High electron mobility in heavily sulfur-doped 4H-SiC Open Access

Silicon carbide (SiC) has attracted much attention as a material for power devices,^1,2 logic and analog devices, and sensors operational at high temperatures^3–7 owing to its superior physical properties. In SiC devices, nitrogen (N) or phosphorus (P) atoms are typically used as donors to form an n-type region. In general, sulfur (S) is not used as a donor in SiC since S has a large ionization energy (⁠ $ΔE$⁠) of 340–520 meV^8,9 and electron concentration in a S-doped region is markedly lower than that in a N- or P-doped region at room temperature.¹⁰ However, recent studies revealed that such a deep donor has a potential to improve device or circuit performance in certain cases. Noguchi et al. have proposed the use of S donors in channel regions of metal-oxide-semiconductor field-effect transistors (MOSFETs).¹¹ By adopting the S-doped channel, low channel resistance and high threshold voltage of SiC MOSFETs have been demonstrated. The authors’ group expected that a S-doped channel in junction field-effect transistors (JFETs) for complementary JFET circuits could suppress a logic threshold voltage shift caused by elevating the temperature.⁹ In spite of such proposals, fundamental studies on S $+$-implanted SiC are very limited.

In this study, the authors investigated the mobility, carrier concentration, and activation ratio in S $+$-implanted layers with various sulfur concentrations (⁠ $N S$⁠) by Hall effect measurements. It is found that the S $+$-implanted SiC exhibits unusually high electron mobility (⁠ $μ e$⁠).

The authors prepared n- and p-type 4H-SiC epitaxial layers with doping concentrations of $2× 10 14$ and $5× 10 14$ cm $− 3$⁠, respectively, grown on n-type substrates. 300-nm-thick S-doped layers were formed by ion implantation (10–350 keV) with a box-like profile with a $N S$ of $1× 10 17$– $1× 10 20$ cm $− 3$⁠. The thicknesses of the S-doped layers assumed for calculating carrier density in Hall effect measurements will be described later. Ion implantation of N $+$ (⁠ $2.3× 10 15$ cm $− 2$⁠, 10–250 keV) was carried out only for the ohmic-contact area. After ion implantation, activation annealing was performed at $1750 ° C$ for 20 min. Dopant depth profiles were obtained by secondary ion mass spectrometry (SIMS). The SIMS measurements were performed by Cs $+$ with an accelrating energy of 15 keV. Schottky barrier diodes were fabricated by deposition of Ni on the S $+$-implanted layers/n-type epitaxial layers for capacitance–voltage (⁠ $C$– $V$⁠) measurement. A clover-leaf-shaped mesa structure was formed on the S $+$-implanted layers/p-type epitaxial layers for Hall effect measurement. The ohmic contacts were formed by depositing Ni and subsequent annealing in vacuum at $950 ° C$ for 2 min.

Figure 1 shows the depth profiles of the implanted S atoms before and after the activation annealing at $1750 ° C$⁠. The dopant profiles at the same $N S$ before and after the annealing are nearly identical, indicating very little thermal diffusion even in the heavily-implanted layers (⁠ $N S>1× 10 19$ cm $− 3$⁠). The slight difference of the SIMS profiles from the samples with $N S$ of $1× 10 18$ cm $− 3$ may originate from the channeling effect during ion implantation, which is negligibly small for performing Hall effect measurements and analyzing the results. To estimate the activation ratio of the implanted S atoms, the net donor concentration was measured by $C$– $V$ measurement (⁠ $N d - C V$⁠), which is plotted in Fig. 2. Twice of $N S$ (2 $N S$⁠) is also shown in Fig. 2. $N d - C V$ is consistent with 2 $N S$ because a S atom works as a double donor,¹² which can be seen for $N S$ of $1× 10 17$ and $1× 10 18$ cm $− 3$⁠. In the case of the heavily-implanted layer (⁠ $N S$ = $1× 10 19$ cm $− 3$⁠), however, $N d - C V$ is about $4× 10 18$ cm $− 3$⁠, which is much lower than 2 $N S$⁠. Note that $C$– $V$ measurement could not be performed on the S $+$-implanted layer with $N S$ of $1× 10 20$ cm $− 3$ due to large leakage current.

Figure 3 depicts Arrhenius plots of the carrier concentration in S $+$-implanted SiC with $N S$ from $1× 10 17$ to $1× 10 20$ cm $− 3$ obtained by Hall effect measurement. The increase in carrier concentration with elevating the temperature is significant, which originates from the large $ΔE$ of S donors (340–520 meV).⁹ Here, the thicknesses of the S-doped layer with $N S$ of $1× 10 19$ and $1× 10 20$ cm $− 3$ were assumed to be 380 and 440 nm, respectively, since $N d - C V$ saturates at $4× 10 18$ cm $− 3$ and the “actual” thickness is defined as the distance from the surface to the depth at $N S$ reaching $2× 10 18$ cm $− 3$⁠. The carrier concentrations in the S $+$-implanted layer with $N S$ of $1× 10 19$ cm $− 3$ are almost the same as those in the S $+$-implanted layer with $N S$ of $1× 10 20$ cm $− 3$ above room temperature, suggesting that these S $+$-implanted layers hold the same donor concentration (⁠ $N D$⁠). Note that the carrier concentrations in the S $+$-implanted layer with $N S$ of $1× 10 18$ cm $− 3$ also look almost the same, but the net donor density is smaller than that of the S $+$-implanted layer with $N S$ of $1× 10 19$ and $1× 10 20$ cm $− 3$ as observed in the $C$– $V$ measurements (Fig. 2). Considering that $N d - C V$ is about $4× 10 18$ cm $− 3$ in the S $+$-implanted layer with $N S$ of $1× 10 19$ cm $− 3$⁠, it is suggested that the solubility or activation limit of S $+$-implanted SiC is about $2× 10 18$ cm $− 3$⁠. The obtained parameters by fitting the curves are summarized in Table I. Here, $N d i$ is the donor density in the $i$-site (⁠ $i=$ h or k), $N comp$ is the compensating acceptor density, $Δ E d i , 1$ is the shallower ionization energy of S donor in the $i$-site, and $Δ E d i , 2$ is the deeper ionization energy of S donor in the $i$-site. The detailed analysis of the temperature dependence of the free carrier concentration using the charge neutrality equation can be found elsewhere.⁹  $N comp$ obtained from the sample with $N S$ of $1× 10 19$ and $1× 10 20$ cm $− 3$ are smaller than those of the samples with lower $N S$⁠. Donor-type defects, which compensate residual acceptors, may be introduced by S $+$ implantation, while the specific mechanism is still under investigation.

Figure 4 illustrates $N S$ (or $N d$⁠) dependence of $μ e$ at room temperature in S $+$-implanted (or N-doped^13,14) SiC. $μ e$ in S-doped SiC is higher than that in N-doped SiC, which is remarkable especially at high $N S$⁠. $μ e$ in S $+$-implanted SiC with $N S$ of $1× 10 17$ cm $− 3$ is 695 cm $2$/V s, which is slightly higher than that of N-doped SiC 599 (cm $2$/V s). At $N S=1× 10 18$ cm $− 3$⁠, $μ e$ in S $+$-implanted SiC (598 cm $2$/V s) is more than twice as high as that in N-doped SiC (268 cm $2$/V s).

Figure 5 shows the temperature dependence of $μ e$ in S $+$-implanted SiC. $μ e$ in N-doped SiC¹⁵ is also shown for comparison. $μ e$ in S $+$-implanted SiC is higher than that in N-doped SiC within a wide temperature range. The relationship between $μ e$ and temperature (⁠ $T$⁠) is often given by $μ e∝ T − β$⁠.¹⁶ By fitting $μ e∝ T − β$ to experimental $μ e$ in S $+$-implanted SiC, $β$ was estimated as 2.56, 2.53, and 2.30 for $N S$ = $1× 10 17$⁠, $1× 10 18$⁠, and $1× 10 19$ (⁠ $1× 10 20$⁠) cm $− 3$⁠, respectively, which are slightly larger than those of N-doped SiC.

The S profiles measured by SIMS revealed that S atoms did not diffuse by activation annealing despite $N S$ exceeding $2× 10 18$ cm $− 3$ (potentially the solubility limit). To investigate crystalline quality and S-atom distribution in the S $+$-implanted layer, cross-sectional transmission electron microscope (XTEM) observation was performed. The XTEM image of the S $+$-implanted layer with $N S$ of $1× 10 20$ cm $− 3$ is depicted in Fig. 6(a). A diffraction pattern taken from the implanted layer revealed that the 4H structure was maintained and any phase transformation was not confirmed. However, many dark contrast regions are clearly observed throughout the S $+$-implanted layer, which are seen in the magnified image of the implanted layer [Fig. 6(b)]. Such dark contrast regions are also formed in heavily-implanted SiC layers and known to be mostly Frank-type stacking faults.¹⁷ Energy dispersive x-ray spectroscopy (EDX) spectra were obtained from the implanted layer along the scanning line shown in Fig. 6(b), which is shown in Fig. 6(c). A notable signal of S atoms is confirmed on the dark contrast region observed in the XTEM image, suggesting that S atoms precipitate near the stacking faults. In general, excess S atoms and stacking faults work as scattering centers for carriers and $μ e$ usually decreases with increasing $N S$⁠. On the other hand, $μ e$ of the S $+$-implanted layer with $1× 10 20$ cm $− 3$ is almost the same as that of the S $+$-implanted layer with $1× 10 19$ cm $− 3$ (Fig. 4), indicating that such excess S atoms and stacking faults do not strongly contribute to electron scattering.

The $β$ values obtained from the temperature dependence of mobility reflect a dominant carrier scattering mechanism. Figure 7 shows $N S$ or $N d$ dependence of $β$ in S-doped SiC determined in this study and the one calculated in N-doped SiC.¹⁵  $β$ in S-doped SiC is not lowered even at high $N d$ while $β$ in N-doped SiC rapidly decreases with increasing $N d$⁠. This indicates that impurity scattering is much less significant in S-doped SiC. Taking into account that the ionization ratio of S donors is much lower than that of N donors at a given temperature due to the large $ΔE$ (⁠ $>340$ meV) of S donors, the impact of ionized-impurity scattering on $μ e$ in S-doped SiC must be smaller than that in N-doped SiC.

The experimental $μ e$ obtained from the SiC with $N d$ of $1× 10 17$ cm $− 3$ [Fig. 8(a)] agrees well with the calculated $μ e$ in the wide temperature range. As the $β$ value implies the dominant scattering mechanism (Fig. 7), the ionized-impurity scattering has a small impact on the total mobility whereas the $μ e$ in N-doped SiC with $N S$ of $1× 10 17$ cm $− 3$ is limited by the ionized-impurity scattering especially at around room temperature.¹⁵ In the case of $N S$ of $1× 10 18$ cm $− 3$ [Fig. 8(b)], on the other hand, neutral-impurity scattering works as the dominant scattering process especially in the temperature range below 300 K. Based on the helium atom model, the mobility determined by neutral-impurity scattering is higher than that based on the hydrogen atom model.¹⁹ The calculated mobility by adopting the hydrogen atom model agrees with the experimental results of the SiC with $N S$ of $1× 10 17$ cm $− 3$ especially in the low temperature range. On the other hand, the electron mobility of SiC with $N S$ of $1× 10 18$ cm $− 3$ is close to the mobility obtained by adopting the helium atom model. The ionization ratio of sulfur depends on the temperature and sulfur works like either a hydrogen atom (partial ionization) or helium atom (full ionization), which may cause the slight deviation. In addition to that, Hall scattering factor may not be unity because of the anisotropy of SiC electronic band structure. Nevertheless, it is surprising that the experimental $μ e$ in the S $+$-implanted SiC with $N S$ of $1× 10 18$ cm $− 3$ is higher than the calculated $μ e$ and the reason for the unexpectedly high $μ e$ is still under investigation. Note that, even with the high $μ e$⁠, the resistivity of the S-doped layer is much higher than that of the N-doped layer with the same thickness because of the significant difference of electron concentration at a given temperature. However, as is discussed in introduction, such S-doped SiC is useful for fabricating n-channel MOSFET exhibiting high mobility or complementary JFET circuits with stable logic threshold voltage in the wide temperature range.

The authors performed Hall effect measurements on S $+$-implanted SiC layers with the doping concentration of $1× 10 17$– $1× 10 20$ cm $− 3$⁠. $C$– $V$ measurement and temperature dependence of carrier concentration indicated that the solubility or activation limit of S $+$-implanted SiC is about $2× 10 18$ cm $− 3$⁠. High $μ e$ of 598 cm $2$/V s was obtained from the S $+$-implanted layer with $N S$ of $1× 10 18$ cm $− 3$⁠, which is appropriately twice of that of N-doped SiC with the same doping concentration. Small ionized-impurity scattering is a possible origin for the high $μ e$ in the S $+$-implanted layer. The $μ e$ calculated with the helium atom model for neutral-impurity scattering fairly reproduced the experimental $μ e$ in the wide temperature range.

This work was supported by a research grant from the Samco Foundation.

The authors have no conflicts to disclose.

Mitsuaki Kaneko: Conceptualization (lead); Formal analysis (supporting); Funding acquisition (supporting); Investigation (supporting); Methodology (supporting); Project administration (lead); Supervision (lead); Validation (lead); Visualization (supporting); Writing – original draft (lead); Writing – review & editing (lead). Taiga Matsuoka: Data curation (lead); Formal analysis (lead); Investigation (lead); Validation (lead); Visualization (lead). Tsunenobu Kimoto: Conceptualization (lead); Funding acquisition (lead); Methodology (lead); Project administration (lead); Resources (lead); Supervision (lead); Writing – review & editing (lead).

The data that support the findings of this study are available from the corresponding author upon reasonable request.