Open Access
Add to BibSonomy
Abstract
In this study, a Ti–diamond–Ti structured ultraviolet photodetector was fabricated on a homoepitaxial diamond layer with an oxygen-terminated surface. The properties of the Ti/diamond schottky contact were measured using X-ray photoelectron spectroscopy, and the barrier height was found to be 1.15 eV. At a bias of 3 V, the responsivity at 210 nm was only 4.29 mA/W, while at 12 V, the responsivity increased rapidly to 51 mA/W. The increase can be ascribed to the photocurrent gain. With the further increase in voltage, an avalanche effect was produced, and the responsivity could reach 1.18 A/W at 50 V. Moreover, the transient response behavior of the photodetector exhibited a good repeatability and response speed.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
A good UV photodetector should satisfy the 5s requirements of high selectivity between UV light and visible light, high signal to noise ratio, high response speed, high stability and high sensitivity [1]. Among many wide bandgap semiconductors, diamond is a promising material in deep-ultraviolet photodetection to meet the 5s requirements owing to its excellent properties such as a wide band gap (5.5 eV), high carrier mobility (3800 cm2/Vs), high thermal conductivity (22 W/cm), low dielectric constant (~5.7), high chemical stability and high radiation hardness [1–4]. Studies on diamond UV photodetectors have mainly focused on film quality and device design [5–8]. Several device structures have been proposed, such as the PIN detector [9], Schottky photodiode (PD) [10], and metal–semiconductor–metal (MSM) detector [11]. Among these devices, the MSM structure has often been applied because of its simple fabrication process. MSM structures can be divided into two types: MSM photoconductor and MSM PD [12]. Compared to the MSM photoconductor, annealing is not required for the MSM PD, and the Schottky barriers formed within the metal/diamond interface do not exhibit high dark currents or noise levels [11]. Teraji et al. applied interdigitated Ti/Au electrodes to an undoped single-crystal diamond epitaxial layer to fabricate a photodetector, where the Ti on the undoped diamond was considered a Schottky contact [13]. In his work, the photocurrent increased linearly with the bias. However, Liao et al. also fabricated a metal–diamond–metal structured photodetector on an undoped diamond epitaxial layer, where tungsten carbide (WC) was used as the electrode, and two different I–V regions could be observed for the photocurrent [14]. At low bias, the I–V curves exhibited sublinear behavior, while at high bias, they were superlinear, leading to a photocurrent gain. The authors ascribed the origin of the photocurrent gain to the formation of interfacial traps between the WC and diamond. In fact, another MSM structure by Liao et al. with an identical electrode configuration did not display such a photocurrent gain [12]. Therefore, the mechanism of photocurrent gain should be further explored.
In this work, we attempted to form a MSM PD structure on interdigitated Ti/Au electrodes on an undoped diamond-oxygen surface. The results indicate that the photocurrent gain is dependent on the photogenerated electron density and the avalanche effect can be produced.
2. Experimental details
2.1. Fabrication of diamond photodetector
An unintentionally doped single-crystal diamond epitaxial layer was used to fabricate the photodetector. First, a high-pressure high-temperature IIa-type single-crystal diamond substrate was ultrasonically cleaned with acetone, alcohol, and deionized water in sequence. Then, an approximately 200-nm homoepitaxial layer was grown on the substrate using microwave plasma chemical vapor deposition. CH4 and H2 were used as reacting gases, and the total flow rate and CH4/H2 ratio were 500 sccm and 1%, respectively. The process pressure, growth temperature, and microwave power were 100 Torr, 900 °C, and 1 kW, respectively. After growth, the sample was boiled in an acid mixture (H2SO4:HNO3 = 1:1 by volume) to change the hydrogen terminations into oxygen terminations. Finally, 30/120-nm-thick Ti/Au interdigitated electrodes were patterned on the epitaxial layer through standard photolithography and radio-frequency magnetron sputtering. The width, length, and depth of the electrode were 10, 200, and 10 μm, respectively, leading to a total active area of 0.036 mm2.
2.2. Characterizations
The quality of the diamond substrate was evaluated by a Raman spectrometer with a 50 × objective lens and 532-nm wavelength laser. The electrical and photoresponse properties of the photodetector were evaluated by an Agilent B1505A power device analyzer, 1000-W Xe lamp source, and monochromator. The incident light power was measured by a commercial UV-enhanced Si detector. Time response behaviors were measured by repeatedly switching the UV light on and off with a metal shutter.
3. Results and discussion
A Raman spectrum of the epitaxial layer is presented in Fig. 1, where a sharp strong peak can be observed at about 1333 cm–1 with an FWHM of 3.96 cm–1. This value presents an acceptable single-crystal diamond quality. In addition to this, another wide weak peak can be observed at about 1428 cm–1, which corresponds to a wavelength of 575 nm in the photoluminescence spectrum, as shown in the inset. In fact, this peak is the fluorescence peak of the NV0 center, and another peak at about 637 nm in the inset is from the NV– center [15], which indicates that some nitrogen impurities were present in the epitaxial layer.
The XPS detection technique can be used to directly determine the barrier heights of the metal/diamond contact. The C1s peak of the diamond covered with interdigitated Ti/Au electrodes was measured and compared to the standard C1s spectrum of diamond. Additionally, a reference Au sample was prepared to calibrate the results. The barrier height (ΦBH) for the Ti/diamond contact can be calculated by the following equation [16]:
where EC1s is the C1s binding energies of the diamond sample covered with interdigitated Ti/Au electrodes, and ∆Ediamond is the energy separation between the valence band maximum and C1s for the undoped oxygen-terminated diamond sample, which has been obtained from another paper: 283.9 eV [17]. and are the Au 4f7/2 binding energies for the diamond sample and reference sample, respectively. The C1s spectrum shown in Fig. 2(a) was obtained from the diamond surface covered by interdigitated Ti/Au electrodes. The component related to the sp3C–C bonds is located at an energy of 283.95 eV, and that located at 283.1 eV is related to carbide impurities. Figure 2(b) presents Au 4f7/2 spectra for the diamond sample and reference sample, where the binding energies are located at 82.8 and 83.9 eV, respectively. From this, the barrier height can be calculated as ϕBH = 283.95 eV – 283.9 eV – (82.8 – 83.9) eV = 1.15 eV. This value is large when compared with the barrier height of 0.63 eV of the Ti ohmic contact on p-type diamond [16]. In fact, an ohmic contact cannot be expected just by Ti deposition on the oxygen-diamond surface [18]. Therefore, the Ti/diamond contact in our work is the Schottky contact. Accordingly, the photodetector is labeled MSM-PD.The I–V characteristics of the MSM-PD with and without 210-nm illumination are presented in Fig. 3(a). It can be observed that the dark current is extremely low, less than 1 pA even under a bias of 30 V. When 210-nm light is induced, the photocurrent is high. It should be noted that the photocurrent can be divided into three parts. When the bias is less than 5 V, in the range of 5–9 V, and more than 9 V, the current increases sublinearly, exponentially, and sublinearly again, respectively. As a result, the current in the high-voltage region is 10 times higher than that in the low-voltage region. This phenomenon means that a photocurrent gain has been achieved for the MSM-PD. The origin of the photocurrent gain can be ascribed to the electron injection at the interface [19]. Because of the symmetrical structure of the electrodes, the band diagram at the equilibrium state is symmetrical, as shown in Fig. 3(b). When a positive bias is applied, the band will bend upward from the anode to the cathode, as shown in Fig. 3(b). For the holes, only one barrier exists in the valence band at the anode, which can be easily collected by the cathodes. However, for electrons, two barriers exist in the conduction band, both at the anode and cathode, inhibiting their transportation. Under illumination with an energy above the band gap, electrons in the valence band will be excited to the conduction band, leaving holes in the valence band, as shown in Fig. 3(b). These photogenerated holes will be collected and contribute to the primary photocurrent at low bias while photogenerated electrons accumulate near the Ti/diamond interface at the anode. When the bias becomes large, the large electric field will enable these electrons to tunnel through the barrier, leading to a photocurrent gain. This mechanism is different from that of Ref. 13, thus, the I–V characteristics are also different.
Fig. 3 (a) I–V characteristics of the detector in dark and 210-nm illumination conditions. (b) Schematic of the band diagram with/without bias.
The I–V characteristics of the MSM-PD under different illumination light wavelengths are presented in Fig. 4(a). When the light wavelength is shorter than 240 nm, the exponential gain region can be clearly observed. In addition to this, a slightly exponential gain region can be observed at 400 and 450 nm. However, their minimum voltages for generating the tunnel effect (called open voltage) are quite different. Under illumination with an energy above the band gap, the observed open voltage is low, while at 400 and 450 nm, the open voltage is high. This difference can be ascribed to the electron density difference. Under 400 and 450-nm illumination, the photogenerated electrons primarily come from the nitrogen defects. In addition, because the defect density is low, relatively fewer electrons can be generated, causing a high open voltage. This viewpoint could be further verified using 210-nm light illumination with different light power densities, and the results are presented in Fig. 4(b). With the decrease in light power, the open voltage gradually increases. As the number of photogenerated electrons is proportional to the light power density, it can be concluded that more electrons will lead to lower open voltages. Therefore, the photocurrent gain is affected by the photogenerated electron density.
Fig. 4 I–V characteristics under (a) different wavelengths and (b) same wavelength but different light power densities.
Because of the existence of photocurrent gain, the responsivity and rejection ratio of the MSM-PD could be improved. The responsivity of the MSM-PD under 34.1 µW/cm2 and 210 nm illumination is presented in Fig. 5(a). At a bias of 3 V, the responsivity is only 4.29 mA/W. However, at a bias of 12 V, which is higher than the open voltage (whose value is 5 V, as shown in Fig. 4(b)), the responsivity increases rapidly to 51 mA/W. In contrast, the responsivity at 400 nm was evaluated, where the 210-nm/400-nm rejection ratio was calculated, and the results are presented in Fig. 5(a). At low bias, the 210-nm/400-nm rejection ratio is relatively small, while at high bias, the rejection ratio increases significantly. Figure 5(b) shows the responsivity depending on the light power at different bias voltages under 210-nm illumination. At a bias of 3 V, the responsivity is almost maintained. At a bias of 7 V, which is located in the exponential voltage gain region, the responsivity increases slowly with light power. At a bias of 12 V, the responsivity increases rapidly under the low light power condition and then increases slowly under high light power. These results indicate that photocurrent gain is helpful in enhancing the responsivity.
Fig. 5 Responsivity of MSM-PD under 210-nm illumination depending on (a) voltage and (b) light power.
The time response behavior of the MSM-PD was investigated by illuminating with discontinuous 210-nm light at 20 V, the results of which are presented in Fig. 6(a). A good repeatability is achieved, where the rise time is 640 ms and the fall time is 34 ms. Liao et al. found that the transient response at high voltage with a high gain becomes relatively slow [14]. This difference also indicates a different photocurrent gain mechanism. In their work, the photocurrent gain is derived from the charge trapping at the metal/diamond interface, leading to hole tunneling at a certain bias. In our work, electron tunneling contributes to the photocurrent gain. As a result, both the I–V characteristics and transient response behavior are different. According to the time response measurement, the noise-equivalent power was calculated to be 1.87 × 10−9 W at 210 nm, indicating a good performance.
Fig. 6 (a) Transient response of MSM-PD. (b) I–V characteristics of the MSM-PD in dark and 210-nm illumination conditions. The dark current is the realistic data, not the absolute value. The inset shows IV characteristics at some other wavelengths.
Apart from the tunneling effect, the avalanche effect is realized for the MSM-PD at high voltage for the first time. With the further increase in the applied bias, the photocurrent will experience a second rapid rise process, as shown in Fig. 6(b). It can be observed that when the applied bias is larger than 43.5 V, the current increases rapidly from 1 to 10 nA. This phenomenon can be ascribed to the avalanche effect. Because of the energy band bending shown in Fig. 3(b), photogenerated electrons in the diamond bulk will move to the anode under the electric field. When the applied bias is high, photogenerated electrons will obtain sufficient energy to induce extra electron ionization. As a result, the avalanche effect is produced, by which the responsivity reaches 1.18 A/W at 50 V. In Bin Zhao et al’s work, the avalanche effect enables the UV photodetector have a high responsivity, high detectivity and high response speed [20]. Therefore, the observed avalanche effect in diamond UV photodetector is promising to enhance the device performance with structure optimized. Other than 210 nm, avalanche effect can also be observed at 210 nm and 215 nm, as shown in the inset in Fig. 6(b). The critical voltage for 210 nm where avalanche effect appears is the smallest. This is in accordance with the tunneling phenomenon, further confirming that photogenerated electron density is a key factor to produce photocurrent gain. The reverse photocurrent–voltage curve is also displayed in Fig. 6(b), in which no avalanche effect is observed. This phenomenon makes the origin of the avalanche effect more complex. It should be noted that the dark currents at negative bias and positive bias are asymmetrical. Thus, the two Schottky barriers may not be symmetrical, leading to the difference in avalanche effect.
4. Conclusions
In this study, because of the existence of photocurrent gain in a diamond ultraviolet photodetector, a responsivity enhancement was achieved. When the bias was larger than the open voltage, photogenerated electrons accumulated near the positively biased interface could tunnel through the barrier, contributing to the photocurrent gain. Further, it was found that the open voltage is related to the number of electrons. When the bias was lower than the open voltage, the responsivity and rejection ratio were small. When the bias was larger than the open voltage, the responsivity and rejection ratio increased. When the voltage increased further, the avalanche effect was produced, leading to a responsivity as high as 1.18 A/W at 50 V under 210-nm illumination. This was the first time the avalanche effect was observed in a diamond photodetector. Moreover, the transient response exhibited a good repeatability and response speed.
Funding
National Natural Science Foundation of China (NSFC) (61627812, 61605155); Technology Coordinate and Innovative Engineering Program of Shaanxi (2016KTZDGY02-03); Postdoctoral Science Foundation of China (PSFC) (2015M580850).
References and links
1. M. Liao, L. Sang, T. Teraji, M. Imura, J. Alvarez, and Y. Koide, “Comprehensive Investigation of Single Crystal Diamond Deep-Ultraviolet Detectors,” Jpn. J. Appl. Phys. 51(9R), 090115 (2012). [CrossRef]
2. S. Salvatori, M. C. Rossi, F. Scotti, G. Conte, F. Galluzzi, and V. Ralchenko, “High-temperature performances of diamond-based UV-photodetectors,” Diamond Related Materials 9(3-6), 982–986 (2000). [CrossRef]
3. R. D. McKeag and R. B. Jackman, “Diamond UV photodetectors: sensitivity and speed for visible blind applications,” Diamond Related Materials 7(2-5), 513–518 (1998). [CrossRef]
4. E. Pace, F. Galluzzi, M. C. Rossi, S. Salvatori, M. Marinelli, and P. Paroli, “Electra-optical properties of diamond thin films as UV photodetectors,” Nucl. Instrum. Methods Phys. Res. A 387(1-2), 255–258 (1997). [CrossRef]
5. S. Salvatori, E. Pace, M. C. Rossi, and F. Galluzzi, “Photoelectrical characteristics of diamond UV detectors: dependence on device design and film quality,” Diamond Related Materials 6(2-4), 361–366 (1997). [CrossRef]
6. Z. Liu, F. Li, W. Wang, J. Zhang, F. Lin, and H.-X. Wang, “Effect of depth of Buried-In Tungsten Electrodes on Single Crystal Diamond Photodetector,” MRS Adv. 1(16), 1099–1104 (2016). [CrossRef]
7. T.-F. Zhu, Z. Liu, Z. Liu, F. Li, M. Zhang, W. Wang, F. Wen, J. Wang, R. Bu, J. Zhang, and H. X. Wang, “Fabrication of monolithic diamond photodetector with microlenses,” Opt. Express 25(25), 31586–31594 (2017). [CrossRef] [PubMed]
8. Z. Liu, J.-P. Ao, F. Li, W. Wang, J. Wang, J. Zhang, and H.-X. Wang, “Fabrication of three dimensional diamond ultraviolet photodetector through down-top method,” Appl. Phys. Lett. 109(15), 153507 (2016). [CrossRef]
9. A. BenMoussa, U. Schühle, K. Haenen, M. Nesládek, S. Koizumi, and J.-F. Hochedez, “PIN diamond detector development for LYRA,” Phys. Status Solidi 201(11), 2536–2541 (2004). [CrossRef]
10. M. D. Whitfield, S. S. M. Chan, and R. B. Jackman, “Thin film diamond photodiode for ultraviolet light detection,” Appl. Phys. Lett. 68(3), 290–292 (1996). [CrossRef]
11. H. J. Looi, M. D. Whitfield, and R. B. Jackmana, “Metal–semiconductor–metal photodiodes fabricated from thin-film diamond,” Appl. Phys. Lett. 74, 3332 (1999). [CrossRef]
12. M. Liao and Y. Koide, “High-performance metal-semiconductor-metal deep-ultraviolet photodetectors based on homoepitaxial diamond thin film,” Appl. Phys. Lett. 89(11), 113509 (2006). [CrossRef]
13. T. Teraji, S. Yoshizaki, H. Wada, M. Hamada, and T. Ito, “Highly sensitive UV photodetectors fabricated using high-quality single-crystalline CVD diamond films,” Diamond Related Materials 13(4-8), 858–862 (2004). [CrossRef]
14. M. Liao, X. Wang, T. Teraji, S. Koizumi, and Y. Koide, “Light intensity dependence of photocurrent gain in single-crystal diamond detectors,” Phys. Rev. B 81(3), 033304 (2010). [CrossRef]
15. M. W. Doherty, N. B. Manson, P. Delaney, F. Jelezko, J. Wrachtrup, and L. C. L. Hollenberg, “The nitrogen-vacancy colour centre in diamond,” Phys. Rep. 528(1), 1–45 (2013). [CrossRef]
16. S. Kono, T. Teraji, H. Kodama, K. Ichikawa, S. Ohnishi, and A. Sawabe, “Direct determination of the barrier height of Ti-based ohmic contact on p-type diamond (001),” Diamond Related Materials 60, 117–122 (2015). [CrossRef]
17. F. Maier, J. Ristein, and L. Ley, “Electron affinity of plasma-hydrogenated and chemically oxidized diamond (100) surfaces,” Phys. Rev. B 64(16), 165411 (2001). [CrossRef]
18. Y. Chen, M. Ogura, S. Yamasaki, and H. Okushi, “Ohmic contacts on p-type homoepitaxial diamond and their thermal stability,” Semicond. Sci. Technol. 20(8), 860–863 (2005). [CrossRef]
19. T. Sugeta and T. Urisu, “High-Gain Metal-Semiconductor-Metal Photodetectors for High-speed Optoelectronic Circuits,” IEEE Trans. Electron Dev. 26(11), 1855–1856 (1979). [CrossRef]
20. B. Zhao, F. Wang, H. Chen, Y. Wang, M. Jiang, X. Fang, and D. Zhao, “Solar-Blind Avalanche Photodetector Based On Single ZnO-Ga2O3 Core-Shell Microwire,” Nano Lett. 15(6), 3988–3993 (2015). [CrossRef] [PubMed]
