As demonstrated by a typical SEM image (Fig. 1(a)), large quantities of InP NWs with a length of about 50–80 μm were achieved on the substrate. Basically, the NWs have small diameters of 29.1 ± 8.3 nm, based on AFM studies (Fig. A2), suggesting a large specific surface area of the NWs. XRD pattern of the product (Fig. 1(b)) was indexed to the zinc blende (ZB) crystal (JCPDS: 65-233) without any detectable wurtzite (WZ) phase. A TEM image of a typical InP NW in Fig. 1(c) shows smooth sidewall and uniform diameter of about 25 nm along its entire length with a Au particle at the tip, indicating the vapor–liquid–solid growth mechanism.^[21,22] The corresponding EDX spectrum inserted in Fig. 3(c) reveals that the NW consists of elements In and P with an equal ratio, very close to the stoichiometric value for a InP compound. Figure 1(d) represents a high-resolution TEM image recorded along the [112] zone axis. This image with the selected area electron diffraction (SAED) pattern inserted documents the high single-crystal quality with the ZB structure of our sample, and its growth direction is [111], which is consistent with the literature.^[23]

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (a) Typical SEM image of numerous InP NWs. (b) XRD pattern of the as-grown NWs. (c) TEM image and EDX spectrum of a single InP NW. (d) The corresponding HRTEM image together with the SAED pattern of the InP NW.

To further investigate the nature of InP NWs, optical characterizations were performed. A typical PL spectrum recorded from a single NW in Fig. 2(a) shows the emission peak located at 870 nm (∼ 1.42 eV), indicating a blue-shift of 50 nm compared with the bulk ZB InP PL (920 nm), which is much more moderate with respect to the defects related emission.^[24] This blue-shift can be explained by quantum confinement effect^[25] correlated well with the average diameter of our samples. Figure 2(b) shows the first-order Raman scattering spectrum of a single InP NW. By means of Lorentzian fitting, the two strong peaks observed can be associated with the transverse optical (TO) mode at 301 cm⁻¹ and the longitudinal optical (LO) mode at 339 cm⁻¹. Comparing with the TO and LO modes of bulk InP, the detected modes in our nanostructure have 2 cm⁻¹ and 6 cm⁻¹ downshifts, respectively, which is ascribed to the quantum confinement effect as well.^[26] In addition, the weak peak at 326 cm⁻¹ probably originates from the surface optical (SO) mode as a consequence of fine size of the InP NW.^[26] These results further confirm the high quality and the large specific surface area characteristics of the achieved InP NWs.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. (a) and (b) Typical μ-PL and Raman spectra of the as-grown InP NW excited with a 632.8 nm laser.

As illustrated in Fig. 3(a), back-gate single NW FETs were assembled to assess the electronic transport properties of the samples. Figure 3(b) shows the output characteristics of a representative device with a NW diameter of 30 nm and a channel length of 5.2 μm between two adjacent electrodes (a SEM image is shown in Fig. 3(a)). With the back-gate voltage (V_GS) rising from −10 V to 10 V, the conductance of the device gradually increases, revealing a good gate-dependence. Meanwhile, the transfer characteristics were investigated at a low drain bias (V_DS) of 0.1 V. As depicted in Fig. 3(c), the drain current (I_DS) increases when the V_GS mounts up, demonstrating a typical n-type semiconducting nature of the NWs, which is in agreement with the result observed by Hui et al.^[27] In this study, the electron mobility μ_e of InP channel was derived from the transfer curve via the following formula:^[28]

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (a) Schematic and SEM image of a back-gate single NW FET. (b) Output characteristics of the device at various VGS. (c) Transfer curves measured at VDS of 0.1 V.

Benefitting from the high mobility, the InP NWs can in principle be available for high performance photodetectors. We first measured its photoresponse properties under V_DS = 0.1 V and V_GS = 0 V on the two-terminal structure as displayed in Fig. 4(a) by turning 520 nm laser illumination with different power intensities ranging from to 103.1 mW/cm². It can be observed that the drain current increases with the increased voltage, at the same time, it increases apparently with the increased light intensities. The spectral responsivity (R_λ) and external quantum efficiency (EQE) of the device can then be obtained by the following equations^[29,30] R_λ = I_ph/PS and EQE = R_λ hc/eλ, where I_ph represents the photocurrent defined as the difference between light current and dark current (V_DS = 0.1 V and V_GS = 0 V), P and S (0.156 μm²) describe the corresponding light intensity and the effective illuminated area of the device, h refers to the Plank’s constant, c is the velocity of light, e is the electronic charge and λ is the incident light wavelength, respectively. An outstanding responsivity value of up to 2200 A·W⁻¹ can be obtained at the intensity of 4.64 mW/cm², although it trends a little downward when the laser intensity increases, the value of 1170 A·W⁻¹ is still pretty high even at 103.1 mW/cm², indicating an excellent conversion ability of optical to electrical signals and a low power consumption of the device (Fig. 4(b)). In addition, the achievable EQE value was estimated to be 2.80 × 10⁵% that can rival or even surpass the previous results.^[17–20]

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. (a) The I–V plots of the device in the dark and under 520 nm laser at different illumination intensities. (b) Responsivity and calculated EQE versus laser intensity at VDS = 0.1 V and VGS = 0 V. (c) Time-resolved switching behavior of the device under laser intensity of 103.1 mW/cm2 at VDS = 0.1 V. (d) Enlarged rise and fall sides over circle of 60–70 s.

To explore the capability of the photodetector for fast-varying optical signals, the time-response cycles of the device were obtained by periodically turning on and off the laser at 10 s intervals, under incident laser intensity of 103.1 mW/cm² at V_DS = 0.1 V and V_GS = 0 V. As plotted in Fig. 4(c), the device current rises up sharply and reaches to a steady state about 0.27 μA once the laser opened, and then falls down quickly to 0.08 μA when the light is blanked, demonstrating an excellent stability and reversible switching properties of the device. This good performance is confirmed by the longer-periodic time photoresponse measurements in Fig. A3. The rise time (τ_r) and fall time (τ_f), defined as the required time for the current rising to 90% and falling to 10% of the peak value, are about 110 ms and 130 ms, respectively, based on the enlarged curve in Fig. 4(d), which have comparability to the values improved by the ferroelectric-enhancing technique.^[31] These good properties (both responsivity and response speed) are particularly attractive merits for optoelectronics and probably result from two unique characteristics of our NW samples. (i) The superior crystal quality. InP NWs with few defects benefit the transport of the carriers, thus increasing the photoconductive gain and facilitating the photocurrent to reach a steady state rapidly in both the rise and fall processes.^[32,33] (ii) The small diameter of the NW (∼ 30 nm). On the one hand, the chemisorption around the NW becomes prominent as a result of the large specific surface area arising from its tiny structure. Consequently, adsorbates attached on the NW surface as hole-traps would prolong the lifetime of photo-generated carriers,^[34] which guarantees the high photo-gain and responsivity of our device when light is turned on. On the other hand, as light is turned off, defects and dangling bonds at the NW surface could increase the recombination rate of carriers, meaning a shorter fall time.^[35] In fact, both high responsivity and fast response speed imply that the surface trapping and recombination relaxation reach an equilibrium for the semiconductor channel.^[36]

Afterwards, we studied the effect of back-gate voltage on the performance of InP photodetectors. Figure 5(a) illustrates the plots of I_DS versus V_GS (V_DS = 0.1 V) under different laser intensities as before. The conductivity of InP channel under illumination in both OFF and ON states increases effectively across the whole gate voltage range employed, indicating that photocurrent dominates over thermionic and tunneling currents across the entire operating range of gate voltages. Figure 5(b) plots the photocurrents with different V_GS extracted from the transfer characterizations. Fitting the plots with power law of I_ph ∼ P^θ,^[35] we can extract the factor θ determining the efficiency of the photo-generated charges to light intensity as a function of V_GS (−10 V to 10 V). At a low gate voltage of −10 V, a linear increase (θ = 0.98) of the photocurrent with the light intensity confirms that the photocurrent is solely determined by the amount of photo-generated carriers.^[37] Whereas, when the gate bias further increases from −10 to 10 V at interval of 5 V, the device puts up a sublinear behavior, that is, the factor θ decreases monotonously with different values of 0.98, 0.92, 0.82, 0.74, and 0.65. The sublinear relation between the photocurrent and the laser intensity may relate to a complex photocurrent-generated mechanism.^[34,35] Figure 5(c) shows the responsivity and EQE recorded as a function of gate voltage under illumination intensity of 103.1 mW/cm² at 0.1 V bias, respectively. Similar to the modulation of channel conductivity, the gate voltage can bring a variation to the responsivity and EQE of InP photodetector. The responsivity of the device is only 138 A·W⁻¹ when V_GS = − 10 V, while V_GS sweeping to 10 V, a preeminent value up to 1436 A·W⁻¹ can be achieved and the EQE also increases by an order of magnitude. This gate-tunable property can provide the potential for the pixelated imaging applications.^[38]

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. (a) Plots of IDS versus VGS (VDS = 0.1 V) under different laser intensities. (b) Relationship between photocurrent and laser intensity at various VGS. The curves are fitted by the power law. (c) Responsivity and calculated EQE as a function of VGS recorded at VDS = 0.1 V under laser intensity of 103.1 mW/cm2, respectively.

The above observed behaviors of our InP NW photodetectors can be explicated by a simple energy band diagram as depicted in Fig. 6.^[31] As shown in Fig. 6(a), the InP device is in its quasi-equilibrium state under bias and illumination without back-gate voltage applied, which can be characterized by a small Schottky barrier at the interface between metal and semiconductor, where band tail drags especially slowly and depletion region extends deep into the channel.^[39] When a negative gate voltage is applied, a larger barrier between InP and metal electrode occurs as compared to the state at zero gate voltage, hindering photo-generated carriers extraction by the applied bias voltage, thus leading to a low responsivity (Fig. 6(b)). Fortunately, since the width of depletion region can expand further with Schottky barrier increasing, the resulting strong local electric field in the semiconductor can separate the electron–hole pairs effectively, which guarantees a good dependence of the photocurrent.^[40,41] At an extreme, the device photocurrent in the OFF state is merely determined by the photo-excited process and easily follows the linear relation to illumination intensity as described before in Fig. 5(b). Inversely, if the applied gate voltage is positive, the Schottky barrier can be lower, allowing more efficient photo-response for the device (Fig. 6(c)).^[42] However, the narrow depletion region cannot separate the electron–hole pairs effectively, which consequently easily deviate the relationship between photocurrent and illumination intensity from linear to sublinear dependence.

Fig. 6.

	Figure Option ViewDownloadNew Window
	Fig. 6. Band diagrams of InP photodetector under bias and illumination when different gate voltages were applied: (a) VGS = 0 V; (b) VGS < 0 V and (c) VGS > 0 V. The EF is the new quasi-Fermi level and ΦB is the Schottky barrier height. The full dots and the empty dots represent the photo-generated electrons and holes, respectively.