light. The electrons are excited from the valence band to the traps formed in the band gap of the SiC [22,23,24,25]. The defects formed in the SiC are usually characterized using a thermal method such as deep-level transient spectroscopy (DLTS); nevertheless, using light illumination prevents the irreversible recovery or additional degradation of irradiated devices. Light-emitting diodes (LEDs) with a center wavelength from 745 (1.67 eV) to 1545 nm (0.80 eV) were employed as the light sources. Visible and infrared spectra features of the LEDs were investigated using compact fiber spectrometers. The averaged intensity of the focused LED light onto the JFETs was 15 μW/mm². As the light energies were located below approximately a half of the 4H-SiC, 3.26 eV [26], the defects formed in the upper/bottom half of the bandgap were mainly excited during the illumination. By illuminating the JFETs with the LEDs, the I_D–V_G curves were measured. Through estimating the V_th from the I_D–V_G characteristics during the illumination, the approximate energy levels of the defects were explored.

3. Results

The I_D–V_G and I_G–V_G characteristics for the non- and irradiated SiC JFETs are shown in Figure 2a–d. Regardless of the device structure, increases in the subthreshold slope and positive shift were observed in the I_D–V_G curves after the irradiation (dashed lines). Values of the drain leakage currents in the I_D–V_G curves were nearly consistent with the I_G leakage currents, indicating leakage paths between the gate and drain electrodes having formed after the irradiation.

The threshold voltage shift (ΔV_th) from the nonirradiated devices was plotted in Figure 3. The initial V_th of the enhancement or depletion type was 0.96 or −5.27 V, respectively. The ΔV_th of the enhancement type increased positively and reached approximately 0.1 V after 15 MGy, whereas for the depletion type, a much larger positive shift was observed. Although the depletion type was a single-gated structure (Figure 1b) and the V_th was expected to be two times larger than the double-gated ones [27], the approximately six times larger ΔV_th at the same dose was of interest.

So as to evaluate the effect of the charges generated on the device surface, irradiated depletion-type JFETs were aged at room temperature. As shown in Figure 4a, the leakage current became as low as that for the nonirradiated device at 1 year after irradiation. The V_th slightly shifted toward the negative direction and was saturated after 1 year (Figure 4b. Given that the leakage paths were formed between the gate and drain electrodes, it was inferred that the leakage currents were caused due to electrons attracted by positively charged passivation oxide. As holes rapped in the oxide were annealed, the apparent positive bias applied to the gate was removed. This resulted in a slightly negative V_th shift. The aging experiment also suggested that a positive V_th shift was derived from the defects formed in the bulk region of the device.

The I_D–V_G hysteresis of the irradiated depletion-type JFETs are shown in Figure 5a. The I_D–V_G curve positively shifted and the hysteresis became larger for 300 mV. The V_th of each sweep direction (forward (FWD) or reverse (REV)) were plotted in Figure 5b. Both V_ths gradually increased and were almost saturated at 300 mV. With the V_G sweep interval at 300 mV, the I_D–V_G curves shifted negatively for a longer delay, as shown in Figure 5c. The FWD V_th plotted in Figure 5d decreased and was saturated after 4 s, while the REV V_th (brank circles) more slowly decreased to 10 s. A decay as slow as several seconds implied that carriers were captured and released via the deep hole traps [28,29,30,31]. Meanwhile, no significant hysteresis appeared for the enhancement type (Figure 5e). Due to the relatively abrupt increase in the drain current that resulted from the double-gated structure (Figure 1a), the sweep interval was limited below 80 mV to obtain sufficient data plots for the V_th estimation. The difference between the FWD and REV V_ths remained within approximately 0.05 V, as shown in Figure 5f, which was much smaller than that shown in Figure 5b.

We assumed the positive charges trapped at the depletion layer served as an apparent positive bias and pushed down the electric potential of the channel [32,33]. In order to verify this hypothesis, we illuminated the irradiated depletion JFETs to excite the electrons from the valence band to neutralize the traps. The I_D–V_G curves characterized during/without the LED illumination are shown in Figure 6a. The V_G sweep interval was 10 mV and the wavelength (energy) of the LED was 750 nm (1.67 eV). The light illumination shifted the curve toward the negative voltage side, while it did not affect the leakage current. The V_th shift from the nonirradiated device (ΔV_th) is shown in Figure 6b. A slight recovery as low as 0.1 V was already observed below 0.8 eV. With increasing light energy, a little steep slope at 1.2 eV and a small gap were observed at 1.4 eV. As the V_th shifted when the LED spectra passed the defect energy level, such a sudden increase or gap indicated that some defects were formed. Given full width at half maximum (FWHM) of the LED spectra of less than 0.1 eV, those defects could be assigned to deep hole traps HK3 (E_v + 1.24 eV) and HK4 (E_v + 1.44 eV) [26].

4. Discussion

The positive charges of the hole traps in the depletion layer were attributed to a positive V_th shift. We evaluated the depletion layer width of the JFETs assuming an abrupt doping profile [27]:

$${\Psi }_{\mathrm{b}\mathrm{i}}=\frac{1}{q}\left(E_{\mathrm{g}}-kT\ln \left(\frac{N_c}{N_D}\right)\right)$$

(1)

$$V_{\mathrm{t}\mathrm{h}}-V_{\mathrm{D}}={\Psi }_{\mathrm{b}\mathrm{i}}-{\Psi }_{\mathrm{P}}=\frac{1}{q}\left(E_{\mathrm{g}}-kT\ln \left(\frac{N_c}{N_D}\right)\right)-\frac{q{N}_{\mathrm{D}}{a}^2}{2{\epsilon }_{\mathrm{s}\mathrm{i}\mathrm{c}}{\epsilon }_0}$$

(2)

$$W_D=\sqrt{\frac{2 {\epsilon }_{sic}{\epsilon }_0({\Psi }_{\mathrm{b}\mathrm{i}}+V_D-V_G)}{q N_{\mathrm{D}}}}$$

(3)

where Ψ_bi is the built-in potential of the gate p–n junction, Ψ_P is the pinch-off potential, W_D is the depletion layer width at the drain side, E_g is the band gap energy, q, k and T are the elementary charge, Boltzmann constant and temperature, respectively, Nc is the effective density of the states in the conduction band, N_D is the donor concentration, a is the channel depth of 0.6 μm and ε_SiC and ε₀ are the relative permittivity of SiC, at 9.8 [26], and permittivity in vacuum, respectively. Taking the drain voltage V_D of 6 V, the effective donor density N_D of the nonirradiated JFET was 4.3 × 10¹⁶ cm⁻³. For simplicity, the apparent positive V_G by the trapped holes was neglected.

The calculated depletion width W_D was plotted in Figure 7a. The W_D decreased as the V_G increased and was smaller for the enhancement type, while the channel depth–depletion width) (dashed lines) increased with an increasing V_G. It was noted that that of the enhancement JFET was less than a half of the depletion type. This resulted from the short depletion width and double-gated structure, as the channel depth was 0.3 μm. Figure 7b shows the REV V_th as a function of the sweep-end voltage. The sweep interval was 300 mV. The REV V_th for the depletion type increased when increasing the sweeping range up to V_G = 3 V. In contrast, the increment for the enhancement type was smaller.

During the FWD sweeping, the holes trapped at the depletion layer were swept out [13] and were then trapped in the source side of the channel. In this case, the holes were trapped in the wider area, as an apparently more positive bias than that trapped in the depletion width was applied [29,30]. Thus, the REV V_th was expected to be nearly proportional to the channel width–depletion width. We curve fit REV V_th plotted for the depletion type and both was nearly consistent. As the electron capture cross-section of hole traps was generally small [29,30], the trapped positive charges pushed down the channel (source) potential. This was attributed to the positive shift of the REV V_th, while for the enhancement type, the narrower depletion width and smaller channel width–depletion width resulted in a smaller REV V_th.

5. Conclusions

The high-dose gamma ray irradiation effects of the SiC JFETs with different gate depletion widths were investigated. A large positive V_th shift and hysteresis were observed for the depletion-type JFET with a larger depletion layer width. With the changing sweep range of the gate voltage and depletion width, the V_th shift was positively proportional to the difference between the channel depth and depletion width (channel depth–gate depletion width). By illuminating the sub-band gap light, the V_th of the irradiated depletion JFETs recovered close to the nonirradiated ones, while the smaller shift and hysteresis were observed for the enhancement type with a narrower width. It could be interpreted that the positive charges generated in the gate depletion layer caused a positive V_th shift. When they were swept out from the depletion layer and trapped in the channel, this gave rise to a further V_th shift and hysteresis in the gamma-irradiated SiC JFETs.