Large-Scale β-Ga₂O₃ Trench MOS-Type Schottky Barrier Diodes with 1.02 Ideality Factor and 0.72 V Turn-On Voltage

Abstract

β-Ga₂O₃ Schottky barrier diodes (SBDs) suffer from the electric field crowding and barrier height lowering effect, resulting in a low breakdown voltage (BV) and high reverse leakage current. Here, we developed β-Ga₂O₃ trench MOS-type Schottky barrier diodes (TMSBDs) on β-Ga₂O₃ single-crystal substrates with halide vapor phase epitaxial layers based on ultraviolet lithography and dry etching. The $1/C^2-V$ plots are deflected at 2.24 V, which is caused by the complete depletion in the mesa region of the TMSBDs. A close-to-unity ideality factor of 1.02 and a low turn-on voltage of 0.72 V are obtained. This is due to the low interface trap density in the metal/semiconductor interface of TMSBDs, as confirmed by the current–voltage ($I–V$) hysteresis measurements. The specific on-resistance calculated with the actual Schottky contact area increases as the area ratio ($AR$) increases because of the current spreading phenomenon. Furthermore, the reverse leakage current of the TMSBDs is smaller and the BV is increased by 120 V compared with the regular SBD. This work paves the way for further improving the overall performance of β-Ga₂O₃ TMSBDs.

1. Introduction

Recently, the ultra-wide bandgap semiconductor β-Ga₂O₃ has attracted extensive attention in power device applications owing to its outstanding material performance [1]. β-Ga₂O₃ has a wide bandgap of 4.9 eV [2], a high critical electric field of 8 MV/cm [3], and its Baliga’s figure-of-merit (3214) is ~10 or 4 times larger than that of 4H-SiC or GaN [4]. Large-scale β-Ga₂O₃ wafers can be fabricated by low-cost melt growth methods such as floating zone, edge-defined film-fed growth, Czochralski, etc. [5,6,7]. High-quality thin films can be epitaxially grown on β-Ga₂O₃ substrates using various methods such as molecular beam epitaxy, metal–organic chemical vapor deposition and halide vapor phase epitaxy [8,9,10,11]. In addition, the chemical stability and thermal stability of β-Ga₂O₃ are excellent [12,13]. Therefore, power devices such as metal–oxide semiconductor field-effect transistors (MOSFETs) [14,15,16,17] and SBDs [18,19,20,21] fabricated on β-Ga₂O₃ wafers have shown strong competitiveness.

On the account of the electric field crowding at the Schottky contact edge and barrier height lowering effect [22], there is still a large gap on the figure-of-merit between β-Ga₂O₃ SBDs and the material limit. For this reason, devices designed with effective electric field management are crucial to power application. Many edge termination methods, such as field plates [18,21], ion implantation [20,23], thermally oxidized termination [24], and heterogeneous termination based p-type NiO [25,26], have been proposed to improve the reverse performance of the device. For field plates, the dielectric is unable to withstand the high-density electric field, resulting in a premature breakdown of the device. The ion implantation and thermally oxidized termination could only realize a weak p-type or high-resistance state, which does not achieve effective electric field management. P-type NiO makes up the gap of the unavailability of p-type doping for β-Ga₂O₃. However, it is necessary to further enhance the lattice mismatch of heterojunction interfaces and improve the stability of NiO materials. Additionally, these three types of methods have little inhibitory effect on the barrier height lowering effect. To address these issues, the trench MOS structure is introduced here [19,27,28,29,30,31]. Since the semiconductor is depleted at the Schottky contact, the trench MOS structure could eliminate the effect of electric field crowding. The peak of the electric field transfers from the Schottky contact edge to the corner of the trench. Moreover, the depletion region generated by the MOS capacitance on the sidewalls of the device effectively reduces the metal/semiconductor surface field and suppresses the barrier height lowering effect [29,32]. In the meanwhile, the reverse leakage current can be effectively reduced and BV can be further enhanced.

There has been a growing interest in β-Ga₂O₃ TMSBDs and significant progress has been made in this field. Sasaki et al. prepared β-Ga₂O₃ TMSBDs via dry etching and chemical–mechanical polishing, and obtained a BV of 240 V [19]. Li et al. utilized field plates and a trench MOS structure for improving the breakdown characteristics, achieved a power figure-of-merit of 0.95 GW/cm² [29]. Jian et al. have shown that TMSBDs exhibited superior performance and stability compared to regular SBDs under various temperatures [30]. Nevertheless, few studies have explored the forward characteristics of β-Ga₂O₃ TMSBDs, such as the ideality factor and turn-on voltage. β-Ga₂O₃ SBDs still face challenges of a relatively high ideality factor (>1.1) and turn-on voltage (>0.8 V) [19,29,30].

In this work, large-scale β-Ga₂O₃ TMSBDs have been demonstrated using ultraviolet lithography and dry etching. A turn-on voltage of 0.72 V and an ideality factor of 1.02 were achieved, which was attributed to the low interface trap density. Compared with regular SBDs, the BV of the TMSBDs was increased by 120 V and the reverse leakage current reduced by several orders of magnitude.

2. Materials and Methods

Figure 1 shows the cross-sectional schematic of the β-Ga₂O₃ TMSBDs. The (001) n-type β-Ga₂O₃ used in this experiment was purchased from Novel Crystal Technology, Japan. The thickness and doping concentration of the Sn-doped β-Ga₂O₃ substrate are 598 μm and ~$5.1\times {10}^{18} {\mathrm{c}\mathrm{m}}^{-3}$, respectively. The Si-doped β-Ga₂O₃ drift layer was grown on the β-Ga₂O₃ substrate via halide vapor phase epitaxy, and its thickness and doping concentration are 11 μm and $3.2\times {10}^{16} {\mathrm{c}\mathrm{m}}^{-3}$, respectively. The mesa widths are 20, 24, 28, and 32 μm, respectively, and the trench width is fixed at 40 μm. The mesa depth is 1.64 μm and the mesa length is 300 μm.

Figure 2 presents the process flow of the fabricated TMSBDs. First, the β-Ga₂O₃ wafers were cleaned with acetone, isopropanol, and deionized water in an ultrasonic environment to effectively remove surface impurities. Then, Ni hard mask patterns were formed by ultraviolet lithography, physical vapor deposition, and lift-off process. Next, a periodic structure composing of trenches and mesas was fabricated on a β-Ga₂O₃ wafer via inductive coupled plasma-reactive ion etching, and the etching gases were BCl₃ and Ar. Then, a Ni hard mask was removed using a solution of hydrochloric acid, and the bottom of the trench was smoothed to alleviate the concentration of the electric field [27,33]. The Ti (20 nm)/Au (100 nm) cathode electrode was deposited on the backside of the β-Ga₂O₃ wafer via electron beam evaporation. The ohmic contact was formed through rapid thermal annealing in N₂ ambient at 470 °C for 60 s. Next, a 50 nm Al₂O₃ film was deposited on the surface of the structure via atomic layer deposition. An overlay technique was used to open a hole at the top of the mesa and a dry etching method was used to remove the exposed Al₂O₃. Finally, the anode electrode of Ni (50 nm)/Au (100 nm) was deposited on the top of the structure. A regular SBD with a circular anode (a diameter of 300 μm) was fabricated for comparison. The capacitance–voltage ($C-V$), forward current density–voltage ($J-V$), and reverse $J-V$ characteristics were measured using a Keysight 4294A meter, a Keysight B1500A semiconductor analyzer, and a Keysight B1505A high-biased analyzer at room temperature, respectively.

Figure 3 shows the optical micrograph of the 24 μm β-Ga₂O₃ TMSBD and regular SBD. The area within dashed red lines is the anode area with a size of 320 μm × 640 μm. The area within dashed blue lines is the Schottky contact area of one mesa. The area ratio ($AR$) is defined as the ratio of the actual Schottky contact area to the anode area. The $AR$s corresponding to TMSBDs with a W_mesa of 20, 24, 28, and 32 μm are 15.8%, 21.1%, 26.4%, and 31.6%, respectively. The TMSBDs fabricated in the [010] orientation had the lowest interface trap density compared to those fabricated in the [100] orientation [28]. Therefore, the trench structures of all devices demonstrated here have the [010] orientation.

3. Results and Discussion

Figure 4a,b shows the $C-V$ and $1/C^2-V$ characteristics for the regular SBD and TMSBDs measured at 1 MHz. It is seen that the $1/C^2-V$ plot of the regular SBD is a standard straight line, while the $1/C^2-V$ plots of TMSBDs are deflected at a certain voltage. We linearly fitted the $1/C^2-V$ plots before and after deflection of the TMSBDs, and calculated the voltage at the intersection of the fitted plots as 2.24 V. According to the designed device structure, we give the following explanation. Before the deflection point, the capacitances of the TMSBDs are mainly composed of the Schottky junction capacitance and the MOS capacitances at the bottom of the trench, the side wall of the mesa, and the top of the mesa. As the voltage increases, the depletion region in the mesa keeps expanding. At the deflection point, the mesa is just completely depleted. After the deflection point, the depletion region extends to the semiconductor region below the mesa, and only the MOS capacitances at the bottom and top of the trench are preserved [19]. As a result, the falling rate of capacitance decreases with the decrease in voltage and the $1/C^2-V$ plots are deflected.

The capacitance of an SBD is given as follows [34]:

$$C={\left[\frac{q{\epsilon }_0{\epsilon }_n{N}_D}{2(V_{bi}-V_r)}\right]}^{1/2}$$

(1)

where $q$ is the electric charge, $N_D$ is the donor concentration, $V_{bi}$ is the built-in potential, $V_r$ is the applied voltage, ${\epsilon }_0$ is the vacuum dielectric constant, and ${\epsilon }_n=10$ is the relative dielectric constant of β-Ga₂O₃ [35]. The $N_D$ was determined to be $1.49\times {10}^{-16} {\mathrm{c}\mathrm{m}}^{-3}$ based on the $1/C^2-V$ plot. The $V_{bi}$ was $0.95 \mathrm{V}$ from extrapolation to $1/C^2=0$. Taking an electron effective mass of β-Ga₂O₃ $m_n^{*}=0.342m_0$ ($m_0$ is the electron rest mass), the local conduction band density of states ($N_c$) of $5.02\times {10}^{-18} {\mathrm{c}\mathrm{m}}^{-3}$ and the energy difference between the conduction band edge and the Fermi energy ($E_C-E_F$) of $0.15 \mathrm{e}\mathrm{V}$ can be obtained [36]. The corresponding Schottky barrier height was $q{\varphi }_B=q{V}_{bi}+\left(E_C-E_F\right)=1.1 \mathrm{e}\mathrm{V}$.

The depletion layer width ($W$) of the Schottky contact at the deflection point voltage of $1/C^2-V$ plots was 1.63 μm using the following equation:

$$W={\left[\frac{2{\epsilon }_n(V_{bi}-V)}{q{N}_{\mathrm{D}}}\right]}^{1/2}$$

(2)

$W$ is 100 nm lower than the mesa depth, consistent with the theoretical analysis above. The depletion of the mesa region is attributed to the depletion of the MOS structure in the sidewall and Schottky contact structure in the top of the devices. Owing to the slower depletion rate of the MOS structure in the sidewall compared to that of Schottky contact structure in the top of the devices, the difference between $W$ and mesa depth will decrease with an increasing width-to-depth ratio of the trench.

Figure 5a,b represents the frequency dependences of $C-V$ and $1/C^2-V$ curves for a 28 μm TMSBD at room temperature. It has been observed that capacitance decreases with increasing frequency. This could be caused by interface traps resulting from material defects, lattice mismatches, and etching defects [37]. At a low frequency, interface trap charges will capture and emit electrons, contributing additional capacitance. The capacitance gradually decreases as the frequency increases from 1 kHz to 100 kHz, while it hardly changes after 800 kHz. This is due to the slow majority of traps near the conduction band edge which cannot follow the AC signal, therefore its capacitance is closer to the ideal value [35,38].

Figure 6a depicts the forward $J-V$ characteristics of the TMSBDs in linear scales. As the $AR$ increases, the forward current density of the TMSBDs increases and the differential specific on-resistance ($R_{on,sp}$) calculated with the anode area decreases, which is attributed to the restricted current conduction path [30]. On the one hand, different $ARs$ correspond to different current conduction paths. On the other hand, negative charges at the fin sidewalls will cause depletion of the sidewall, thereby reducing the current density of the TMSBDs. The turn-on voltage of the TMSBDs is about 0.72 V taking the voltage at the forward current of $100 \mathrm{m}\mathrm{A}/{\mathrm{c}\mathrm{m}}^2$. It has no obvious relationship with the $W_{mesa}$ or $AR$. The minimum $R_{on,sp}$ of $24.15 \mathrm{m}\mathsf{\Omega }\cdot {\mathrm{c}\mathrm{m}}^2$ is measured on a TMSBD with a $W_{mesa}$ of 32 μm, while the maximum $R_{on,sp}$ of $34.36 \mathrm{m}\mathsf{\Omega }\cdot {\mathrm{c}\mathrm{m}}^2$ is acquired on a TMSBD with a $W_{mesa}$ of 20 μm, inconsistent with the discussion above.

Figure 6b shows the forward $J-V$ characteristics of the regular SBD and 20–32 μm TMSBDs in semi-logarithmic scales. The subthreshold swing (SS) of 60.71 and 60.49 mV/dec and the ideality factors of 1.020 and 1.018 are extracted at TMSBDs and the regular SBD, respectively. Compared with the previous work, ideal factors of the TMSBDs in this research are competitive [19,29,30]. The ideality factors of both types of devices are close to unity, revealing that the carrier transport mechanism is dominated by thermionic emission (TE). The reverse saturation current density formula of the TE mechanism is $J_S=A^{*}{T}^2{e}^{-q{\varphi }_B/kT}$, where $A^{*}=41.1 \mathrm{A}\cdot \mathrm{c}{\mathrm{m}}^{-2}\cdot {\mathrm{K}}^{-2}$ is the effective Richardson’s constant obtained when electron effective mass is $m_n^{*}=0.342m_0$, T is Kelvin temperature, ${\varphi }_B$ is the Schottky barrier height, and k is the Boltzmann’s constant. The reverse saturation current densities of TMSBDs and the regular SBD are $5.92\times {10}^{-8}$ and $8.76\times {10}^{-8} \mathrm{A}/{\mathrm{c}\mathrm{m}}^2$, respectively, and the extracted Schottky barrier heights are 0.82 and 0.81 eV, respectively. The barrier height of the TMSBDs is not much different from that of the regular SBD, which could be owing to the large $W_{mesa}$ of the fabricated TMSBDs weakening the additional Schottky barrier introduced by the MOS junction of the sidewalls of the mesa. In addition, the barrier height of regular the SBD measured by $J-V$ measurement is 0.29 eV lower than that of the $C-V$ measurement, showing consistency with previously reported studies [1,27,36].

Figure 7 shows the $I-V$ characteristics of the 28 μm TMSBD and regular SBD under hysteresis measurements, which were obtained under the forward voltage sweeping from 0 to 5 V and then backward with a limiting maximum current of 100 mA. During the forward sweeping, the interface traps will capture electrons, which decreases the current growth rate and increases the turn-on voltage for the $I-V$ curve. However, during the backward sweeping, most of the electrons in the traps are too late to be released to capture new electrons, resulting in better $I-V$ characteristics of the device. Therefore, there is a hysteresis voltage ($V_h$) in the forward and backward $I-V$ curves [38]. Additionally, the $V_h$ will increase with the interface trap density. For the regular SBD, $I-V$ hysteresis curves overlap completely, indicating that there are few interface traps in the device and the $I-V$ characteristics tend to be ideal. For the TMSBDs, the $I-V$ hysteresis curves show a higher $V_h$, indicating that TMSBDs have a higher interface trap density compared with the regular SBD. It is seen that the $V_h$ here is only 40 mV, suggesting that the interface trap density of TMSBDs is very small. Due to the small interface trap density of the prepared TMSBDs, the ideality factor of the device is close to unity. In addition, the Schottky barrier height of β-Ga₂O₃ is primarily determined by the interface states rather than the actual metal [39]. Shallower trap levels increase the forward bias diode current and reduce the diode Schottky barrier height and turn-on voltage [40].

Figure 8a shows the statistics of extracted $R_{on,sp}$ for the TMSBDs. The $R_{on,sp}$ and $AR$ of trench devices can be modeled as follows [27]:

$$R_{on,sp}=R_w+\frac{R_{w/o}-R_w}{AR}$$

(3)

where $R_w$ is the $R_{on,sp}$ below the mesa region of TMSBDs, and $R_{w/o}$ is the $R_{on,sp}$ of the regular SBD with the same electrode shape and area as TMSBDs. $R_w$ and $R_{w/o}$ were determined to be 13.5077 and 16.77 $\mathrm{m}\mathsf{\Omega }\cdot {\mathrm{c}\mathrm{m}}^2$, respectively, from a fitting plot to the model (3) shown by the black dash line in Figure 8a. Additionally, the fitting degree of the plot $R^2$ is 0.9952, which is very close to 1, illustrating that the fitting result is reliable.

In practice, the electrons flow mainly through the mesa region, as shown in dashed blue lines in Figure 3a. The actual Schottky contact area is used to evaluate the current density as $J_{on}=I_{on}/(m·A_{sch})$, where $I_{on}$ is the forward current, $A_{sch}$ is the Schottky contact area of one mesa, and $m$ is the number of mesas in a device. The specific on-resistance $R_{on,sp}^{\prime }$ based on the actual Schottky contact area was calculated using $R_{on,sp}^{\prime }={\left(d{J}_{on}/d{V}_f\right)}^{-1}$, where $V_f$ is the applied voltage. The extracted $R_{on,sp}^{\prime }$ is statistically shown in Figure 8b. It has been predicted that the $R_{on,sp}^{\prime }$ increases with increasing $AR$, which results from the current spreading mechanism. As the $AR$ decreases, the contact area decreases, the current spreading weakens, and the current density increases, then the $R_{on,sp}^{\prime }$ decreases [28].

Figure 9a depicts the representative reverse $J-V$ characteristics of the regular SBD and TMSBDs in semi-logarithmic scales. Owing to the small Schottky barrier height and the unrestricted barrier height lowering effect, the regular SBD has a large reverse leakage current and a small BV. Compared with the regular SBD, the TMSBDs had lower reverse leakage currents. This illustrates that the introduction of the trench MOS capacitance weakens the influence of the barrier height lowering effect, reducing the reverse leakage current of the devices. On the other hand, it suppresses the electric field crowding and improves the BV of the devices. The maximum BV of the TMSBDs was 340 V, which is 120 V higher than that of the regular SBD. Figure 9b shows the BV statistics of the TMSBDs. Unlike small-scale trench SBDs in the past research [27], the correlation between the BV and $W_{mesa}$ of the large-scale TMSBDs has not been found. The electric field concentration is most prominent at the anode electrode edge [19]. It should be noted that introducing a thicker field-plate dielectric at the anode electrode edge can further enhance the BV.

4. Conclusions

In summary, we have designed and fabricated large-scale TMSBDs based on ultraviolet lithography and dry etching. The introduction of trench structures causes a deflection in the $1/C^2-V$ plots. At the deflection point, the theoretically calculated depletion layer depth is 100 nm smaller than the actual trench depth, indicating that the sidewalls of the trench play an assisted role in depletion. It has been found that the hysteretic voltage between the forward and backward $I-V$ curves is small. This reveals that TMSBDs have a low interface trap density, contributing to a lower ideality factor of 1.02 and a turn-on voltage of 0.72 V. The specific on-resistance calculated with the anode area is inversely proportional to the $AR$, which is attributed to the increase in the current conduction path. The specific on-resistance determined by the actual Schottky contact area rises proportionally with the $AR$ due to the current spreading phenomenon. In addition, compared with the regular SBD, the reverse performance of large-scale TMSBDs was greatly improved, which shows the important role of the trench MOS structure in suppressing the electric field crowding and barrier height lowering effect of β-Ga₂O₃ SBDs.