Spectral response of a ZnMgBeSe Schottky photodiode at Ϫ 10 V bias. Theoretical curves corresponding to different quantum efficiencies ͑ ␩ ͒ 

Context in source publication

Context 1

... radiation detection has a number of applications, es- pecially in cases where the UV component of light should be detected in an infrared or visible background. Involved fields include environmental ͑ air quality monitoring, gas sensing, control of ozone layer thickness ͒ , medical ͑ personal UV ex- posure dosimetry ͒ , spatial ͑ research on solar and stellar spectra ͒ , and aeronautic domains ͑ material research, missile plume detection ͒ . 1,2 Classical semiconductors such as silicon and gallium arsenide can be used in UV detection, their long wavelength response being corrected with filters; however, they exhibit poor radiation hardness characteristics and lead to rather high dark currents. Wide-band gap materials are a much more attractive choice for selective UV detection. Main efforts are currently directed to III–V nitrides ͑ GaN, 3.39 eV; AlN, 6.2 eV ͒ , silicon carbide ͑ 3C-SiC, 2.39 eV; 6H-SiC, 2.86 eV ͒ , and diamond ͑ C, 5.5 eV ͒ . ZnSe, and more precisely ZnMgBeSe, are also promis- ing wide-band gap materials. Indeed, as shown in Fig. 1, ZnMgBeSe quaternary alloys can be grown lattice matched onto GaAs substrates with a direct band gap ranging from 2.75 to more than 3.8 eV. These compounds present high crystalline quality, 3 which should lead to high sensitivity and detectivity. Besides, the use of a GaAs substrate facilitates device integration with GaAs optoelectronics. So far, ZnMgBeSe p - i - n photodiodes have been fabricated with high responsivities of 0.22 A/W at 420 nm 4 and 0.17 A/W at 450 nm, 5 and ZnMgBeSe vertical Schottky photodiodes with a cutoff at 430 nm and a responsivity of 75 mA/W have also been published. 6 However, there is no report on visible-blind ZnMgBeSe devices. This is mainly due to the difficulty to achieve p -type doping in these materials. 7 However, this limitation does not prevent from developing Schottky barrier photodiodes, which, moreover, present some advantages on p – n diodes, such as a simpler fabrication process and a higher short-wavelength sensitivity. In this letter, we report on planar geometry Schottky photodiodes designed for UV detection. They are based on ϳ 1- ␮ m-thick nonintentionally doped ZnMgBeSe layers grown by molecular-beam epitaxy on semi-insulating ͑ 001 ͒ GaAs substrates. By x-ray diffraction measurements around the ͑ 004 ͒ reflection of GaAs, full widths at half maximum of 28 arc sec are obtained for these quaternary layers thus con- firming a high crystalline quality. Devices are processed in two steps. The Schottky contact consists of a semitransparent 60-Å-thick gold layer deposited by joule evaporation on top of the structure. The ohmic contact is made with molten indium on the layer. All structures have a total area of 5.5 mm 2 and an optical area ( A opt ), that is with the semitransparent metallic contact, of 4 mm 2 . These detectors are not antireflection coated. Spectral responsivity measurements are performed with a 150 W Xe arc lamp in a synchronous detection scheme. Calibration is made with a pyrometer. Figure 2 reports the spectral response of a typical detector measured at Ϫ 10 V bias, with front-side illumination through the semitransparent Schottky contact. The response is very flat above the band gap, due to the position of the depleted region on top of the structure: this is an advantage of Schottky barrier devices in comparison with p – n photodiodes. The cutoff at 380 nm is very sharp with a visible rejection rate of 5 ϫ 10 3 , which is a consequence of the high crystalline quality of the structure. A maximum responsivity of 0.17 A/W is obtained at 375 nm corresponding to a quantum efficiency of 54%. This is close to the theoretical limit, since we can roughly estimate that 25% of the incident light is lost by reflection on the surface and 10% by absorption in the semitransparent contact layer. Besides, it should be noted that this detector is well suited for UV–A detection with a quantum efficiency which remains over 50% over the whole UV–V range, that is from 380 to 315 nm. This maximum responsivity compares favorably with other results reported on II–VI Schottky detectors based on ZnS, ZnSSe, and ZnSTe layers grown on GaP substrates. ZnS- and ZnSSe-based structures have responsivities of 0.08 A/W at 335 nm and 0.09 A/W at 370 nm, respectively. 8 ZnSTe devices exhibit a high response of 0.13 A/W at 320 nm but the tellurium incorporation leads to a significant decrease of the rejection rate and generates a large cutoff in the spectral response. 9 ZnMgBeSe photodetectors also compare advantageously with 6H-SiC ͑ 0.15 A/W ͒ , 10 GaN ͑ 0.10–0.18 A/W ͒ , 11 and AlGaN ͑ 0.07 A/W ͒ . 12 Commer- cial GaN-based UV detectors exhibit a responsivity of 0.13 A/W at 365 nm. 13 The UV/visible contrast is also compa- rable with other materials: rejection rates in the 10 – 10 range are usually obtained in GaN-based devices. 14 The inset of Fig. 3 reports maximum responsivities for different bias. The response increases significantly up to 10 V as a consequence of the depleted region broadening. A 62% quantum efficiency is thus obtained at high bias ͑ over 30 V ͒ . Nevertheless, low bias operation ͑ 2 or 3 V ͒ is enough to get quantum efficiencies over 30%. Measurements at different optical excitation power have also been performed. Figure 3 shows the linearity obtained between the collected photocurrent and the optical power without any saturation over five decades. Time response is measured by using the third frequency of a Nd–yttrium–aluminum–garnet laser ͑ 355 nm ͒ with 10 ns gaussian pulses. Figure 4 reports the photocurrent decay measured at Ϫ 10 V bias with a 10 k ⍀ load resistance. This decay is not completely exponential, consisting of a major exponential component and a slow minor tail. The inset of Fig. 4 shows ␶ 90 → 10 values ͑ time from the signal to fall from 90% to 10% of its maximum value ͒ for several load resistances, R L . A linear variation ␶ 90 → 10 ϭ ␶ 0 ϩ R L C is obtained, indicating an RC -limited time response. Extrapolating to low load, a minimum time response ␶ 0 ϭ 1.5 ␮ s is obtained, which corresponds to a bandwidth BW ϭ 230 kHz. Noise characterization is performed with a SR350 lock-in amplifier, whose background noise power spectral density is 10 A /Hz. Figure 5 reports noise power spectral densities, S n , measured at Ϫ 2 and Ϫ 3V bias. In all cases, 1/ f noise is dominant at low frequencies. The fits indicated on Fig. 5 correspond to S n ϭ 2.5 ϫ 10 Ϫ 24 / f 1.2 ( V ϭ 2 V ͒ and S n ϭ 10 Ϫ 23 / f 1.2 ( V ϭ 3 V ͒ . The normalized detectivity D * is then obtained ...