1. Introduction

Gallium nitride (GaN) is one of the most promising materials for the fabrication of high-sensitivity visible-blind ultraviolet (UV) photodetectors (PDs) [1,2]. The wide band gap property of GaN also makes it a potential candidate for fabricating UV PDs for application in extreme conditions. In the past decade, various types of GaN-based PDs have been proposed [1,2]. Compared with bipolar junction PDs, the layer structure of Schottky barrier (SB) PD devices is much easier to fabricate. However, leakage current in SB PDs is higher because of the large thermionic emission current in SB PDs, as compared to the diffusion current in bipolar PDs for a given built-in voltage. To reduce the leakage current, and consequently the detectivity, in GaN-based SB PDs, a high Schottky barrier height at the metal/semiconductor interface must be achieved [3–5]. In addition to the work function of contact metals, the leakage current of SB PDs also depends strongly on the properties of the topmost semiconductor layer. Previous studies have shown that the gate leakage current in GaN-based SB diodes can be significantly reduced by utilizing a low-temperature-grown GaN layer inserted between the metal contact and high-temperature-grown GaN layer [5,6]. Dielectric films, including SiO₂, Si₃N₄, Ta₂O₅, Ga₂O₃, and Gd₂O_3, on GaN have been adopted to fabricate GaN-based metal–insulator–semiconductor (MIS) diodes [7]. The gate dielectric layers in GaN-based MIS structures can be prepared by physical vapor deposition, thermal oxidation, or chemical vapor deposition. In the present study, the gate dielectric oxide of gallium oxide (GaO_x) was grown on n-GaN epitaxial layers by alternating current (AC) bias-assisted photoelectrochemical (PEC) oxidation of n-GaN in H₂O. This approach for the formation of GaO_x layers differs from the conventional PEC method, which has a low growth rate and uses diluted phosphorus acid (H₃PO₄) and alkaline solutions [8]. The AC bias-assisted PEC oxidation features a low-temperature process and a high growth rate to form the GaO_x layer on GaN epitaxial layers. In the current paper, the optical and electrical properties of the fabricated GaN-based MIS PDs with GaO_x gate layer formed by AC bias-assisted PEC oxidation are reported.

2. Experiments

The GaN wafers used in this study were grown on c-face (0001) sapphire (Al₂O₃) substrates in a vertical low-pressure organometallic vapor phase epitaxy (OMVPE) reactor (Emcore D180). The layered structure of the GaN epitaxial wafers comprised a 30 nm thick GaN nucleation layer grown at 550 °C, a 2.3 μm thick Si-doped n⁺-GaN (n = 6 × 10¹⁸cm⁻³) layer, and a 1 μm thick lightly doped n-GaN (n = 3 × 10¹⁷cm⁻³) grown at 1050 °C. An Hg lamp with a power intensity of 350 mW/cm² was used as the UV light source in the PEC oxidation experiments. Before oxidation, deionized H₂O was first sprayed onto the surface of the mold. The front side of n-GaN sample was then adhered to the front side of the mold by gradually increasing the imprint pressure to 9 kg/cm² in order to fill the mold pattern entirely with H₂O. For the AC bias-assisted process, each bias cycle comprised a positive voltage pulse followed by a negative voltage pulse at a frequency of 2 Hz and a duty cycle of 90%. The growth rate of the PEC GaO_x layers was 13.5 nm/min when the AC bias was 5 V. The GaO_x layers were selectively grown on the n-GaN samples as the insulator layers for fabricating GaN MIS PDs. A semitransparent Ni/Au (5/10 nm) contact layer was deposited on the GaO_x layer to serve as the gate metal, while Cr/Au (50/200 nm) contact metal was subsequently deposited onto the exposed n⁺-GaN layer and gate metal to serve as the Ohmic contact and electrode pad, respectively [9,10]. In this study, GaN MIS diodes with 70 nm and 130 nm thick GaO_x gate layers were labeled PD-1 and PD-2, respectively. Figure 1(a) and (b) show the schematic structure of the fabricated devices and a representative photograph of the fabricated devices, respectively. Devices without the GaO_x gate layer were also prepared (labeled PD-3) for comparison. The room temperature current-voltage (I-V) characteristics of the fabricated PDs were measured by an HP4156 semiconductor parameter analyzer. Spectral responsivity of these GaN MIS PDs was measured using a Xe arc lamp and a calibrated monochromator as the light source.

 

Fig. 1 (a)schematic device structure (b)representative photograph of the fabricated circular photodetectors (c) schematic band diagram for depicting the mechanism of defect-assisted tunneling around the metal/GaO_x/n-GaN interfaces.

Download Full Size | PDF

3. Results and discussions

Figure 2 shows the typical reverse I-V curves taken in dark and under illumination (λ = 350 nm) from the proposed PDs. The dark current of PD-3 increases rapidly as the reverse bias increases, and is much larger than the PD-1 and PD-2. The observed dark current in the PD-1 is near constant when the reverse bias does not exceed 3 V. When the reverse bias exceeds 4 V, the dark current increases rapidly with an increase of bias, behaving similarly to PD-3. However, as the thickness of GaO_x gate layer increases to 130 nm (i.e., PD-2), the dark current shows only a slight increase, as low as ~10⁻¹¹ A, even when the bias reaches 10 V. This could be attributed to the fact that the resistive GaO_x gate layers between Ni and GaN could result in a thicker and higher potential barrier, as compared to conventional Schottky barrier PDs (i.e., PD-3). In other words, the insertion layer of GaO_x is able to effectively suppress the dark current, indicating that the PEC-formed gallium oxides possess a reasonably good insulating property [7]. Under illumination, the relative higher photocurrent-to-dark-current ratio (R_photo/dark) is observable in both PD-1 and PD-2. For example, the contrast ratios are around 2 × 10⁴ and 9 × 10⁶ for PD-1 and PD-2, respectively, when the PDs are biased at −3 V. At the same condition, the R_photo/dark obtained from the PD-3 is only 10. In addition to the relative low dark current, the photocurrents of PD-1 and PD-2 are significant higher than that of PD-3. The relative higher photocurrent may have been due to the fact that the GaO_x gate layers play the role of passivation layer to deaden the surface defects, thereby reducing the recombination events occurring at the metal/semiconductor interface during the light illumination. In other words, the passivated surface defects no longer recombine with the photogenerated carriers, leading to an enhancement of photocurrent in the PD-1 and PD-2 in comparison to those of PD-3. The passivation effect becomes more significant as the thickness of GaO_x gate layers increases. As revealed in extant literature, the insertion of a thin SiO₂ layer at the metal–GaN interface results in a significant decrease of the dark current density; however, the photocurrent is also reduced. This reduced photocurrent could be attributed to the fact that most photogenerated holes in reverse-bias condition are blocked by the SiO₂ layer with band gap of ~9 eV. Only portions of the holes are transported to the metal electrode via tunneling process through the defect states within the gap of SiO₂. In this study, a significant reduction of photocurrent, compared with the PD-3, was not observed from the PD-1 and PD-2. This result is very different from the conventional GaN-based MIS PDs with SiO₂ or SiN_x as insulators [11–13]. This difference could be attributed the fact that the band gap of GaO_x is significantly lower than that of SiO₂. A possible band gap of the GaO_x layer is approximately 4.9 eV at room temperature. As shown in Fig. 1(c), the relative lower bang gap may allow the photogenerated carriers transporting via the mechanism of thermionic-field emission associated with defect-related states within the band gap of GaO_x layer. The defects might be the oxygen and gallium vacancies, which play the donor- and acceptor-like states, respectively [14,15]. In this study, the dark current decreased with an increase of the thickness of GaO_x gate layer. However, the photocurrent would also decrease when the thickness of GaO_x gate layer exceeded 130 nm. In other words, the optimized thickness of the GaO_x layer for the application of GaN UV detector was around 130 nm.

 

Fig. 2 Typical reverse I-V curves taken in dark and under illumination from the experimental PDs. The incident light wavelength was 360 nm.

Download Full Size | PDF

Figure 3 shows the typical spectral responses of PD-1, PD-2, and PD-3. The cutoff was found to occur at around 365 nm (the absorption edge of GaN) for all the fabricated PDs. Under reverse bias and with incident light wavelength of 350 nm, the measured responsivities of PD-1 and PD-2 were found to be much higher than that of PD-3. This result is consistent with the photocurrents displayed in the Fig. 2. Such a dramatic difference of responsivity between the MIS type (i.e., PD-1 and PD-2) and the GaO_x-free PDs (i.e., PD-3) could be attributed to the fact that the PD-3 may have relatively more trap states at the Ni/ GaN interface for recombination with the photogenerated carriers. However, marked bias-dependent spectral responses were observed from the fabricated PDs. For UV illumination, electron-hole pairs were generated from the near band edge absorption. The origin of bias-dependent spectral responses in PD-3 could be due to the difference of transit times of holes and electrons through the trap states at the Ni/GaN interface. For PD-1 and PD-2, the origin of bias-dependent spectral responses could be due the accumulation of photogenerated holes at the GaO_x/GaN interface, as schematically shown in Fig. 1(c). This mechanism is similar to the origin of high optical gain observed from heterojunction bipolar transistors [16]. As shown in Fig. 2, the relative lower dark current of PD-1 and PD-2 also reflect the results of high UV-to-visible rejection ratio (R_UV/vis). As shown in Fig. 3, the values of R_UV/vis (350/450 nm) at a given reverse bias of 4 V are approximately 3 × 10⁴, 3 × 10⁵ and 40 for PD-1, PD-2, and PD-3, respectively. The high R_UV/vis of PD-2 were comparable with conventional GaN-based UV PDs with R_UV/vis of around 10³~10⁵ [1–6]. However, as revealed in extant literatures, the R_UV/vis values of GaN-based MIS UV PDs with insulating gate layer were well below 10⁴ [11–13,17]. A significant bias-dependent spectral response in the visible region was observed from the PD-3. This result can be attributed to the photoresponse caused from the defects at the Ni/GaN interface of PD-3. The defect-related trap states existing within the band gap of GaN could result in appreciable photoresponse. This phenomenon occurs even if the exciting photon energy is below band gap, especially when the reverse bias is increased because the Schottky barrier thickness decreases with an increase of reverse bias, thereby enhancing trap-assisted tunneling of photogenerated carriers. For PD-1 and PD-2, the surface defect-related states may be passivated by the GaO_x gate layer. As a result, below-band gap photon excitation did not generate pronounced carriers to contribute photocurrent. Clearly, the bias-dependent spectral response in visible region could be further suppressed as the thickness of GaO_x gate layer is increased to 130 nm.

 

Fig. 3 Spectral responses of (a)PD-1, (b)PD-2 and (c)PD-3 measured at different reverse biases.

Download Full Size | PDF

4. Conclusions

GaN-based MIS UV photodetectors with a GaO_x gate layer formed by alternating current bias-assisted photoelectrochemical oxidation of n-GaN in H₂O have been demonstrated in the current paper. The insertion of GaO_x gate layer between the gate metal (Ni/Au) and n-GaN has been proven to significantly reduce leakage current and lead to a significant enhancement of UV-to-visible rejection ratio (R_UV/vis). In addition, a marked bias-dependent spectral response in visible region could be effectively suppressed using the insertion layer; thus, the degradation of R_UV/vis at high reverse bias becomes insignificant.