An ultra-fast WSe₂ homojunction photodiode with a large linear dynamic range towards in-sensor image processing

Abstract

The versatile photoresponse tunability of two-dimensional (2D) semiconductors achieved by tuning the gate voltage has opened promising pathways for in-sensor visual processing. However, the current limited dynamic range and slow response speed of the gate-tunable 2D photodetectors inhibit their implementation under challenging lighting conditions. Here, using a facile, efficient, and universal localized electrostatic screening strategy, we demonstrate an electrostatic screening enabled single-gate-tunable in-plane homojunction photodiode based on WSe₂ with bipolar high dynamic ranges (HDRs) and ultrafast photoresponse. The demonstrated WSe₂/PdSe₂ van der Waals (vdW) stacking in-plane homojunction photodiode leveraging the efficient band alignment and less interface recombination inherited from its vdW nature exhibits a large physical linear dynamic range (LDR) of up to 142 dB and an ultrafast response time down to 8 ns. These superior properties ensure that the device captures high-precision HDR images and performs in-sensor image processing with low latency. Our results provide an effective strategy for constructing 2D photodetectors with tunable positive/negative responses and high LDRs, which are promising for in-sensor visual processing of scenes with HDRs.

---

New concepts

This work introduces a novel and universally applicable electrostatic screening strategy for constructing 2D in-plane homojunction photodiodes with single-gate tunable photoresponse. The approach is demonstrated using a WSe₂/PdSe₂ van der Waals stacking system, which shows ultrafast photoresponse (down to 8 ns) and a large linear dynamic range of up to 142 dB. Unlike previous research that relies on complex multi-gate or heterojunction design strategies, this single-gate, electrostatically controlled strategy simplifies the device architecture while achieving high tunability and performance. The proposed concept also overcomes the limitations of slow response speed and narrow dynamic ranges typically observed in 2D photodetectors, particularly those with gate-tunable characteristics. Furthermore, the demonstrated device performs in-sensor image processing operations, such as edge enhancement, embossing, and blurring, with low latency and high accuracy, highlighting its potential for real-time, high-precision HDR imaging applications in dynamic lighting environments. This work provides new perspectives on how to design faster and more efficient 2D photodetectors, thereby advancing the field of in-sensor computing and visual processing.

1. Introduction

Machine vision systems consist of image acquisition, preprocessing, and analysis modules fulfilling the tasks of object detection, recognition, etc. similar to a human vision system.^1–3 This perceptive mode of optical information becomes one of the most important features for diverse intelligent scenarios, including robotics, unmanned aerial vehicles, and autonomous driving. Particularly, recently emerged in-sensor image sensing, which involves extracting optical information while acquiring images,^4–7 also known as artificial vision, provides a promising pathway towards the development of next-generation intelligent machine vision systems. Such a novel optoelectronic sensing paradigm requires photodetectors with ingenious tunable photoresponse properties that can mimic the neuromorphic vision process characterized by weight-updated synaptic behaviors and have remarkable advantages in terms of data latency, security and energy efficiency.^5,8,9 For practical applications of machine vision in an open-world scenario, unpredictable lighting conditions with huge variations of light illumination intensity for different areas of a scene may distort the captured images mainly due to the limited optical dynamic range of the photodetectors.^10–13

To achieve high dynamic range (HDR) imaging, strategies such as extra auxiliary lighting, multiple exposure, or image fusing by post-processing algorithms that have been adopted in conventional machine vision can also be employed in in-sensor imaging systems.^14,15 However, this inevitably further increases the complexity of the hardware and the processing overhead, resulting in additional time delays. Therefore, photodetectors for high-performance in-sensor imaging with tunable responsivity, large linear dynamic ranges (LDRs), and fast response speed are highly desired for accurate detection and real-time tracking of targets in complex lighting scenarios. However, the conventional complementary metal–oxide–semiconductor (CMOS) image sensors based on chemically doped silicon photodiodes face considerable challenges due to their electrical non-tunability of the photoresponse. Encouragingly, promising advances have been made in in-sensor computing using an electrostatically doped dual-gate p–i–n silicon photodiode.¹⁶

Alternatively, two-dimensional (2D) materials show strong light–matter interaction, high carrier mobility, and versatile gate-tunable physical properties, making them highly promising for high-performance photodetectors and intelligent imaging.^17–21 For example, owing to the strong gate tuning of the channel conductivity and therefore the photoresponsivity, the bilayer MoS₂ phototransistors show a large HDR of up to 140 dB²² and a visual adaptation range of up to 199 dB²³ for in-sensor imaging. However, such defect-involved gate-induced carrier transport typically shows non-linear and sub-linear photoresponses with slow response speed (milliseconds to tens of seconds), which may increase the data processing time. In contrast, a 2D photodiode intrinsically possesses high-speed characteristics because of the built-in electric field. The gate-tunable reversible 2D photodiodes are generally implemented with bipolar photoresponse using two device architectures: local electrostatic doping induced homojunctions by physically separated gates (splitting gates)^{1,5,16,23–25} and band-alignment-tunable heterojunctions by vertical stacking (global gates).^8,26–31 The former shows the response speed approaching nanoseconds but with a limited LDR (<100 dB). Moreover, this device architecture will substantially increase the complexity of electrical wiring required for the pixel-array-based imager. The latter has also been widely demonstrated. For instance, a notable demonstration of in-sensor imaging was presented by Pi et al., where they showed a band-alignment tunable 2D/2D van der Waals (vdW) stacking heterostructure (PdSe₂/MoTe₂) capable of simultaneous broadband image sensing and convolutional processing.⁸ In addition, 2D/3D heterostructures have also been applied for achieving gate-controlled reconfigurable photoresponse. For example, Yang et al. developed a novel approach using multi-terminal mixed-dimensional graphene–Ge heterostructure devices for in-sensor dynamic computing,⁹ enabling accurate detection and tracking of dim targets in low-light and complex scenarios. Despite these remarkable advancements, the response speed of such devices is still limited to microseconds to milliseconds. Therefore, a facile and general method to obtain single-gate controlled 2D reconfigurable photodetectors with large LDRs and ultrafast response is still lacking.

In this article, we report a reconfigurable, ultrafast, and HDR 2D photodetector for in-sensor visual processing devices based on a van der Waals tungsten diselenide (WSe₂) in-plane homojunction, which can be constructed by a facile and general localized electrostatic screening strategy. Owing to its high-quality interface, the demonstrated vdW WSe₂/PdSe₂ in-plane homojunction photodiode exhibits single-gate-tunable positive and negative photoresponses with ultrafast response speeds down to 8 ns and large physical LDRs of up to 142 dB. These properties allow efficient target detection and in-sensor visual processing in scenes with significant variations in brightness. Our results demonstrate that the utilization of HDR photodetectors substantially enhances target recognition accuracy compared with traditional non-HDR image sensors. Additionally, we showcase various in-sensor visual processing operations, including edge enhancement, embossing, and blurring, further highlighting the capabilities of our HDR in-sensor visual processing sensor.

2. Results and discussion

2.1. Overview and simulation of a localized electrostatic screening strategy

To realize a single-gate controlled in-plane 2D reconfigurable homojunction photodetector, we employ a typical ambipolar 2D semiconductor (for example, WSe₂) with a semi-screened gate instead of physically separated two splitting gates.^1,5,16 The schematic illustration of the device structure is shown in Fig. 1a. Due to the presence of the underneath screening layer with a much higher carrier concentration^32–34 (>10¹⁸ cm⁻³, green region), the electrostatic doping by applying gate voltage on WSe₂ (gray-white region) is locally screened according to the Debye shielding effect (Fig. S1 (ESI†)). In this way, a WSe₂ in-plane homojunction along the edge of the screening–unscreening region is created as shown by the dotted lines. Due to the ambipolar conducting properties of WSe₂, the polarity of the WSe₂ in-plane homojunction reverses under positive and negative gate voltages, resulting in bipolar tunable photoresponse properties (Fig. 1b). By simulating the potential distribution of the device under negative and positive gate voltages through Silvaco TCAD simulation (Fig. 1c and d), we can visually observe the screening effect. The WSe₂ in the blue region (Fig. 1c) and the red region (Fig. 1d) is electrostatically doped with p⁺⁺ and n⁺⁺ by negative and positive gate voltages, respectively, while the WSe₂ in the gray region is unaffected by the applied gate voltage and behaves as a weak intrinsic p-type (see details in Fig. S5c (ESI†)) semiconductor. The band energy extracted from the white dashed lines in Fig. 1c and d indicates the formation of p⁺⁺/p (Fig. 1e) and n⁺⁺/p (Fig. 1f) WSe₂ in-plane homojunctions, also corresponding to Fig. 1b(i) and (ii), respectively. More details of the Silvaco TCAD simulations are given in Code S1 (ESI†).

Fig. 1 Schematic and simulation of the in-plane homojunction formed by a localized electrostatic screening strategy. (a) Schematic of the WSe₂ in-plane homojunction bipolar photodiode formed using the localized electrostatic screening strategy. (b) Band alignment tunable WSe₂ in-plane homojunction. V_g < 0 V (i), V_g > 0 V (ii). (c) and (d) Silvaco TCAD simulated potential distribution in the WSe₂ in-plane homojunction under (c) negative and (d) positive gate bias. (e) and (f) Corresponding energy band diagram along the white dashed lines in (e) and (f), respectively.

2.2. Device structural characterization and electrical properties

Our photodetector with a gate-tunable photoresponsivity and a large physical LDR was achieved by leveraging a localized electrostatic screening-enabled WSe₂ in-plane homojunction, where the 2D PdSe₂ layer acts as both a screening layer and an electrode for charge collection. Fig. 2a presents the schematic of the WSe₂/PdSe₂/h-BN vdW stacking device on a SiO₂/Si substrate, where the Si acts as a bottom gate electrode. A dry transfer method was employed to fabricate the vdW stacks. Detailed information on the device fabrication can be found in the Experimental section. Fig. 2b shows the optical microscopy image of the device, with the WSe₂, PdSe₂, and h-BN edges marked with red, purple, and green dotted lines, respectively. The introduction of a 2D insulating layer material, h-BN, between WSe₂ and SiO₂ was intended to improve the interfacial contact between them.^35,36 The thicknesses of each layer, h-BN, PdSe₂, and WSe₂, were measured using an atomic force microscope (AFM) (Fig. 2c), yielding measurements of 16, 154, and 27 nm, respectively. We further verified the h-BN, PdSe₂, and WSe₂ stacking crystals by Raman spectra (Fig. S2 (ESI†)). Additionally, the clean interface between these 2D materials was also confirmed by high-resolution transmission electron microscopy (HRTEM) characterization and the corresponding energy-dispersive spectrometry (EDS) mapping on the cross-section of the fabricated stacks (Fig. S3 (ESI†)).

Fig. 2 Device structure, characterization, electrical properties, and band-alignments of the WSe₂ in-plane homojunction. (a) Schematic of the WSe₂ in-plane homojunction photodetector on a SiO₂/Si substrate, where the PdSe₂ layer acts as both a screening layer and an electrode for charge collection. (b) Optical microscopy image of the WSe₂ in-plane homojunction photodetector, where the red, purple, and green dashed regions indicate the WSe₂, PdSe₂, and h-BN nanosheets, respectively. (c) AFM image of the WSe₂/PdSe₂/h-BN stacks. The inset shows the height profile of the WSe₂/PdSe₂/h-BN stacks along the red solid line in (c). (d) and (e) Experimental (d) and simulated (e) logarithmic scale I_ds–V_ds curves of the WSe₂ in-plane homojunction at V_g = −80 V (red) and 80 V (blue). (f) Band alignment of the WSe₂ in-plane homojunction under different voltages. V_g = −80 V, V_ds = −1 V (i). V_g = −80 V, V_ds = 1 V (ii). V_g = 80 V, V_ds = −1 V (iii). V_g = 80 V, V_ds = 1 V (iv).

Fig. 2d shows the logarithmic scale current–voltage (IV) curves of the device, specifically between the source electrode (1) and drain electrode (4), under positive and negative gate voltages (−80 V and 80 V). The two curves exhibit a typical rectification characteristic with opposite polarities. Furthermore, the ideality factors are extracted by fitting the forward currents of the device utilizing a modified form of the Shockley diode formula expressed by the Lambert W function.^23,37 The ideality factors are found to be 1.4 (V_gs = −80 V, red line) and 1.2 (V_gs = 80 V, blue line), respectively. These ideality factors approaching 1 indicate that the forward current flow process in the device is mainly dominated by carrier diffusion rather than carrier recombination.^38,39 Based on the actual device geometry and material parameters shown in Table S1 (ESI†), we obtained similar rectification characteristics by Silvaco TCAD simulation (Fig. 2e and Code S2 (ESI†)), implying a possible homojunction diode transport behavior. Moreover, as shown in Fig. S4a and b (ESI†), the polarity and rectification ratio of the device can be dynamically modulated by the gate voltage. The excellent rectification characteristics of the device are further emphasized by the maximum rectification ratios of the photodiodes, which are 3 × 10⁵ at V_gs = 80 V and 7 × 10³ at V_gs = −80 V, respectively (Fig. S4b (ESI†)).

To elucidate the underlying mechanism behind the observed rectification behavior, the transport properties between the metal electrode and WSe₂ or PdSe₂ were also measured, as presented in Fig. S5a and b (ESI†). Both near-linear current–voltage curves indicate a quasi-ohmic contact⁴⁰ characteristic of the metal contacts, suggesting that the rectification behavior is independent of the metal contacts. It is worth noting that the PdSe₂ flake exhibits a high carrier concentration of 10¹⁸ cm⁻³ at 300 K,⁴¹ corresponding to the large I_ds (∼10⁻⁶ A at V_ds = 0.1 V, V_gs = 0 V) in the output and transfer curves of the PdSe₂ flake-based transistor (Fig. S5b and d (ESI†)). The high carrier concentration in the PdSe₂ flake effectively screens the electric field originating from the bottom gate,⁴² preventing the modulation of the carrier concentration and polarity in the WSe₂ layer above the PdSe₂ flake. Conversely, the carrier concentration and polarity in the WSe₂ layer above the SiO₂ insulator can be effectively adjusted using the bottom gate voltage (Fig. S5a and c (ESI†)). Therefore, a homojunction is formed between the WSe₂ layer above SiO₂ and the WSe₂ layer above PdSe₂ flakes, enabling gate-tunable characteristics observed in the device. Fig. 2f illustrates the band alignments under various bias conditions. Details of the band alignment were also investigated by ultraviolet photoelectron spectroscopy (UPS) and Kelvin probe force microscopy (KPFM) of WSe₂ and PdSe₂, as shown in Fig. S6 and S7 (ESI†) respectively. A small Schottky barrier of 0.22 eV could be formed by direct contact without considering electrostatic screening. However, this small Schottky barrier cannot be the dominant mechanism of the diode-like IV behavior, as verified by the latter photocurrent mapping measurement (see the following section, Section 2.3). When the gate voltage (V_g) is set to −80 V, the Fermi energy (E_F) of the WSe₂ layer above SiO₂ shifts toward the valence band edge (E_V), while the E_F of the WSe₂ layer above PdSe₂ flakes remains unaltered (slightly shifted towards the E_V). Consequently, an in-plane p⁺⁺–p homojunction of WSe₂ is formed (Fig. S8a (ESI†)). As can be seen in Fig. 2f(i) and (ii), V_ds < 0 V represents the forward bias condition, whereas V_ds > 0 V corresponds to the backward bias of a diode. Moreover, owing to the bipolar characteristics (Fig. S4c (ESI†)), the WSe₂ in-plane homojunction can also be modulated by gate voltage from the p⁺⁺–p junction to the n⁺⁺–p junction. Specifically, when V_g is set to 80 V, the E_F of WSe₂ shifts toward the conduction band edge (E_C), resulting in the formation of an n⁺⁺–p homojunction (Fig. S8b (ESI†)). In this case, V_ds > 0 V signifies the forward bias condition, while V_ds < 0 V corresponds to the backward bias of a diode, as depicted in Fig. 2f(iii) and (iv).

2.3. Photoresponse of the in-plane homojunction photodiode

The presence of gate-tunable polarity of rectification in the device allows gate bias tuning polarity and values of the photocurrent or responsivity R. This capability is crucial for in-sensor visual processing. The photoresponse of the device was measured at 520 nm. Fig. 3a presents the I_ds–V_ds curves under light illumination at different gate voltages (from −80 V to 80 V), where the photoresponse current at zero bias is the short circuit current (I_sc). The device demonstrates a positive photoresponse (I_sc > 0) when V_g < 0 V, while it exhibits a negative photoresponse (I_sc < 0) at V_g > 0 V. This behavior can be attributed to the reverse of the internal electric field of the device caused by the gate voltage. To gain deeper insights into the photoresponse mechanism, photocurrent mappings were conducted. The I_sc mappings at −80 V and 80 V are shown in Fig. 3b and c, respectively. The region exhibiting photosensitivity is located near the edge of the WSe₂/PdSe₂ stacks, encompassing an approximate area of 144 μm² (indicated by the white dashed area in Fig. 3b and c). This result indicates that the built-in electric field is confined to the edge of the WSe₂/PdSe₂ region rather than the entire contact region (the typical feature of metal–semiconductor Schottky junctions⁴³ and vertical stacking heterojunctions⁸). The I_sc mapping at zero gate voltage (Fig. S9 (ESI†)) further rules out the possibility of the metal–semiconductor Schottky junction and the WSe₂/PdSe₂ heterojunction contributing to the photoresponse. Furthermore, both the spectral response (Fig. S10a (ESI†)) of the device, corresponding to the WSe₂ absorption spectrum (Fig. S10d–f (ESI†)), and the unpolarized photoresponse (Fig. S10b and c (ESI†)) of the device confirm that the light absorption of the device is primarily attributed to the multilayer WSe₂. Therefore, we can conclude that the photoresponse of the device originates from the localized electrostatic screening-enabled WSe₂ in-plane homojunction. Except for the polarity of the photocurrent, the amplitude of the photocurrent can also be adjusted by the gate voltage, as depicted in Fig. 3d. Thus, the gate voltage serves as a control parameter that enables modulation of both the polarity and magnitude of the photocurrent.

Fig. 3 Gate-tunable positive and negative photovoltaic responses of the WSe₂ in-plane homojunction. (a) I_ds–V_ds curves under different gate voltages at 520 nm laser irradiation (134 mW cm⁻²). (b) and (c) The scanning I_sc mapping of the WSe₂ in-plane homojunction at 520 nm at V_g = −80 V (b) and 80 V (c). The photoresponse area of the device is approximately 144 μm² (white dashed area). (d) Gate-tunable positive and negative photoresponse under illumination at 520 nm with a power density of 125 mW cm⁻². (e) and (f) Rise and fall times for positive (f) and negative (i) photovoltaic modes of nanosecond pulsed laser. (g) Positive and negative photovoltaic current–time curves under illumination at 520 nm with a power density of 125 mW cm⁻². There was no significant degradation in device performance after ten months in the atmosphere.

Moreover, owing to the formation of high-quality in-plane homojunctions and clean interfaces from vdW stacking, our homojunction photodiode also has an ultrafast response time. To evaluate the response speed of the homojunction, we conducted measurements of the transient I–t curves using a high-speed measurement system (Fig. S11 (ESI†)) consisting of a nanosecond pulsed laser, a broadband amplifier, and a high-speed oscilloscope. The positive photoresponse exhibits rise and fall times down to 8 ns and 10 ns, respectively (Fig. 3e), while the negative photoresponse demonstrates rise and fall times of 8 ns and 9 ns, respectively (Fig. 3f), corresponding to a 3-dB cut-off frequency of above 20 MHz.³⁰ Note that such an ultrafast response speed surpasses most of the reported photodetectors (refer to Table S2, ESI†). This characteristic is particularly significant for rapid in-sensor visual image processing with low latency. In addition, the positive and negative photocurrents generated in the device are stable as can be seen from the current–time curves under modulated light illumination (Fig. 3g). There was no significant degradation in the photoresponse current of the device after 10 months in the atmosphere. Furthermore, we have fabricated several devices with identical structures but different material thicknesses, which exhibit similar gate-tunable photoresponses, as demonstrated in Fig. S12 (ESI†). These devices exhibit gate-tunable photoresponses that are independent of thickness variations, which may originate from the thickness-insensitive band structure of multilayer (>15 nm) transition metal dichalcogenides (TMDCs).^33,44 This high reproducibility of positive and negative photoresponses across multiple devices verifies the reliability of the observed phenomena.

It is worth noting that the approach to create WSe₂ in-plane homojunctions using a localized electrostatic screening is a universal method. By replacing the thick PdSe₂ flake with a graphite or even a 50 nm Au electrode (Fig. 4), we can also achieve similar gate-tunable positive and negative photoresponse characteristics. Schematics, optical microscopy images, and AFM images of the WSe₂/graphite and WSe₂/Au stacks are shown in Fig. 4a, d and Fig. S13a and d (ESI†). The scanning I_sc mappings and transient I–t curves of the WSe₂/graphite (Fig. 4b, c and Fig. S13b, c (ESI†)) and WSe₂/Au (Fig. 4e, f and Fig. S13e, f (ESI†)) stacks demonstrate that positive and negative photocurrents originated from the in-plane homojunction of WSe₂. These results indicate that PdSe₂ is unnecessary for the bipolar characteristics and therefore the reversible photoresponse. However, the power-dependent positive and negative photoresponses⁴⁵ of WSe₂/graphite and WSe₂/Au stacks (Fig. S14 (ESI†)) may limit their applications in in-sensor image processing. In contrast, the excellent performance of the WSe₂ in-plane homojunction photodiode formed by the PdSe₂ screening layers benefits from the efficient band alignment and less interface recombination inherited from vdW nature. In addition, we note that the large curvature of the stacking edge can also lead to a strong local electric field, which will further enhance the charge carrier transport across the junction.⁴⁶ However, such edge electric field enhancement that dominates the photocurrent spatial distribution is ruled out by using very thin graphite (≤10 nm, which still retains semi-metallicity) as a screening layer (Fig. S15 (ESI†)).

Fig. 4 Localized electrostatic screening strategy with graphite and Au as screening layers. (a) and (d) Schematic and optical microscopy image of the WSe₂ in-plane homojunction photodetector on a SiO₂/Si substrate, where the graphite (a) and Au (d) act as the screening layer. (b) and (c) The scanning I_sc mapping of the WSe₂ in-plane homojunction with a graphite screening layer at V_g = −80 V (b) and 80 V (c). (e) and (f) The scanning I_sc mapping of the WSe₂ in-plane homojunction with an Au screening layer at V_g = −80 V (e) and 80 V (f).

Next, we measured and discussed the optical dynamic range (DR) of the WSe₂/PdSe₂ in-plane homojunction photodiode. The DR for a photodetector is defined as:^3,13,47

where P_min (mainly affected by the quality of materials, responsivity, and temperature¹⁰) and P_max (mainly affected by recombination rate⁴⁸) denote the maximum signal level that the sensor can detect before saturation and the minimum detectable signal level above the sensor's noise floor, respectively. A large DR allows the photodetector to accurately capture the scenes with huge illumination differences using a single exposure, resulting in detailed high-quality images. Notably, a definition of the ten-times ratio of P_min and P_max in the log scale is also used to define the DR.^10,49,50 In practical applications, achieving a linear response in photodetectors is crucial for accurately converting optical signals into electrical signals and avoiding distortion.¹⁰ The LDR represents the power range within which a photodetector maintains a linear response to incident light. Additionally, compared with the sublinear relationship between photocurrent and light intensity observed in most 2D phototransistors, a photodiode with a linear relationship is more suitable for implementing in-sensor processing.^5,8Fig. 5a and b presents the positive and negative photoresponses obtained under various light intensities, under a bias voltage of 0 V. The extracted I_sc as a function of light intensity is shown in Fig. 5c and d. Both positive and negative photoresponses exhibit linear behavior until deviating from linearity at high light intensities. The photoresponsivity R is expressed as:¹⁰where I_sc is the photocurrent at zero bias, Φ is the light intensity, and S is the photosensitive area. The photoresponsivities for positive and negative photoresponses remain constant in the range of P_min and P_max, but decrease when Φ exceeds P_max, as shown in Fig. 5c and d. In these cases, the LDR at −80 V V_g is 122 dB, while it reaches 142 dB at 80 V V_g. The minimum incident light power P_min (= Φ_min × S) on the device is 18.7 pW (−80 V) and 3.3 pW (80 V), respectively, which approaches the noise equivalent power (NEP) of the system without bandwidth normalization (Fig. S16 (ESI†)). The ultra-low NEP at zero-bias operating voltage, high-quality homojunctions, and clean interfaces contribute to the large LDR of the device. The insets of Fig. 5c and d visually illustrate the large LDR of the device, covering a wide illumination range from full moon to sunny noon conditions (Fig. S17 (ESI†)). Such a large LDR of the device at multiple gate voltage (Fig. S18 (ESI†)) highlights its potential for HDR visual processing. We compared the photodetection performance, including response speed and LDR values, of our electrostatic screening-enabled homojunction photodiode with other state-of-the-art 2D reconfigurable bipolar photodetectors with either splitting gates or global gates (Fig. 5e; more detailed parameter comparisons are shown in Table S2 (ESI†)). The results indicate that our semi-screened single gate-controlled bipolar photodiodes combine ultrafast response speed and large LDRs, which could be promising for in-sensor image processing under harsh lighting conditions with low latency.

Fig. 5 Gate-tunable linear photoresponse characteristics of a WSe₂ in-plane homojunction photodiode over a large brightness range. (a) and (b) Positive (a) and negative (b) photoresponses under different Φ (520 nm) at V_g = −80 V and 80 V, respectively. (c) and (d) Φ dependence of the extracted I_sc and responsivity at V_g values of −80 V (c) and 80 V (d). The I_sc–Φ curves of the device exhibit nearly perfect linearity within the light power density at ±80 V gate voltage, corresponding to a constant R of 29.8 mA W⁻¹ (c) and 47.9 mA W⁻¹ (d), respectively. The LDR values of the device are 122 dB (c) and 142 dB (d), respectively. The inset photos visually illustrate the large LDR of the device, covering a wide illumination range from full moon to sunny noon conditions. (e) A plot of the LDR versus response time for reconfigurable photodetectors. Based on the device structure, reconfigurable photodetectors were classified into three categories, namely, splitting gates (brown diamonds), global gates (blue triangles), and semi-screened gates (red pentagram). S, I, G, and Scr. correspond to the semiconductor, insulator, gate, and screening layers, respectively.

2.4. HDR in-sensor image processing

Finally, we show that our gate-tunable bipolar in-plane homojunction photodiode can be used for HDR in-sensor image processing. We simulated such image processing in a driving scenario at a tunnel exit based on a single high-performance device,^5,8 where the illuminance contrast between the tunnel interior (less than 10 lx) and the exterior (over 10⁵ lx) is ∼100 dB.^46,47,49 To capture images of a high-speed vehicle when entering/exiting tunnels using a single image sensor, the sensor must possess an HDR exceeding 100 dB^51–53 with a fast response speed (>120 fps). Otherwise, the resulting images will be distorted (such as in Fig. 6a). It is notable that human eyes have a HDR of about 140 dB⁵⁴ through the visual adaptation of retinal cells and adjustment of the pupil diameters, a remarkable ability to adapt to various lighting conditions, although this process of light/dark adaptation takes a long time (about 1–30 min) and is nonlinear⁵⁵ (Fig. S17 (ESI†)). Typical modern high-speed CMOS image sensors (>120 fps) in digital cameras and smartphones (∼60 fps) usually have an LDR of around 60 to 80 dB.³ In contrast, our WSe₂/PdSe₂ in-plane homojunction photosensor offers a larger physical LDR and an ultrafast response speed and has the ability to achieve positive and negative photovoltaic responses by tuning the gate voltage and hence it is more suitable for implementing in-sensor processing in scenes with a wide range of brightness levels.^5,8

Fig. 6 HDR in-sensor image processing. (a)–(c) Schematic of extracting edge features in a scene with large differences in brightness using HDR in-sensor visual processing. From left to right corresponds to the scene of tunnel exit (a), the reconfigurable image sensor (b), and the output processed image (c), respectively. The insets of (a) and (c) are zoomed-in views of the HDR image and edge extraction image of the van near the tunnel exit, respectively (red-bordered rectangular areas). (d) Schematic of in-sensor processing using our WSe₂ bipolar photodiode. The image projected on the sensor is partitioned into pixels, with each pixel further divided into subpixels. The light intensity incident on each subpixel is denoted as P_j. The output photocurrent (I_ph) from each pixel is described as I_ph = ∑P_jR_j, where R_j represents the photoresponsivity of the subpixel sensor (also considered as the weight of the kernel matrix), which can be tuned by a gate voltage. (e)–(g) Grayscale visual imaging (e), edge enhancement (f), and embossing (g) for HDR (>120 dB) WSe₂ in-plane homojunction devices. (h)–(j) Grayscale visual imaging (h), edge enhancement (i), and embossing (j) for non-HDR (<100 dB) reconfigurable photosensors. The realization of these functions by individually updating the gate voltage for devices. The different colors/numbers in the array correspond to different responsivities/kernel weights. [Image attribution: https://www.sony.com/en/SonyInfo/vision-s/safety.html. Licensed by Sony Group Corporation.]

To enable in-sensor visual-like image processing, each pixel in the image sensor is divided into a 3 × 3 subpixel array (the red dashed rectangle in Fig. 6b), corresponding to a 3 × 3 kernel matrix. By adjusting the gate voltage of each subpixel, i.e., the bipolar WSe₂ in-plane homojunction photodiode, the responsivity can be tuned from negative to positive, allowing for the implementation of various types of kernels (Fig. S19 (ESI†)). It should be emphasized that the current gate voltage can be largely reduced by a high-k gate dielectric⁵⁶ to meet the voltage requirement in digital logical circuits. The large LDR of the photodiodes (122 dB/142 dB, red/blue lines in Fig. 6c) completely covers the luminance range inside and outside the tunnel (Fig. S17 (ESI†)). Their ultrafast response speed (8 ns) also ensures their real-time imaging capability. As shown in Fig. 6d, each subpixel receives incident light with an intensity of P_j. The output photocurrent (I_ph) of each pixel is obtained by summing the photocurrents from all the subpixels (I_ph = ∑P_jR_j), where R_j represents the photoresponsivity of the subpixel. The photoresponsivity, which is analogous to the weight of the kernel matrix, can be adjusted by applying a gate voltage, enabling control over both the amplitude and polarity. Fine adjustments to the polarity and responsivity of these subpixels allow the pixel to function as an image processing operator, realizing high fidelity, high precision, and low latency in-sensor visual processing tasks like edge extraction of a forward truck, as demonstrated in Fig. 6c. Fig. 6e–g and Fig. S20 (ESI†) show images generated using different kernels with high LDR detectors, including original image capture, edge detection, and emboss processing. Additional examples of HDR in-sensor visual processing can be found in Fig. S21 (ESI†). In comparison to non-HDR sensors (Fig. 6h–j), our device demonstrates a clear advantage in in-sensor visual processing, particularly in scenes with significant illuminance variation. This advantage stems from the large physical LDR of the WSe₂ in-plane homojunction device, which cannot be replaced by software-based HDR processing techniques or conventional HDR CMOS image sensors.

3. Conclusions

In summary, we have successfully constructed a WSe₂ in-plane homojunction photovoltaic photosensor with a large physical LDR and single-gate-tunable optoelectronic properties through a localized electrostatic screening approach. Moreover, this localized electrostatic screening strategy for constructing in-plane homojunctions is facile, efficient, and universal. The WSe₂/PdSe₂ in-plane homojunction photodiode demonstrates remarkable single-gate-tunable positive and negative photoresponse characteristics with an ultrafast response time down to 8 ns. Additionally, its low NEP of 3 pW and high saturation optical power of 40 μW contribute to a remarkable LDR of up to 142 dB. Compared to the state-of-the-art non-HDR photosensors, it could enhance recognition accuracy and offer distinct advantages in the intelligent in-sensor visual processing of scenes with a large brightness range and low data latency.

4. Experimental section

4.1. Device fabrication

Multilayer WSe₂ and PdSe₂ nanosheets were obtained by mechanical exfoliation from single crystals purchased from Nanjing Muke Nanotechnology Co., Ltd. Multilayer h-BN nanosheets were mechanically peeled off from the bulk crystal purchased from SixCarbon Tech. Shenzhen. The WSe₂/PdSe₂/h-BN stacks were prepared by a polyvinyl chloride-assisted full dry transfer method⁵⁷ in an accurate transfer system (Metatest, E1-T). Source and drain electrode areas of the WSe₂/PdSe₂ photodetector were defined on SiO₂ (285 nm)/Si substrates by electron beam lithography (EBL, Tescan SHB) and maskless ultraviolet photolithography (DL10, Zeptools) technologies, followed by thermal evaporation of In/Au (5/50 nm) electrodes. The photoresponse area of the WSe₂/PdSe₂ stacking device is approximately 144 μm².

4.2. Material characterization

The surface morphologies of the devices were characterized using an optical microscope (Olympus, BX51). AFM and Kelvin probe force microscopy (KPFM) patterns were characterized using the scanning probe microscope (Park NX10) in non-contact modes and electrical modes, respectively. Raman spectra were recorded on a Raman spectrometer (Andor SR-500i-A-R Raman system with a 532 nm excitation laser). The energy band structures were tested by UPS (Thermofisher, Escalab Xi+).

4.3. Device characterization

The photoelectric response properties were tested using the Metatest ScanPro system (Metatest, ScanPro Advance) equipped with a Keithley 2636B source-meter system (Tektronix Inc). We used 520 nm monochromatic lasers as light sources and the lasers were focused by a 50× objective lens with Gaussian beam diameters of 2 μm and 32 μm for the single mode (for scanning photocurrent mapping) and multimode lasers, respectively. Both 520 nm lasers were triggered in DC or AC mode by a digital modulator (DG822, RIGOL). Tuning the neutral filter with different OD values of the lasers allows the light intensity to vary from 10⁻³ to 10⁵ mW cm⁻². A xenon lamp (C12122-020-85, Hamamatsu) with a grating spectrometer (mSpec300, Metatest) was used for spectral responsivity testing. All light intensities were measured using photodiode power sensors (S130VC and S132C, Thorlabs). The device response speed was obtained using a nanosecond pulsed 405 nm laser, a signal generator (DG852 Pro, RIGOL), an amplifier (FEMTO DHPCA-100), and an oscilloscope (DHO4404, RIGOL). The device noise currents were measured using a Keysight B1500A semiconductor analyzer. The calculation of NEP is detailed in Fig. S16 (ESI†). All the measurements were performed at atmospheric pressure and room temperature.

Author contributions

Chun Li: conceptualization, formal analysis, funding acquisition, methodology, project administration, resources, supervision, writing – reviewing and editing; Changyong Lan: conceptualization, formal analysis, funding acquisition, methodology, resources, supervision, writing – reviewing and editing; Shaofeng Wen: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, writing – original draft, writing – reviewing and editing; Shuren Zhou: formal analysis; Yimin Gong: formal analysis; Rui Zhang: formal analysis, visualization; Xinyu Jia: formal analysis; Lingkang Kong: formal analysis; Haodong Fan: formal analysis; Yi Yin: funding acquisition; and Yong Liu: funding acquisition.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (Grant No. 61975024 and 62074024), the Natural Science Foundation of Sichuan Province (Grant No. 2022NSFSC0042), and the Sichuan Science and Technology Program (Grant No. 2023YFH0090 and 2023NSFSC0365).

References

1. L. Mennel, J. Symonowicz, S. Wachter, D. K. Polyushkin, A. J. Molina-Mendoza and T. Mueller, Nature, 2020, 579, 62–66.
2. T. Wang, M. M. Sohoni, L. G. Wright, M. M. Stein, S.-Y. Ma, T. Onodera, M. G. Anderson and P. L. McMahon, Nat. Photonics, 2023, 17, 408–415.
3. F. Liao, Z. Zhou, B. J. Kim, J. Chen, J. Wang, T. Wan, Y. Zhou, A. T. Hoang, C. Wang, J. Kang, J.-H. Ahn and Y. Chai, Nat. Electron., 2022, 5, 84–91.
4. T. Wan, B. Shao, S. Ma, Y. Zhou, Q. Li and Y. Chai, Adv. Mater., 2023, 35, 2203830.
5. Y. Zhou, J. Fu, Z. Chen, F. Zhuge, Y. Wang, J. Yan, S. Ma, L. Xu, H. Yuan, M. Chan, X. Miao, Y. He and Y. Chai, Nat. Electron., 2023, 6, 870–878.
6. W. Pan, J. Zheng, L. Wang and Y. Luo, Engineering, 2022, 14, 19–21.
7. Y. Liu, R. Fan, J. Guo, H. Ni and M. U. M. Bhutta, Intell. Comput., 2023, 2, 0043.
8. L. Pi, P. Wang, S.-J. Liang, P. Luo, H. Wang, D. Li, Z. Li, P. Chen, X. Zhou, F. Miao and T. Zhai, Nat. Electron., 2022, 5, 248–254.
9. Y. Yang, C. Pan, Y. Li, X. Yangdong, P. Wang, Z.-A. Li, S. Wang, W. Yu, G. Liu, B. Cheng, Z. Di, S.-J. Liang and F. Miao, Nat. Electron., 2024, 7, 225–233.
10. Y. Wei, C. Lan, S. Zhou and C. Li, Appl. Sci., 2023, 13, 11037.
11. S. Zhou, S. Wen, H. Fan, Y. Wei, Y. Yin, C. Lan, C. Li and Y. Liu, ACS Photonics, 2024, 11, 1810–1820.
12. S. Goossens, G. Navickaite, C. Monasterio, S. Gupta, J. J. Piqueras, R. Pérez, G. Burwell, I. Nikitskiy, T. Lasanta, T. Galán, E. Puma, A. Centeno, A. Pesquera, A. Zurutuza, G. Konstantatos and F. Koppens, Nat. Photonics, 2017, 11, 366–371.
13. H. Ko, S. Park, H. J. Son and D. S. Chung, Chem. Mater., 2020, 32, 3219–3228.
14. P. E. Debevec and J. Malik, SIGGRAPH ’08: ACM SIGGRAPH 2008 classes, 2008, 31, 1–10.
15. A. Jacobs, Adv. Build. Energy Res., 2007, 1, 177–202.
16. H. Jang, H. Hinton, W.-B. Jung, M.-H. Lee, C. Kim, M. Park, S.-K. Lee, S. Park and D. Ham, Nat. Electron., 2022, 5, 519–525.
17. Y. Liu, X. Duan, H.-J. Shin, S. Park, Y. Huang and X. Duan, Nature, 2021, 591, 43–53.
18. Y. Liu, X. Duan, Y. Huang and X. Duan, Chem. Soc. Rev., 2018, 47, 6388–6409.
19. C. Liu, H. Chen, S. Wang, Q. Liu, Y.-G. Jiang, D. W. Zhang, M. Liu and P. Zhou, Nat. Nanotechnol., 2020, 15, 545–557.
20. D. Akinwande, C. Huyghebaert, C.-H. Wang, M. I. Serna, S. Goossens, L.-J. Li, H.-S. P. Wong and F. H. L. Koppens, Nature, 2019, 573, 507–518.
21. H. Kim, S. Z. Uddin, D.-H. Lien, M. Yeh, N. S. Azar, S. Balendhran, T. Kim, N. Gupta, Y. Rho, C. P. Grigoropoulos, K. B. Crozier and A. Javey, Nature, 2021, 596, 232–237.
22. Y. Zhou, J. Fu, T. Wan, L. Xu, S. Ma, J. Chen, X. Miao, Y. He and Y. Chai, IEDM, 2022 DOI:10.1109/IEDM45625.2022.10019350.
23. J. Jiang, W. Xu, F. Guo, S. Yang, W. Ge, B. Shen and N. Tang, Nano Lett., 2023, 23, 9522–9528.
24. Y. Wang, H. Sun, Z. Sheng, J. Dong, W. Hu, D. Tang and Z. Zhang, Nano Res., 2023, 16, 12713–12719.
25. Y. Zhang, K. Ma, C. Zhao, W. Hong, C. Nie, Z.-J. Qiu and S. Wang, ACS Nano, 2021, 15, 4405–4415.
26. J. Yang, Y. Cai, F. Wang, S. Li, X. Zhan, K. Xu, J. He and Z. Wang, Nano Lett., 2024, 24, 5862–5869.
27. H. Wang, Y. Li, P. Gao, J. Wang, X. Meng, Y. Hu, J. Yang, Z. Huang, W. Gao, Z. Zheng, Z. Wei, J. Li and N. Huo, Adv. Mater., 2024, 36, 2309371.
28. J. Huang, K. Shu, N. Bu, Y. Yan, T. Zheng, M. Yang, Z. Zheng, N. Huo, J. Li and W. Gao, Sci. China Mater., 2023, 66, 4711–4722.
29. Y. Zhang, L. Wang, Y. Lei, B. Wang, Y. Lu, Y. Yao, N. Zhang, D. Lin, Z. Jiang, H. Guo, J. Zhang and H. Hu, ACS Nano, 2022, 16, 20937–20945.
30. Y. Zhang, W. Shen, S. Wu, W. Tang, Y. Shu, K. Ma, B. Zhang, P. Zhou and S. Wang, ACS Nano, 2022, 16, 19187–19198.
31. F. Liu, X. Lin, Y. Yan, X. Gan, Y. Cheng and X. Luo, Nano Lett., 2023, 23, 11645–11654.
32. C. M. Koo, P. Sambyal, A. Iqbal, F. Shahzad and J. Hong, Two-Dimensional Materials for Electromagnetic Shielding, WILEY-VCH GmbH, Weinheim, 2021.
33. Y. Wang, J. Pang, Q. Cheng, L. Han, Y. Li, X. Meng, B. Ibarlucea, H. Zhao, F. Yang, H. Liu, H. Liu, W. Zhou, X. Wang, M. H. Rummeli, Y. Zhang and G. Cuniberti, Nano-Micro Lett., 2021, 13, 143.
34. H. Han, B. Zhang, Z. Zhang, Y. Wang, C. Liu, A. K. Singh, A. Song, Y. Li, J. Jin and J. Zhang, Nano Lett., 2024, 24, 8602–8608.
35. X. Luo, K. Andrews, T. Wang, A. Bowman, Z. Zhou and Y.-Q. Xu, Nanoscale, 2019, 11, 7358–7363.
36. C. R. Dean, A. F. Young, I. Meric, C. Lee, L. Wang, S. Sorgenfrei, K. Watanabe, T. Taniguchi, P. Kim, K. L. Shepard and J. Hone, Nat. Nanotechnol., 2010, 5, 722–726.
37. B. W. H. Baugher, H. O. H. Churchill, Y. Yang and P. Jarillo-Herrero, Nat. Nanotechnol., 2014, 9, 262–267.
38. Y. Yang, N. Huo and J. Li, J. Mater. Chem. C, 2018, 6, 11673–11678.
39. J. Sun, Y. Wang, S. Guo, B. Wan, L. Dong, Y. Gu, C. Song, C. Pan, Q. Zhang, L. Gu, F. Pan and J. Zhang, Adv. Mater., 2020, 32, 1906499.
40. C. He, M. Cheng, T. Li and W. Zhang, ACS Appl. Nano Mater., 2019, 2, 2767–2775.
41. W. Nishiyama, T. Nishimura, K. Ueno, T. Taniguchi, K. Watanabe and K. Nagashio, Adv. Funct. Mater., 2022, 32, 2108061.
42. T. Shen, J. Liu, X. Liu, P. Cheng, J. C. Ren, S. Li and W. Liu, Adv. Funct. Mater., 2022, 32, 2207018.
43. I.-H. Ahn, J. Ahn, D. K. Hwang and D. Y. Kim, Results Phys., 2021, 21, 103854.
44. X. Yao, Y. Wang, X. Lang, Y. Zhu and Q. Jiang, Phys. E, 2019, 109, 11–16.
45. Z. Gao, R. Jiang, M. Deng, C. Zhao, Z. Hong, L. Shang, Y. Li, L. Zhu, J. Zhang, J. Zhang and Z. Hu, Adv. Mater., 2024, 36, 2401585.
46. F. Wu, H. Tian, Y. Shen, Z. Hou, J. Ren, G. Gou, Y. Sun, Y. Yang and T.-L. Ren, Nature, 2022, 603, 259–264.
47. H. T. Chandran, S. Mahadevan, R. Ma, Y. Tang, T. Zhu, F. Zhu, S.-W. Tsang and G. Li, Appl. Phys. Lett., 2024, 124, 101113.
48. X. Meng, Y. Du, W. Wu, N. B. Joseph, X. Deng, J. Wang, J. Ma, Z. Shi, B. Liu, Y. Ma, F. Yue, N. Zhong, P. Xiang, C. Zhang, C. Duan, A. Narayan, Z. Sun, J. Chu and X. Yuan, Adv. Sci., 2023, 10, 2300413.
49. J.-M. Liu, Principles of photonics, Cambridge University Press, New York, 2016.
50. X. An, F. Liu, Y. J. Jung and S. Kar, Nano Lett., 2013, 13, 909–916.
51. E. Setyaningsih and J. Pragantha, IOP Conf. Ser.:Mater. Sci. Eng., 2019, 508, 012077.
52. Z. Wang and H. Gao, Electronics, 2023, 12, 443.
53. P. Zhang, G. Liang, M. Zhang and P.-J. Zhang, Highways & Automotive Application, 2023 DOI:10.20035/j.issn.1671-2668.2023.04.007.
54. Sony Group Portal, https://www.sony.com/en/SonyInfo/vision-s/safety.html.
55. M. Azimipour, D. Valente, K. V. Vienola, J. S. Werner, R. J. Zawadzki and R. S. Jonnal, Opt. Lett., 2020, 45, 4658.
56. S. Wen, R. Zhang, Z. Yang, S. Zhou, Y. Gong, H. Fan, Y. Yin, C. Lan, C. Li and Y. Liu, ACS Appl. Nano Mater., 2023, 6, 16970–16976.
57. Y. Wakafuji, R. Moriya, S. Masubuchi, K. Watanabe, T. Taniguchi and T. Machida, Nano Lett., 2020, 20, 2486–2492.

---

Footnote

† Electronic supplementary information (ESI) available: Raman characterization, HRTEM and EDS mapping of WSe₂/PdSe₂/h-BN stacks; electrical properties of the WSe₂ in-plane homojunction; output and transfer curves of a multilayer WSe₂ and PdSe₂ FET; UPS and band structure of WSe₂; AFM and KPFM measurements of thick WSe₂ and PdSe₂; band alignment of WSe₂ at different gate voltages; photocurrent mapping of the WSe₂ in-plane homojunction at zero gate voltage; spectral responsivity and polarization sensitivity of the WSe₂ in-plane homojunction; a high-speed photoresponse testing system of the device; device reproducibility; demonstration of the universality by WSe₂/graphite and WSe₂/Au devices; power-dependent positive and negative photoresponses of WSe₂/graphite and WSe₂/Au devices; demonstration of the universality by the WSe₂/thin graphite device; measurements of noise current and NEP; illuminance levels for daily life and DR of the human eye; the large LDR of the device at multiple gate voltages; more modes of in-sensor image processing; performance comparison of 2D-based reconfigurable photodetectors; Silvaco TCAD simulation of potential and band energy distributions; and Silvaco TCAD simulation of I_ds–V_ds curves at different V_g values (PDF). See DOI: https://doi.org/10.1039/d4nh00656a

---