1. Introduction

Along with the rise of graphene as a conductive two-dimensional (2D) carbon nanosheet, research in wide band gap semiconductor boron nitride nanosheets have received increasing attention recently. Excellent properties of BNNSs make them useful for applications in harsh environments. The 2D BNNS was first synthesized in 2004 [1]. The synthesizing techniques reported include plasma sputtering [2], chemical vapor deposition (CVD) [3–7], chemical exfoliation [8,9], chemical blowing [10], ball milling [11], and liquid exfoliation of a bulk hexagonal boron nitride (h-BN) [12]. Excellent results have been reported with the CVD method, although the growth rate was low, and the required high processing temperatures up to 1000 °C would vaporize the impurities and result in internal thermal stresses between the deposited BNNS and the substrate.

As expected, the quality and thickness of the synthesized BNNS material play an important role in its electrical and optical properties. Short-pulse laser plasma deposition (LPD) has been shown to be a promising cost effective technique for the synthesis of high purity 2D BNNSs. The advantages offered include a relatively low process temperature that accommodates different types of substrates with reduced thermal stresses; a high growth rate; and a digitized thickness control down to atomically thin layers. In addition, the built-in magnetic filter removes any impurities during synthesis process [13]. Recently, the nanoscale morphology and the structure evolution of BNNS synthesized with pulsed LPD have been characterized [14,15].

For ultraviolet (UV) applications ranging from biological and chemical sensing to space astronomy, boron nitride (BN), as a wide band gap material, is an attractive alternative to semiconductor based gallium nitride (GaN), silicon carbide (SiC) and diamond UV detectors, due to its excellent chemical and thermal properties. Different photonic devices based on BN bulk and film materials have been reported, including DUV emitters [16–18] and photodetectors [19], neutron detectors [20], biological sensors [21] and diode lasers [22]. It also provides an excellent material platform for integrated nanophotonic devices [23–25].

Recently, 2D BNNS based DUV photodetectors have emerged as hot research topics. One of the reasons for using super thin, 2D BNNS materials is to avoid possible heat accumulation and to achieve high-performance operation at high temperature. How to obtain a high purity and high crystallinity of the BNNSs is critical for DUV applications. Most of previous works focus on the lateral structure DUV photodetectors [26,27] where the same metal is employed as the anode and cathode on the top surface. The small gap between the alternating metal contacts formed by a densely placed interdigitated finger structure establishes highly localized electric field lines. The carriers generated in the BNNS active layer away from the surface are difficult to diffuse into the high field region before collection.

In recent years, major efforts have been conducted successfully towards developing BNNS energy bandgap width modulation techniques. As a result, BNNSs can now be synthesized with a wide range of energy bandgap widths as low as 2 eV or less [14,15,32,33]. In this paper, we report on a new approach to fabricate the vertical structure MSM DUV detector based on the bandgap engineered 2D h-BNNS deposited onto metal molybdenum (Mo) substrate. The detailed characterization of the BNNS is given first, followed by the design, fabrication and testing of the BNNSs based photodetector.

2. Short pulse CO₂ laser plasma deposition of boron nitride nanosheets

The LPD system used to synthesize 2D BNNS samples consists of a CO₂ pulse laser, a beam delivery system and a vacuum chamber [28,34] whose schematic is shown in Fig. 1. The 10.6 μm CO₂ laser, after passing through a 30 cm focal length ZnSe lens, was focused onto the target at a 45° angle with respect to its surface normal. The pyrolytic h-BN target of 2” dia. × 0.125” thick, a purity of 99.99%, a density of 1.94 g/cm³, was mounted on a stage of six-axis adjustment (x, y z, tip, tilt and rotating at 200 rpm along its surface normal). The focused laser beam, 5 J per pulse at 5 Hz repetition rate, creates a spot size of ~2 mm in diameter on the target, corresponding to a laser energy density of ~160 J/cm².

 

Fig. 1 Schematic of the CO₂ pulsed laser plasma deposition system.

Download Full Size | PDF

Due to its chemical and thermal stability, molybdenum (Mo) was chosen as the substrate. Each substrate of 1.0 cm dia. × 0.3 mm thick was polished by diamond nanoparticles and cleaned properly. During deposition the substrate temperature was maintained at 400 ± 5°C. The LPD has been studied by a number of numerical models [29]. The ablation rate can be calculated using the kinetic model described in Keidar et al [30]. Because laser generated plasma deposition produces both the ionic deposition and molecular deposition, the distance between the target material and substrate is one of the key parameters in the deposition process. In this experiment, the distance between the target and substrate was set to 4 cm, and the deposition was completed in about 15 minutes, corresponding to ~4500 laser pulses.

3. Characterization of the BNNS sample deposited onto molybdenum substrate

The nanoscale morphologies of the synthesized BNNSs/Mo samples have been characterized using scanning electron microscopy (SEM) and transmission electron microscopy (TEM). As shown in Fig. 2(a), the sample consists of a large amount of overlapped nanosheets covering the whole surface of the substrate. Each continuous single crystal, few-atomic-layer-thick 2D BNNS has an average area of a few μm². As indicated by the SEM image of the sample cross section view in Fig. 2(b), the thickness of the synthesized BNNSs is about 1.5 μm. Hence the obtained sample, referred to as BNNSs, is made of a large number of randomly orientated single crystal BNNS [27]. When the laser generated plasma plume approached the substrate surface, extra thermal activation in addition to the substrate heating was provided to the sample by energy exchange through collisions between the active ions, clusters and atoms of B and N. This process helped to grow single crystal BNNS with limited layers that coalesced into a thin film and transformed into larger patches of the BNNSs.

 

Fig. 2 Images of the BNNSs deposited on Mo substrates, (a) low magnification top-view SEM with overlapping BNNSs; (b) cross section view SEM of ~1.5 μm thick sample on Mo substrate; (c) typical TEM with visible curved/wrinkle and folding structures; and (d) high resolution TEM image of the edge area of a BNNS whose interlayer spacing is ~0.33 nm.

Download Full Size | PDF

In order to understand and study crystalline properties and morphologies, the sample was simply scratched off from the substarte and then transferred to the grids for TEM measurements. Figure 2(c) shows the typical TEM image of the sample that appears highly transparent with a clear edge, indicating a good crystalline structure of each BNNS. Small curved/wrinkle and folding structures are also observed. High resolution TEM was also used to image the edge area of BNNS. As shown in Fig. 2(d), the interlayer spacing is estimated around 0.33 nm, and the thickness of the single crystal BNNS is limited to 20-30 layers (less than 10 nm). Thicker BNNSs can be synthesized with an extended deposition time. However, the growth through nucleation eventually forms an opaque BNNS sample with random edge and poor crystalline structures, once beyond a critical thickness.

The BNNSs/Mo sample has also been analyzed by Raman spectroscopy using triple monochromator with 514 nm Ar⁺ laser as the excitation wavelength [34]. The laser beam was focused onto the sample surface. In Fig. 3(a), a clear Raman active E_2g mode at approximately 1365 cm⁻¹ corresponds to the in-plane vibrations of B₃N₃ hexagonal structure of BNNS. As confirmed in Fig. 3(b), the XRD spectrum has a sharp peak centered at 2θ ≈26.9°, corresponding to an interlayer spacing of 0.33 nm of h-BN lattice plane [34]. A small peak in Fig. 3(b) positioned at 2θ ≈27.9° is related to B₂O₃ content presented in the BNNS sample caused by the residual gas in the chamber. The largest peak is related to the Mo substrate. Figure 3(c) shows the Fourier transform infrared (FTIR) spectrum of the sample in the transmission mode [34]. The distinct features include two bands at 1200 cm⁻¹ and 1429 cm⁻¹, respectively. The weak peak at 1200 cm⁻¹ corresponds to the B₂O₃-stretching and B-O deformation mode, while the strong peak at 1429 cm⁻¹ is associated with the in-plane E_1u B–N bond stretching vibration of sp²-bonded h-BN phase. No evidence related to the BN in its cubic crystallographic structure (c-BN) was found in the Raman, XRD or FTIR measurements.

 

Fig. 3 Spectra of the BNNS deposited on Mo substrate: (a) Raman, (b) XRD, and (c) FTIR.

Download Full Size | PDF

Our measurements showed that the LPD was a versatile and scalable tool to produce high-quality h-BN nanosheets with a high yield and a digitized thickness control for the DUV photodetector applications. Since the performance of a DUV photodetector is mainly determined by the electron-hole pairs generated in the active layer, the desired thickness for total absorption should be optimized for low defects, low surface roughness, and good crystal structure homogeneity. In this paper, a 1.5 μm thick h-BN sample was chosen as a compromise between high quality h-BN nanosheets and responsivity of detector [26,27].

4. Vertical structure Al/BNNS/Mo deep UV photodetector

Among different photodetector design structures like p-i-n, Schottky photodiode and photodiode array (PDA), the planar configuration of MSM photoconductors was chosen for its simple design and fabrication process. The process flow of the prototypical DUV photodetector is shown in Fig. 4(a). After the metal aluminum was deposited as the top electrode, the device was annealed for 2 hours at 600°C in air to stabilize the metal contacts. The vertical structure was completed by the Mo substrate as the bottom electrode. In contrast to the conventional lateral structure with interdigitated finger electrodes on top surface, this vertical Al/BNNS/Mo configuration increased the overlap between the carriers and the electrical field and decreased the shadowing of the active area by the single-side metal electrodes. As shown in Fig. 4(b), the two electrodes are connected to a bias voltage V_s through a precision load resistor R. The current in and voltage variations across the resistor R are monitored during the experiment.

 

Fig. 4 Vertical structure MSM BNNS deep UV photodetector: (a) fabrication process flow, (b) schematic structure; and (c) the electric current as a function of the bias voltage.

Download Full Size | PDF

Under UV illumination, the conductance of the BNNS active layer changed as a function of incident UV intensity. The current-voltage (I-V) of the fabricated detector was characterized using an HP-Agilent 6268B power supply and an HEWLETT 34401A electrical meter. Figure 4(c) shows the typical dark current at room temperature as a function of the bias voltage. The dark current was less than 300 nA at 4 V and −200 nA at −5 V applied voltage, respectively. To assess the detector responsivity, a pen-ray DUV lamp (UVP, LLC) at 254 nm wavelength was used. When exposed with the UV light at an intensity of 1.0 mW/cm², the detector of an active surface area ~20 mm² had a responsivity of 1.5 mA/W at −5 V bias. Although a higher bias voltage led to a larger responsivity, the dark current increased too.

Figure 5 shows cyclical tests of the response strengths when the fabricated photodetector was illuminated with a pulsed UV light source of 50% duty cycle at the same intensity. The peak photocurrent increased with the increased bias but the response time was kept almost the same when the device was either positively or negatively biased. The actual response time was mainly dominated by the large capacitance from the large active area. The prototypical detector was also tested under different illumination intensities and operating temperatures. At −1.5 V bias, Fig. 6(a) shows the detector photocurrents when exposed to UV light intensities of 1.0 mW/cm², 0.28 mW/cm² and 0.16 mW/cm², respectively. The prototype showed good repeatability and stability, and the induced photocurrent was directly attributed to the absorption of DUV. Following the decrease of the UV light intensity down to 0.16 mW/cm², the generated photocurrent was only 6 nA. The signal-to-noise ratio also greatly decreased.

 

Fig. 5 Dynamic response time of the detector under 254 nm illumination at room temperature, when the device was (a) forward biased and (b) reverse biased at different voltages.

Download Full Size | PDF

 

Fig. 6 The measured detector photocurrent under 254 nm illumination, (a) at different radiation intensities; and (b) at different operating temperatures.

Download Full Size | PDF

It was also noticed that at the beginning of the measurement, the response time to the UV pulses was slightly longer and then it stabilized to a shorter value after many cycles elapsed. This phenomenon was probably due to the oxygen and water molecules adsorbed on the detector surface that reacted to the UV light and the photo-generated charge carriers [31]. Additional experiments have been performed at different operating temperatures, ranging from room temperature (RT) to 300°C, as shown in Fig. 6(b). The device is still functioning at 130°C, although the signal-to-noise ratio decreases by 3 times. At higher operating temperature, the signal-to-noise ratio deteriorates further. At 300°C, the thermal noise completely dominates the fabricated photodetector output, as the weak photocurrent signal has fully been merged into strong thermal noise. Limited by the available continuous UV light source, we could only test the detector at few UV wavelengths. The detector gave no response to the incident UV light at 300 nm or any longer [26,27].

5. Conclusions

High purity h-BN nanosheets have been synthesized by using pulsed LPD technique. This technique offers advantages of low processing temperature, digital control of the sample thickness, high yield and scalability. The SEM, TEM, Raman, XRD and FTIR measurement results have clearly indicated the sample was made of many single crystal randomly oriented h-BN nanosheets. The prototypical BNNSs based DUV detector with the vertical MSM configuration has been fabricated which demonstrated a stable operation from room temperature to 100°C, although the device design and fabrication could be further optimized. The experimental results have indicated that the BNNS is a promising band gap controllable semiconductor material for the solar-blind DUV photodetector and image array applications where operating at higher temperatures is required to save the energy and to prevent possible depositions of foreign materials on the cold detector surface.