Nitride-based ultraviolet metal-semiconductor-metal photodetectors with a low-temperature GaN layer

Journal of Electronic Materials, May 2003 by Sheu, J K, Kao, C J, Lee, M L, Lai, W C, Et al

The GaN metal-semiconductor-metal (MSM) ultraviolet (UV) photodetectors with a low-temperature (LT)-GaN layer have been demonstrated. It was found that we could achieve a two orders of magnitude smaller, photodetector-dark current by introducing a LT-GaN layer, which could be attributed to the larger Schottky-barrier height between the Ni/Au metal contact and the LT-GaN layer. It was also found that photodetectors with the LT-GaN layer could provide a larger photocurrent to dark-current contrast ratio and a larger UV-to-visible rejection ratio. The maximum responsivity was found to be 3.3 A/W and 0.13 A/W when the photodetector with a LT-GaN layer was biased at 5 V and 1 V, respectively.

Key words: LT-GaN, dark current, MSM, UV photodetector

Gallium nitride (GaN) is one of the most promising materials for the fabrication of high-sensitive, visible-blind ultraviolet (UV) detectors because it has a large direct-bandgap energy (3.41 eV at room temperature) and a high saturation-electron drift velocity (310 cm/s).1 The superior radiation hardness and high-temperature resistance of GaN also makes it a suitable material for UV detectors working in extreme conditions. In the past few years, various types of GaN-based photodetectors have been proposed, such as p-n junction diode,2 p-i-n diode,3,4 p-[pi]-n diode,5 Schottky-barrier detector,6 and metal-semiconductor-metal (MSM) photodetector.7-9 Among these devices, the MSM photodetector has an ultra-low intrinsic capacitance, and its fabrication process is compatible with field-effect-transistor (FET)-based electronics. Thus, one can easily integrate GaN-MSM photodetectors with GaN FET-based electronics to realize GaN-based optoelectronic-integrated circuits. To achieve a high-performance MSM photodetector, it is necessary to reduce dark current, which originates from carrier leakage occurring at the metal/semiconductor interface. Previously, it has been shown that one can significantly reduce gate leakage in GaAs FETs by using a low-temperature (LT)-GaAs layer.10 A similar concept should also be applied to GaN-based devices. It is known that LT-GaN exhibits an ultrahigh resistivity, which cannot be reduced even with thermal annealing. Such a semi-insulating property makes LT GaN suitable to serve as the top layer of MSM photodetectors. By depositing a LT-GaN layer on top of an n-GaN layer, we should be able to achieve a high Schottky-barrier height at the metal/semiconductor interface, reduce leakage current, and thus, improve the performance of a GaN-based MSM photodetector.

The GaN samples used in this study were all epitaxially grown on c-face sapphire substrates by metal-organic chemical vapor deposition.11-29 Trimethylgallium (TMGa) and ammonia (NH^sub 3^) were used as the source materials of Ga and nitrogen, respectively. After annealing the sapphire substrate at 1,100[degrees]C in an H^sub 2^ ambient to remove surface contamination, a 30-nm-thick LT-GaN nucleation layer was deposited onto the sapphire substrate at 550[degrees]C. It should be noted that measured resistivity of the LT GaN was larger than 10^sub 9^ ohm/[white square], which is the detection limit of our instrument. The temperature was then raised to 1,050[degrees]C to grow a 2-[mu]m-thick, high-temperature (HT) GaN-epitaxial layer. Typical room-temperature carrier concentration of this HT GaN-epitaxial layer was 2 x 10^sup 16^/cm^sup 3^. A 30-nm-thick, LT-GaN top layer was then grown on top of the HT GaN at 550[degrees]C (sample A). For comparison, samples without the LT-GaN top layers were also prepared (sample B). The Ni (100 nm)/Au (100 nm) bilayer metal-contact electrodes were then deposited onto these samples by e-beam evaporation for the fabrication of GaN-MSM UV photodetectors. Figure 1 shows the schematic of sample A with a LT-GaN top layer. The device consists of two interdigitated-contact electrodes fabricated through standard lithography and etching. The fingers of the contact electrodes were 4-[mu]m wide and 100-[mu]m long with a spacing of 10 [mu]m.

Figure 2 shows the room-temperature current-voltage characteristics of samples A and B measured in dark and under illumination. It can be seen that both dark current and photocurrent increase rapidly with voltage when the applied voltage is small. However, these currents start to saturate at about 3 V. Such a result suggests that these currents were probably related to generation-recombination centers and were limited by carrier lifetime ([tau]).30 It should be noted that the measurement limit of our experimental setup was around 1 pA. Thus, currents smaller than 1 pA could not be detected. In other words, the dark current of sample A actually should be smaller than 1 pA when the applied bias was less than 1.5 V. Furthermore, it was found that the dark current measured from sample A (curve a) was about two orders of magnitude smaller than that observed from sample B (curve c). Such a result could be attributed to the large Schottky-barrier height for the Au/Ni/LT-GaN/n-GaN contacts. As a result, we could effectively reduce dark current of GaN-MSM photodetectors by employing the LT-GaN layer.31 We also found that the measured photocurrents were about the same for these two samples when the incident-light wavelength was [lambda] = 360 nm, as shown in curves e and f. Such a result also indicates that sample A has a higher photocurrent to dark-current contrast ratio and thus higher detection sensitivity in the UV region because the dark current of sample A was much smaller than that of sample B. When the incident-light wavelength was [lambda] = 400 nm, it was found that photocurrent and dark current were almost the same for sample B, as shown in curves c and d. This is due to the fact that almost no absorption will occur when the incident-photon energy is smaller than the bandgap energy of the crystalline HT-GaN layer. In contrast, photocurrent measured with the same [lambda] = 400-nm incident light was about two times larger than the dark current for sample A, as shown in curves a and b. This could be attributed to the absorption of the LT-GaN layer. It is known that LT GaN exhibits a microcrystalline nature.32 When the LT GaN is illuminated, subgap absorption, originating from deep-level states in the material, could also absorb incident light. Thus, we could still observe an enhanced photocurrent even with [lambda] = 400 nm.