Silicon photodiode-competitive 2D vertical photodetector

The flexible form factor of electronic devices plays a crucial role in Sensor 4.0, as it allows for integration onto non-flat or dynamically changing surfaces such as human skin, textured surfaces, curved robotic surfaces, and more¹. While silicon photodiodes are widely used in electronic devices for photosensors due to their affordability, reliability, and high performance, they are challenging to utilize in flexible applications. Although flexible photodetectors based on silicon nanomembranes have been reported², achieving performance comparable to their bulk counterparts is difficult due to the weak absorption in silicon nanomembranes. Tremendous efforts have been paid to explore next-generation flexible photodetectors with emerging nanomaterials such as organic material^3,4, organic-inorganic hybrid perovskite material⁵, quantum dot⁶, and two-dimensional (2D) materials⁷. Among these, 2D materials possess inherent advantages in high absorption coefficient⁸, high carrier mobility⁹, good stability¹⁰, excellent flexibility¹¹, and broadband detection^12,13,14,15. Various strategies, such as heterostructure construction^16,17, chemical doping¹⁸, electrical doping¹⁹, and contact engineering²⁰ have been employed to develop high-performance photodetectors based on 2D materials. Some devices^{21,22,23,24,25,26} have showcased exceptional photodetection performance, including rapid response times of up to 5.5 picoseconds²⁶, low dark current of 15 pA²⁵, and high responsivity of 365 mA/W²¹. However, achieving these advanced performances within a single device poses significant challenges. Furthermore, the operational stability of 2D photodetectors is a critical concern as atomically thin crystals are highly sensitive to atmospheric conditions²⁷. Additionally, many device structures are not suitable for flexible applications due to the requirement of additional brittle bulk semiconductors or dielectric films²³.

The performance limitations of 2D material-based photodetectors can be attributed to three main factors. Firstly, the internal quantum efficiency of photocarriers is often low due to the relatively high binding energy²⁸, short carrier lifetime below 1 ns²⁹, and long channel lengths typically exceeding 1 μm in most 2D material-based devices. Secondly, the doping engineering of 2D materials is not as advanced as that of conventional three-dimensional bulk crystals. Despite the existence of diverse reported heterostructures^18,30,31,32, effectively suppress dark current remains an ongoing challenge. Lastly, the trapping of photocarriers in defect states is a common issue in devices based on 2D materials³³, leading to relatively slow response times in the range of milliseconds or even slower³⁴. Taking into account the aforementioned issues, it is vital to make a short channel length and minimize defects at the electrode interface in order to minimize photocarrier loss and ensure reliable photocurrent. Furthermore, to achieve low dark current, effective suppression of majority carrier transport under dark condition is required. The multiple aspects must be considered in parallel to realize a high-performance photodetector that can be competitive with silicon photodiodes.

In this study, we propose a silicon photodiode-competitive flexible 2D vertical photodetector, which can be achieved by simply integrating van der Waals metal electrodes and 2D materials along its vertical direction. The prototype devices have been demonstrated, which consist of n-type van der Waals electrodes (Ti and Ag) and p-type semiconducting material WSe₂. By leveraging the high hole Schottky barrier in the van der Waals contact, we successfully suppress majority carrier drift and achieve a low dark current of 0.8 pA under a 0.5 V bias voltage (corresponding to a dark current density of 4 × 10^-6A/cm⁻²). A high photocurrent of 1.2 μA was generated under the light power of 6.8 μW with 458 nm laser illumination, and the responsivity of 0.18 A/W as well as the external quantum efficiency of 49% were achieved. The photocurrent is attributed to the photocarrier drift, as evidenced by its strong linear dependence on light power. Furthermore, the actual response time defined as time difference from 90% to 10% of maximum photocurrent variation reveals fast response time ~1 μs. Ultrafast photocurrent measurement, which can show intrinsic picosecond level response time of the device³⁵, showed the intrinsic response time of this device can achieve 337 ps, which is due to the ultrashort channel length. We extensively evaluated the photodetection performance of the device under different bias voltages, light powers, wavelengths, temperatures, and bending conditions. The device demonstrates robust suppression of dark current and reliable photocurrent, making it highly stable even at high temperatures and flexible conditions under ambient circumstances. Our findings provide a facile and practical approach for designing high-performance flexible photosensors based on 2D materials.

Efficient photocarrier extraction and dark current suppression in Ti/WSe₂/Ag photodetector

Figure 1a, b illustrate our strategies for achieving a highly sensitive photodetector with a high light on-off ratio based on the 2D vertical device. To maximize the light on/off ratio, the device relies on key factors including low dark current and high photocurrent in which achieving enhanced photocurrent necessitates efficient photocarrier collection and an optimized device architecture. In Fig. 1a, we demonstrate the efficient extraction of photocarriers in the ultra-short channel in the 2D vertical device. Assuming a 2D material thickness of 50 nm, the ultra-short channel length allows for a mean drift length of 25 nm for the photocarriers²⁶. Under a 0.5 V bias voltage (ignoring contact resistance), the electric field can reach up to 0.1 MV/cm, facilitating fast photocarrier extraction due to the short drift length and strong electric field. Additionally, by leveraging a strong electric field and a short channel length, photocurrent loss mechanisms, such as recombination, can be minimized, thereby enhancing the photocurrent²⁶. The scheme for achieving a low dark current is illustrated in Fig. 1b, where we showcase the carrier transport under dark and light conditions in precisely designed van der Waals contacts. By integrating a low work function van der Waals metal electrodes with a p-doped 2D semiconductor, a high Schottky barrier for the majority carrier (holes) and a negligible Schottky barrier for the photoexcited minority carrier (electrons) can be achieved³⁶. Consequently, the high Schottky barrier effectively suppresses the dark current, while the excited minority carriers efficiently contribute to the photocurrent.

a Schematic of efficient and fast photocarrier collection in the vertical device based on 2D materials. The electric field is calculated at V_ds = 0.5 V, neglecting the contact resistance. b Schematic of carrier and photocarrier transport at the semiconductor/metal interface under dark and light conditions. Here, the work function of the metal matches the conduction band of the semiconductor, and the semiconductor is of P-type material c Schematic of the device fabrication process involving van der Waals electrode integration. The 2D material and top metal electrode are transferred using a PDMS and probe-tip-assisted transfer method. d Optical microscopy (OM) image of device #3. Scale bar: 3 μm. e Scanning transmission electron microscopy (STEM) image of the cross-section of device #2. Scale bar: 10 nm. f Energy-dispersive X-ray spectroscopy (EDS) image corresponding to the region in (e). g Semi-log scale current-voltage (IV) curves of device #1 under dark and light conditions. A 458 nm laser was focused on the device with a power of 6.8 μW for light condition. The beam spot size is 1 μm.

Full size image

The fabrication of the device was accomplished entirely via van der Waals staking, specifically employing the probe tip assisted metal film transfer method^37,38. In Fig. 1c, we demonstrate the process of sandwiching the WSe₂ between the flat Ti and Ag van der Waals contacts, which were pre-deposited on mica and SiO₂/Si, respectively. To provide more specific details, the Ti electrode was deposited onto a freshly cleaved mica substrate using the sputtering method, with a metal mask utilized to achieve the desired pattern for the flat bottom electrode. Subsequently, the WSe₂ layer was transferred onto the Ti bottom contact, followed by the transfer of the Ag/Au layer (10 nm/150 nm) onto the WSe₂ layer. For more information on device fabrication, please refer to Supplementary Fig. 1 and the “Method” section.

We present the device geometry and configuration in Fig. 1d–f with the optical microscopy (OM) image, cross-section scanning transmission electron microscopy (STEM) image, and cross-section energy dispersive spectroscopy (EDS) image. The layer configurations: Au, Ag, vdW gap, WSe₂, TiO_x, Ti, and mica, specifically, can be clearly observed in Fig. 1e, f. The 3 nm-thick TiO_x layer in the device was formed due to the oxidization of Ti in the ambient condition during the device fabrication process and adds an additional barrier for carrier transport, the function of which will be discussed in detail below. We fabricated four devices (device #1-#4) with identical Ti/WSe₂/Ag structures and demonstrated the details in Supplementary Fig. 2.

We measured the IV curve of the device under dark conditions and 458 nm laser illumination, as depicted in Fig. 1g. Under dark conditions, the current level is only several pA, however the current can be up to 1 μA under light illumination, resulting in a light on/off ratio of ~10⁶. It is worth mentioning that the photocurrent in our device exhibits a strong linear dependence on the light power density, as shown in Supplementary Fig. 3. This suggests that the photocurrent generation mechanism is related to photocarrier transport. Other photocurrent mechanisms, such as defect trapping induced photogating and photocarrier-induced Schottky barrier changes, which typically exhibit nonlinear photocurrent dependence on light power^33,39, can be ruled out. Moreover, negligible hysteresis was observed even in a strong light power density of 860 mW/cm² (Supplementary Fig. 4), demonstrating minimal photocarrier accumulation in the van der Waals electrode interface. All fabricated devices exhibit low dark current and high photocurrent, as demonstrated in Supplementary Fig. 2, indicating good reproducibility of our device structures. We further conducted tests on various other vertical device configurations of Pt/WSe₂/Ag, Ti/MoTe₂/Pt, and Pt/MoTe₂/Ag that adopted our strategy, and observed low dark current and sensitive photoresponse simultaneously, indicating our strategy is generally applicable for 2D material-based vertical photodetector engineering. However, it is important to note that further comprehensive engineering is necessary to minimize dark current and maximize photocurrent in these structures.

Photocurrent generation mechanism in the Ti/WSe₂/Ag photodetector

The working mechanism of the device was further investigated by analyzing the IV curve under dark and light conditions. Figure 2a illustrates the IV curves of device #1, showing an asymmetric IV curve under dark and relatively symmetric IV curve under light, which suggest the distinct carrier transport mechanisms. Under dark conditions, the IV curve exhibits an asymmetric shape with a relatively small current of 2.4 pA at a bias voltage of 1 V and a relatively larger current of −207 pA at a bias voltage of −1 V. The additional TiO_x layer, formed by the oxidation of Ti, contributes to hole blocking at positive bias voltages (left panel of Fig. 2b, c), resulting in the asymmetric shape of the IV curve under dark conditions. However, under light conditions, the IV curves become relatively symmetrical, indicating a similar photocurrent generation mechanism for both positive and negative bias voltages. The photoexcited electrons can be easily extracted by the electrodes easily owing to the low Schottky barrier for electrons (right panel of Fig. 2b, c). The efficient photocarrier extraction leads to a high photodetection performance. As a result, at a bias voltage of 1 V, a high photocurrent of 1.25 μA is achieved under 458 nm laser illumination. The high responsivity of 0.18 A/W and external quantum efficiency (EQE) of 49% can be calculated with the equation of ({rm{R}}={I}_{{ph}}/P,{rm{and}}; {rm{EQE}}={rm{R}}{hc}/elambda ,) respectively, where R, I_ph, P, h, c, e, and λ are the responsivity, photocurrent, incident power, plank constant, speed of light, elementary charge, and wavelength of light. It is worth noting that while our device achieves fast and efficient photocarrier collection, additional enhancements such as increasing light absorption through the use of optical cavities^40,41 or antireflection coatings⁴² and optimizing the electrode interfaces²⁶ to minimize interface recombination could be explored to further improve responsivity. Additionally, integrating carrier multiplication^43,44 mechanisms present a promising approach to further boost the performance.

a The linear scale current-voltage (IV) curves corresponding to Fig. 1g. b, c Schematic of photocarrier transport in the device under dark conditions (left panel) and light conditions (right panel) at −1 V and 1 V bias voltage, respectively. d The optical microscopy (OM) image (top panel) and schematic of the device structure (bottom panel) for illustrating the position of focused beam illumination. Regions I, II, and III represent the WSe₂ sandwiched between the Ti and Ag electrodes, WSe₂ in contact with the Ag electrode only, and WSe₂ separated from the Ti and Ag electrodes, respectively. e, f The linear and semi-log scale IV curves obtained by illuminating focused light on different regions of the device. The color of the IV curve corresponds to the color of the beam spots in (c). g–j Schematic of photocarrier contribution to the photocurrent when light illuminated to different region of device and different bias voltages of −1 V and 1 V was applied. The transport of photoexcited electron was considered here because the holes cannot contribute to the photocurrent due to the large hole Schottky barrier.

Full size image

We further investigated the photocurrent generation mechanism by illuminating the focused laser on different region I, region II and region III of the device, as illustrated in Fig. 2d. The corresponding IV curves under different light illumination spots are presented in Fig. 2e, f, using both linear and semi-log scales, respectively. When the light is illuminated in region I (WSe₂ sandwiched between the Ti and Ag electrodes), a large photocurrent can be generated under both positive and negative bias voltages. However, if the light is focused on region III (WSe₂ separated from both the Ti and Ag electrodes), the photocurrent decreases by two orders of magnitude due to the negligible electric field in this region. Interestingly, when the light is illuminated on the region II (WSe₂ in contact with the Ag electrode only), the photocurrent exhibits a strong difference depending on positive and negative bias voltage. A substantial photocurrent of 0.4 μA is generated under positive V_ds, decreasing only by one-third compared to the light-illuminated in region I. Despite the weak electric field applied in region II, the efficient extraction of photogenerated electron can be achieved due to the negligible electron Schottky barrier. However, when a negative bias voltage is applied, the photocurrent is only around 18 nA. This relatively low photocarrier efficiency can be attributed to the long distance between the photocarrier generation region and the electrode edge, resulting in inefficient extraction of the photogenerated electrons.

We have summarized the explanation and illustrated the photocurrent generation mechanism under different bias voltages in Fig. 2g–j. When the light is illuminated in region I, the photoexcited electrons can be extracted by either the Ti or Ag electrode depending on the bias voltage. In region II, the photogenerated electrons can only be extracted by the bottom Ag electrode when a positive bias voltage is applied. Under the negative bias voltage, the electron should transport to the Ti electrode for extraction. The relatively small electric field and long transport distance decrease its extraction efficiency. When the light is illuminated in region III, the photoexcited electrons are also difficult to be extracted due to the negligible electric field and long distance to the electrode edge. Therefore, in our device, photocurrent generation primarily occurs in region I and region II (only under positive bias voltage for region II). It is important to note that the photocarrier transport direction is predominantly vertical in our device, with an ultrashort photocarrier transport length.

Photodetection performance characterization of the Ti/WSe₂/Ag photodetector

To evaluate the photodetection performance of our device, we measured the IV curve by illuminating the entire device with a large-size beam at different light powers ranging from 24 μW/cm² to 1.1 W/cm², as shown in Fig. 3a. The photocurrent and responsivity under negative and positive bias voltages were extracted and plotted in Fig. 3b, c, respectively. The photocurrent exhibits a strong linear dependence on the light power density, while the responsivity of the device remains constant across different power levels and various bias voltages. At a bias voltage of 1 V, the device achieves a linear dynamic range of 86 dB, as indicated by Equation LDR = 20 log (P_max/P_min), where P_max (P_min) is the highest (lowest) light intensity. The strong linearity of the photocurrent with respect to light power density is further confirmed by the time-dependent photocurrent measurement, as shown in Supplementary Fig. 6. The strong linear dependence is consistent for different wavelength lasers of 640 nm and 725 nm as shown in Supplementary Fig. 7. Need to note that the photocurrent generation at V_ds = 0 V and the higher photocurrent observed at positive bias voltage in Fig. 3a can be attributed to the photocurrent generation in region II under positive bias voltage, as discussed in Fig. 2h.

a IV curves of device #1 under dark and different light power densities from 24 μW/cm² to 1.1 W/cm². b, c The photocurrent (top panel) and responsivity (bottom panel) under different bias voltages under positive (Fig. 2b) and negative (Fig. 2c) bias voltages. Data extracted from (a). d Time-dependent photocurrent with different wavelengths of light under power density of 70 mW/cm² at V_ds = 0.5 V. e Time-dependent photocurrent under 458 nm laser power density of 24 μW/cm², the power illuminated to the device area is 7.2 pW. f Ultrafast photocurrent measurement of device #1 at V_ds = 1 V. The data were fitted with the exponential decay (red curve), yielding a decay time of 337 ps. g Comparison of the photodetector performance with the previously reported 2D material-based photodetectors^{21,22,24,25,26,32,45,49,50,51}. MSM refers to metal-semiconductor-metals. This comparison does not consider 2D/3D semiconductor heterostructures.

Full size image

The device exhibits excellent photodetection performance across a wide range of wavelengths, spanning from ultraviolet light at 350 nm to near-infrared light up to 1200 nm. In the ultraviolet and visible range, the device achieves a high on/off ratio of above 10⁴. However, as the wavelength increases into the near-infrared range, the WSe₂ absorption decreases significantly due to weak indirect band absorption. For wavelengths longer than 1000 nm, the WSe₂ absorption becomes negligible, and the photocurrent can be ascribed to the transfer of hot electrons from the metal electrode to the WSe₂ conduction band^45,46. The hot electrons with high kinetic energy are excited from the metal surface and overcome the Schottky barrier to inject into the WSe₂. Consequently, even in the absence of photo-induced exciton generation, the generation of photocurrent remains feasible by infrared light. The responsivity at different wavelengths was measured using photocurrent spectroscopy and was demonstrated in Supplementary Fig. 8. Taking advantage of the low dark current under positive bias voltage, a clear photocurrent of 1 pA can be observed from a dark current of 0.8 pA when exposed to 7.2 pW incident light on the device. The detectivity of the device can be calculated as 5 × 10¹² Jones using the formula ({D}^{* }={{RA}}^{1/2}/{(2e{I}_{{dark}})}^{1/2}), where R is the responsivity, A is the active area, e is the elementary charge, and I_{_dark} is the dark current.

We conducted continuous wave (CW) laser-based photocurrent response speed measurements by amplifying the photocurrent with a current preamplifier and recording the data using an oscilloscope as mentioned in previous literature^12,13 and defined it as the actual response time of our device. The rise/fall time obtained from the response to increase/decrease from 10% to 90%/90% to 10% of maximum photocurrent reveals fast response time (actual response time) of 1.2/0.8 μs, as shown in Supplementary Fig. 9. It is important to note that the bandwidth of our current preamplifier limits this response time. To get the intrinsic high response speed (intrinsic response time) of this device, we performed ultrafast photocurrent measurements of our device using a home-built pump-probe-based photocurrent measurement setup³⁵. The ultrafast response time of 337 ps was measured under the bias voltage of 1 V, as demonstrated in Fig. 3e. This value is consistent with the previous paper where ultrafast photocurrent was also measured in a similar vertical structure based on WSe₂⁴⁷.

We compared the photodetection performance of our device, specifically the dark current, response speed, and responsivity, with previous results and presented the comparison in Fig. 3h. Generally, photodetectors based on 2D materials with metal-semiconductor-metal (MSM) structures exhibit high responsivity but suffer from slow response speeds due to the defect-assisted photogating effect. However, the vertical device design in our device minimizes this effect through the strong electric field and fast photocarrier extraction, allowing for faster response speeds and a high dynamic range. Despite the simple MSM configuration, our device outperforms many heterostructure-based devices in terms of response speed and dark current, while achieving comparable photoresponsivity. The photoresponsivity, response speed, and dark current are competitive with those of commercial silicon photodiodes, specifically FDS015 in Thorlabs. Need to note that the data point with red color edge in Fig. 3f, was obtained from the pump-probe-based ultrafast photocurrent measurement.

Robust photocurrent under high temperature and bending conditions

We have demonstrated the robustness of the photocurrent in our device under bending conditions in Fig. 4a, b. The device was fabricated on a flexible mica substrate which have the thickness of ~10 μm and the mica substrate was affixed to a scotch tape to enable bending (Fig. 4a). The overall thickness of the device, including the electrodes, is ~200 nm. Therefore, even with strong bending of the substrate with a 3 mm radius, minimal strain is applied to the device and the channel material⁴². The IV curve was measured under dark and light illumination in the flat condition, bending condition, and after 1000 bending cycles, as shown in Fig. 4b. Notably, no significant changes are observed in the IV curve for both dark and light illumination, indicating the device’s resilience to bending. We further assessed the working stability by monitoring the I_ds with light on/off at a frequency of 0.05 Hz. As depicted in Fig. 4c, the dark current and photocurrent maintain their initial values consistently under bending conditions and even after 1000 bending cycles. Negligible changes are observed over a measurement period exceeding 20 min at a bias voltage of 1 V, showcasing the high working stability of our device even in ultrashort channel. This excellent photocurrent reliability is attributed to efficient and robust photocarrier transport within the ultrashort channel and strong electric fields. Additionally, the channel material is sandwiched between the two electrodes, which also serves as a protective layer, minimizing the influence from the surrounding atmosphere.

a Optical microscope (OM) image and schematic of device #3 for bending measurement. b IV curve of dark and light conditions (458 nm light, 50 mW/cm²) at flat, bending (bending radius of 3 mm), and after 1000 times bending with a bending radius of 6 mm. c Time-dependent photocurrent at V_ds = 0.5 V under the same bending conditions as shown in (b). d IV curve under dark and light conditions (458 nm light, 500 mW/cm²) at different temperatures ranging from 24 °C to 100 °C. e Time-dependent photocurrent at V_ds = 1 V under different temperatures of 24 °C, 50 °C, and 100 °C, respectively. f Temperature-dependent dark current of our device under a different bias voltage of 0.4 V, 0.6 V, and 1 V.

Full size image

The device exhibits excellent working stability even at high temperatures. Figure 4d shows the IV curve under dark and light illumination at different temperatures ranging from 24 °C to 100 °C. A noticeable distinction in current between dark and light conditions is evident, with an on/off ratio exceeding 10³. To further showcase the device’s ability to function reliably in high-temperature environments, we conducted time-dependent photocurrent measurements at 100 °C, as illustrated in Fig. 4e. The results exhibit consistent operation even under such elevated temperatures. Notably, the dark current remains below 100 pA at 100 °C across various bias voltages, affirming that our device is well-suited for high-temperature photodetection.

While our device demonstrates excellent overall performance, it is important to address a critical limitation. The transferred contact generally exhibits poor long-term stability, and it is very sensitive to temperature change, which makes our device unstable against high temperature annealing (Supplementary Fig. 11). Therefore, it is crucial to explore alternative van der Waals electrode fabrication techniques that can achieve reliable van der Waals contacts for practical applications. Given that stable van der Waals contacts are a common objective of the 2D device community, we anticipate that this issue can be resolved in the near future. In Supplementary Table S1 and Supplementary Fig. 12, we provide a comprehensive comparison of photodetectors fabricated using various materials, including silicon, 2D materials, organic-inorganic hybrid perovskite materials, organic materials, and quantum dots. We assess the devices based on their response speed, detectivity, responsivity, stability, and flexibility. The 2D materials-based photodetector exhibits favorable properties in all five aspects, while the Si photodiode poses challenges for flexible applications, and both the organic photodetector and perovskite photodetector suffer from stability drawbacks. The quantum dot-based photodetector demonstrates its weakness in device performance such as the response speed and detectivities.

We have successfully demonstrated a practical approach to construct a high-performance flexible photodetector using 2D materials. By integrating van der Waals contacts along the vertical direction of the 2D material, we have achieved a high Schottky barrier for majority carriers. The Ti/WSe₂/Ag vertical device configuration has enabled us to achieve a low dark current of 0.8 pA and a high external quantum efficiency of 49%, thanks to efficient photocarrier extraction within a short carrier lifetime of 337 ps. The device exhibits a large linear dynamic range of 86 dB and a high detectivity of 5 × 10¹² Jones. Moreover, we have demonstrated the robustness of the device’s photocurrent and dark current under high-temperature and bending conditions. The superior photodetection performance of our device is attributed to the engineered carrier/photocarrier transport in the ultrashort channel along the vertical direction of 2D materials. The concept presented in our approach holds great potential for advancing 2D material-based photodetection technologies and contributing to the development of next-generation flexible photodetectors.

Device fabrication

Ag/Au thin metal films with 10 nm and 150 nm thicknesses are deposited onto a SiO₂/Si substrate using thermal deposition. These metal films are then cut with a probe tip³⁸ to create patterns for the bottom electrodes. The patterned metal films are transferred onto a mica substrate, followed by the deposition of bottom electrodes (Ti or Pt) with sputtering deposition method. Afterward, the metal masks are removed from the mica substrate. A WSe₂ flake is transferred onto the bottom Ti electrode with the assistance of PDMS, and an Ag/Au electrode is transferred on top of the WSe₂ flake to serve as the top electrode. All transfer processes are conducted under ambient conditions. The same device fabrication process is carried out for the Pt/WSe₂/Ag, Pt/MoTe₂/Ag, and Ti/MoTe₂/Pt devices. For the flexible device, the mica substrate is exfoliated to a thin substrate with a thickness of 10 μm. It is then fixed securely onto scotch tape to enable bending. The mica substrate was attached to a metal rod with a 3 mm radius to conduct the bending test.

Device characterization

All photocurrent measurements were carried out in ambient conditions using a home-built probe station equipped on the inverted optical microscope set up, allowing for bottom-side light illumination. Power-dependent photocurrent measurements were performed using CW lasers with wavelengths of 458 nm, 640 nm, and 750 nm. Wavelength-dependent photocurrent measurements were conducted using white light (EQ-99X-FC) and bandpass filters (FKB-VIS-10 and FKB-IR-10, Thorlabs). A home-built photocurrent setup⁴⁸ was utilized for photocurrent spectroscopy measurements. A mechanical flip stage was employed to control the on/off switching of the light, while a source meter unit (Keithley 2636B) was used for applying the bias voltage and recording the current. The thickness measurement was performed using an atomic force microscope (AFM5000II, Hitachi). To investigate temperature-dependent photocurrent, a home-built temperature control stage was utilized. Ultrafast photocurrent measurements were carried out using a pump-probe photocurrent setup. A femtosecond pulse laser with a duration of 13 fs (Coherent Vitara-T, 80 MHz repetition rate) operating at a central wavelength of 800 nm was utilized in a pump-probe configuration. The laser beam was split into two beams using a 50/50 beamsplitter to form the pump and probe beams. To avoid interference effects, the polarization direction of the pump beam was adjusted to be perpendicular to that of the probe beam using a half-waveplate. The delay time of the probe beam was controlled by a mechanical delay stage (Newport UTS150CC and ESP301 controller). The probe beam was modulated with a frequency of 100 Hz using a mechanical chopper. The pump and probe beams were then recombined using another beamsplitter and focused onto the device under study through a microscope objective (Mitutoyo, MY50X-825). The resulting photocurrent generated in the device was measured with a lock-in amplifier (SR830) after the amplification with a current preamplifier (DLPCA-200, FEMTO).