Introduction

Photodetection technology is critical to many applications including optical communication, imaging, and information processing. Among the various material systems explored, two-dimensional (2D) layered materials have expanded the design freedom and functionality of photodetectors, thanks to their unique electronic and optical properties and tunable band structures1,2. One challenge for using atomic-thin 2D materials for photodetection applications is low light absorption3,4,5. Several approaches have been explored to enhance light-matter interaction in 2D materials by utilizing structures like photonic crystals, plasmonic, and nanocavity resonators6,7,8,9. However, these approaches often rely on complex nanostructure that raises concerns about cost and scalability in large-area fabrication10,11.

In contrast, optical cavity engineering utilizing Fabry–Pérot (F-P) cavities, particularly asymmetric F-P cavity, offers a relatively simple and scalable method with uniformly enhanced absorption in 2D materials. Asymmetric F-P interference has been demonstrated to significantly boost the optical gain in MoS2, yielding photoluminescence (PL) enhancement of up to 22 %10. Moreover, the asymmetric cavity designs have demonstrated versatility for a range of emerging technologies, including color printing, flexible displays, wearable electronics, and other optoelectronic applications12,13. Typically, such optical cavities consist of 2D material on a dielectric spacer and a reflective mirror, forming asymmetric F-P interference. By optimizing the dielectric thickness, significant absorption enhancement in 2D materials at specific wavelengths can be achieved to boost the photoresponse of the photodetectors14. Despite the potentials, only a limited number of studies have employed cavity substrates to boost photodetection performance, and most of them utilize hexagonal boron nitride (h-BN) as the dielectric layer15,16,17,18,19,20, which poses challenges for precise thickness control and device reproducibility.

Van der Waals (vdWs) heterostructures enable precise control of band alignment and optoelectronic properties at the atomic scale21,22,23,24,25,26,27. The improved charge separation and suppressed recombination at the heterojunction of vdWs heterostructure can elevate the performance of the photodetector28,29,30,31,32. For example, the broken-gap SnSe2/MoTe2 heterostructure exhibited multifunctional behavior, including Esaki and backward diode operation, as well as a gate-tunable photovoltaic response33. The WSe2/SnSe2 heterostructure photodetector achieved high responsivity of 588 AW−1 due to band-to-band tunneling enabled by its near-broken to type-III band alignment34. In particular, type II band alignment with large band offset is highly desirable to facilitate the extraction of photogenerated charge carriers35,36. For instance, the strong interlayer coupling and large type-II band offset in the SnS2/MoS2 heterostructure enabled efficient charge separation and transport, resulting in a high photoresponsivity of 1.36 AW−1 35. Similarly, the large band offset in the MoS2/SnSe2 heterostructure enabled efficient carrier separation, yielding a high responsivity of 2.3 AW−1. Additionally, doping engineering introduced electronic states that allowed solar-blind ultraviolet detection with a high UV-to-visible rejection ratio of 250, a responsivity of 0.5 AW−1, and a response time of 0.15 ms36.

Here, we report a MoS2/SnSe2 heterostructure photodetector integrated with an asymmetric F-P cavity comprising an aluminum (Al) mirror and aluminum oxide (Al2O3) dielectric spacer. This architecture enables synergistic enhancement via both cavity-assisted light absorption and favorable band alignment with large band offset for more efficient extraction of the photogenerated carriers. The device exhibits exceptional photoresponse enhancement across a broad spectral range from visible to near-infrared (λ = 450-900 nm). At 450 nm wavelength and Vds = 2 V, the photodetector achieves ultrahigh responsivity (4.7 × 103 AW−1), external quantum efficiency (EQE) (1.3 × 106 %), and specific detectivity (1.8 × 1014 Jones), outperforming previously reported values. It also maintains a fast response time of 150 ms, comparable to a similar MoS2/SnSe2 device on SiO2 substrates but with dramatically enhanced sensitivity. Notably, the device can operate under zero bias (i.e., at Vds = 0 V) due to the built-in electric field at the heterojunction. Our results shed new light on combining large band-offset 2D heterostructures with cavity engineering for high-performance photodetection and energy-efficient optoelectronic devices.

Results

Cavity design for enhanced absorption

The cavity is engineered with Al as the back reflector and Al2O3 as the dielectric material. The optimal thickness of the Al2O3 layer was determined using the finite difference time domain (FDTD) method to achieve the maximum absorption at the heterostructure (Fig. 1a). The simulations revealed that a very thin layer of Al2O3 (∼10–40 nm) yields maximum absorption in the wavelength range of 300–500 nm. However, to achieve a consistently enhanced absorption across a broader spectral range of 300–900 nm, the optimized Al2O3 thickness was determined to be 50 nm. Figure 1b presents a schematic of the heterostructure photodetector fabricated on a cavity consisting of a 100 nm thick Al mirror and a 50 nm thick Al2O3. MoS2 and SnSe2 were selected for the heterostructure for their type II band alignment and large band offset, which together promote efficient extraction of photogenerated charge carriers35,36. Figure 1c illustrates the energy band structures of individual MoS2 and SnSe2 layers before and after contact. The Fermi level of SnSe2 lies ~87 meV below its conduction band minimum (CBM), while that of MoS₂ is positioned about 0.4 eV below its CBM37,38.

Fig. 1: Cavity design and simulation.
Fig. 1: Cavity design and simulation.
Full size image

a Al2O3 thickness-dependent absorption enhancement of MoS2/SnSe2 heterostructure. b Schematic of the cavity and the heterostructure photodetector at the optimized thickness of Al2O3 (50 nm). c The schematic of type II energy band alignment in the MoS2/SnSe2 heterostructure before and after contact, showing the large band offset. d FDTD simulation for the reflectance spectra of the heterostructure for monolayer MoS2 and different thicknesses of SnSe2. e FDTD simulation for the reflectance spectra of the heterostructure for thin SnSe2 and different thicknesses of MoS2. f Propagation path of light using cavity substrate and SiO2/Si substrate.

To investigate the impact of MoS2 and SnSe2 thickness on the reflection spectra of the heterostructure, FDTD simulations were performed (Fig. 1d, e). The results show that increasing the SnSe2 thickness beyond 6.5 nm diminishes the cavity-induced absorption enhancement due to weaker optical coupling. On the other hand, increasing MoS2 thickness up to 4 nm slightly improves the absorption. These trends indicate that SnSe2 thickness is the limiting factor, and it should be minimized for optimal performance. As a result, a heterostructure comprising a monolayer of MoS2 and a few layers of SnSe2 (around 3-4 layers) was chosen to maximize the absorption.

The mechanism behind the enhanced absorption utilizing the cavity substrate is elucidated in Fig. 1f. The inclusion of the Al film enhances light-matter interaction in the MoS2/SnSe2 heterostructure by forming an asymmetric F-P cavity, effectively extending the optical path length for incident light within the heterostructure. As a result, light absorption is significantly increased across a broad spectral range. In contrast, conventional SiO2/Si substrates allow a substantial portion of the incident light to transmit through the SiO2 layer and be absorbed by the silicon substrate, resulting in reduced absorption efficiency for the MoS2/SnSe2 layer. This is evident from the PL and reflectance measurements of individual materials and the heterostructure on both the optical cavity and SiO2/Si substrates.

Optical characteristics and photoresponse enhancement

Figure 2 presents the PL spectra of the MoS2/SnSe2 heterostructure and the individual materials on the SiO2 and the cavity substrate. Firstly, we compared the PL spectra of monolayer (1 L) MoS2 on SiO2 and cavity substrates under identical excitation conditions (25 µW, 532 nm laser), as shown in Fig. 2a. MoS2 exhibits two main PL peaks near 670 nm and 620 nm, corresponding to the A and B excitons, respectively39. The PL intensity of 1 L MoS2 on the cavity substrate is more than twice that observed on the SiO2 substrate, indicating the significant enhancement in optical absorption enabled by the cavity structure. SnSe2, an indirect bandgap semiconductor in both monolayer and bulk forms, with bandgaps of ~1.69 and 1.07 eV, respectively, naturally exhibits weak PL40. However, when few-layer (FL) SnSe2 is placed on the cavity substrate, a broad PL emission centered around 615 nm becomes apparent even at low excitation power (10 µW), as shown in Fig. 2b. This enhancement, absent in SnSe2 on SiO2 even at higher excitation power (25 W), as shown in Fig. 2c, underscores the role of the cavity in boosting optical absorption. Interestingly, the PL spectrum of the MoS2/SnSe2 heterostructure on the cavity resembles that of 1 L MoS₂ alone, likely influenced by charge transfer at the type II interface and non-radiative energy transfer (NRET) within the heterostructure (Fig. 2b). The latter effect has been reported to enhance the PL when SnSe2 thickness is ≤10 nm41. In this study, the thickness of 3-4 L SnSe2 (~2.5–3.5 nm) was selected to optimize both optical coupling and cavity absorption. In comparison, Fig. 2c shows the PL spectra of the individual materials and the heterostructure on the SiO2 substrate, where PL quenching at the heterostructure is attributed to the thicker SnSe2 (21 nm)41, as indicated in Supplementary Fig. S1 in the Supporting Information. With increasing SnSe2 thickness, the spectral overlap and dipole coupling between SnSe2 absorption and MoS2 excitons are reduced, leading to progressive PL quenching)36,41. When the SnSe2 thickness increases further (e.g., to 43 nm), the PL quenching becomes even more pronounced42. The reduced NRET (or Förster resonance energy transfer) efficiency arises from diminished spectral overlap and increased interlayer separation, which weaken dipole–dipole coupling and limit energy transfer to radiative recombination in MoS241.

Fig. 2: PL and reflectance spectra.
Fig. 2: PL and reflectance spectra.
Full size image

a Comparison of PL spectra of monolayer (1 L) MoS2 on SiO2 and cavity substrates under identical excitation (25 µW, 532 nm laser). Insets are the optical microscopic images of the samples. The scale bars correspond to 10 µm. b PL Spectra of the 1 L MoS2, FL SnSe2 (3 L∼2.5 nm), and their heterostructure on cavity substrate (excitation power of 10 µW). Insets are the OM image of the heterostructure and enlarged PL spectrum of FL SnSe2. The scalebar is 10 µm. c PL Spectra of the 1 L MoS2, FL SnSe2 (~21 nm), and their heterostructure on SiO2 substrate (excitation power of 25 µW). Insets are the OM image of the heterostructure and enlarged PL spectrum of FL SnSe2. The scalebar is 10 µm. d The measured reflectance spectra of the parent materials and heterostructure on the cavity substrate compared to the simulated reflectance spectra of the heterostructure.

Reflectance measurements further substantiate the cavity-enhanced absorption. Figure 2d shows the reflectance spectra of the individual materials and the MoS2/SnSe2 heterostructure on the cavity substrate. As expected, all exhibit significantly reduced reflectance, indicating enhanced absorption compared to their counterparts on the conventional SiO2 substrate (Supplementary Fig. S1c, d, Supporting Information). The experimental reflectance spectra also show good agreement with the FDTD simulations, validating the cavity design. The minor deviation at the A and B exciton peaks of MoS2 in the simulated reflectance is likely due to differences between the actual and referenced refractive index data used for simulations43. In addition, the consistently lower reflectance values predicted by the model, compared to experiment, are commonly observed in 2D-material-based F-P cavities14,18. These discrepancies mainly arise from scattering induced by surface roughness in transferred layers and slight variations in material optical properties that are not accounted for in ideal simulations.

The photoresponse of 1 L MoS2/4 L SnSe2 heterostructure photodetector integrated with cavity substrate was characterized across the visible to near-infrared range (λ = 450–900 nm) using a supercontinuum laser source (see methods section for details). Figure 3a, b presents the current-voltage (I-V) characteristics at 450 and 800 nm wavelengths as a function of laser powers, respectively. The photodetector exhibits a strong photoresponse across a range of illumination powers due to the enhanced light absorption enabled by the cavity structure. Full I-V characteristics across different wavelengths and powers, along with the corresponding photocurrents, are provided in Supplementary Figs. S2 and S3 of the Supporting Information. Figure 3c illustrates the energy band diagram of the heterostructure photodetector under illumination. The large band offset inherent to the type II alignment facilitates efficient charge separation, where the photoexcited electrons in the conduction band of MoS2 are transferred to SnSe2 while holes move from the valence band of SnSe2 to MoS2. This promotes efficient carrier extraction and contributes to the device’s high photoresponse efficiency.

Fig. 3: Photoresponse of the MoS2/SnSe2 heterostructure photodetector on cavity substrate.
Fig. 3: Photoresponse of the MoS2/SnSe2 heterostructure photodetector on cavity substrate.
Full size image

The power-dependent I-V characteristics at λ = 450 nm (a), and λ = 800 nm (b). Inset in (a) is the optical microscopic image of the device. The scalebar corresponds to 10 µm. c The energy band diagram of the MoS2/SnSe2 heterostructure under illumination. d Normalized temporal photoresponse of the device at different illumination wavelengths (Vds = 2 V) and the corresponding response time extracted at each wavelength. e Temporal photoresponse under different illumination wavelengths and powers at Vds = 0 V.

The dynamic photoresponse of the device under 1.0 Hz laser modulation and 2 V bias across various wavelengths is shown in Fig. 3d. The photodetector exhibits consistent response time in milliseconds at all measured wavelengths. Despite the typical trade-off between responsivity and response speed, the device maintains a fast response alongside its ultrahigh responsivity (to be discussed later). In comparison, a similar device on a SiO2 substrate exhibits comparable speed but much lower responsivity, underscoring the critical role of the cavity substrate42. Moreover, the photodetector exhibits a photovoltaic response (Vds= 0 V) with excellent switching behavior at different illumination powers and wavelengths (Fig. 3e), driven by the built-in electric field at the heterojunction. However, no photovoltaic response is observed beyond 700 nm (the band edge of MoS2). This is due to light being predominantly absorbed by the SnSe2 layer, while the large band offset hinders efficient hole transfer to MoS2. In addition, the built-in electric field is insufficient to separate carriers, resulting in dominant recombination under zero bias. To further validate this architecture, a photodetector of 1 L MoS2/3 L SnSe2 was characterized as shown in Supplementary Figs. S4–S6 in the Supporting Information. The device exhibited similar behavior, confirming that the cavity enhances both absorption and carrier extraction in bandgap-engineered heterostructures.

Figure 4a–c show the power-dependent responsivity (R), external quantum efficiency (EQE), and detectivity (D*) of the photodetector under a bias of 2 V. As expected, the photodetector demonstrates ultrahigh responsivities across the entire wavelength range, with a peak value of 4.7 × 103 AW−1 at λ = 450 nm and extremely low laser power of 14 nW. This is accompanied by an enhancement of the EQE, peaking at 1.3 × 106 % at 450 nm. The observed ultra-high EQE arises from a strong photoconductive gain (G) mechanism, where G = τlifetime / τtransit. Here, τlifetime denotes the extended carrier lifetime, and τtransit is the channel transit time (τtransit = L²/ (μ V), where L is channel length, μ is the mobility, and V is the applied bias voltage). Under bias, the staggered alignment plays a pivotal role in enabling this gain, where photoexcited electrons from the valence band of MoS2 transfer to the conduction band of SnSe2, while holes remain in MoS₂, promoting efficient spatial separation and reducing recombination. This process is further enhanced by the increased optical absorption enabled by the asymmetric F-P cavity, which raises the initial photogeneration rate and synergistically amplifies the effective gain, leading to the observed ultrahigh EQE.

Fig. 4: Performance metrics of the MoS2/SnSe2 heterostructure photodetector on cavity substrate.
Fig. 4: Performance metrics of the MoS2/SnSe2 heterostructure photodetector on cavity substrate.
Full size image

The power-dependent responsivity (R) (a), external quantum efficiency (EQE) (b), and detectivity (D*) (c), of the device under different illumination wavelengths in photoconductive mode (Vds = 2 V). The power-dependent R (d), and EQE (e) under different illumination wavelengths in photovoltaic mode (Vds = 0 V). (f) Comparison of the R, D*, and response time (τ) of the SnSe2/MoS2 photodetector on a cavity substrate with other reported 2D-material-based photodetectors with optical cavities.

The detectivity of the photodetector was calculated using Eqs. (5) and (6) (methods section). The root-mean-square noise current (ISD) is calculated to be 1.34 × 10−11 AHz−1/2 (Supplementary Fig. S7a, Supporting Information). A maximum detectivity of 1.8 × 1014 Jones was recorded at 450 nm (Fig. 4c). Performance metrics for the photodetector with 3 L SnSe2 are presented in Supplementary Fig. S6. At λ=450 nm and laser power of 0.2 µW (Vds = 1 V), the device achieved a R of 8.3 × 102 AW−1, EQE of 2.3 × 105 %, and a D* of 1.5 × 1013 Jones. Comparatively, the heterostructure photodetector with 4 L SnSe2 exhibited a R of 1.5 × 103 AW−1, EQE of 4.2 × 105 %, and D* of 5.8 × 1014 Jones under identical conditions (i.e., illumination power of 0.2 µW and applied bias of 1 V). Both devices showed consistent performance, with only minor variations attributable to flake thickness and contact quality. Importantly, cavity-assisted absorption enhancement was observed in both cases, confirming the reproducibility and robustness of the synergistic design. Besides, the photodetector also delivers notable photovoltaic performance under zero bias (Vds = 0 V), with R and EQE values on par with leading self-powered photodetectors reported in literature44,45,46,47 due to the high built-in potential at the heterojunction. At 450 nm (Vds = 0 V), it achieves a R of 43.6 mAW−1 and EQE of 12 % (Fig. 4d, e). Supplementary Fig. S7b compares responsivity with and without the optical cavity across 450–900 nm, alongside reported cavity-based photodetectors. The enhancement in responsivity and detectivity from cavity-assisted absorption is evident across all wavelengths (Supplementary Fig. S7c, Supporting Information). These findings highlight the crucial role of optical cavity design and band alignment in optimizing TMD heterostructure photodetector performance.

Beyond the enhanced performance of our photodetector on cavity substrate compared to that on SiO2/Si substrate, it also outperforms previously reported 2D-material-based cavity photodetectors in terms of responsivity (Fig. 4f and Table 1). The significant improvement across all parameters is attributed to the asymmetric F-P cavity, which effectively extends the optical path length and boosts light absorption within the heterostructure (Fig. 1f). Moreover, the detectivity rivals the highest reported value of a MoS2/NiO/Au photodetector (responsivity of 703 AW−1)18 This aligns with the enhanced absorption of our designed cavity, particularly around 450 nm. To further emphasize the significance of our design, we compare the performance of the MoS2/SnSe2 heterostructure photodetector with conventional semiconductor devices. At λ = 450 nm, the photodetector exhibits ultrahigh responsivity of 4.7 × 10³ AW−1, substantially surpassing commercial silicon photodiodes, which typically exhibit a responsivity of ~0.4–0.5 AW−1 at similar wavelengths48. While commercial devices achieve faster response times in the nanosecond range, their low responsivity and detectivity at low light levels highlight the advantages of the cavity-assisted 2D heterostructure approach.

Table 1 Comparison of the performance of 2D-material-based cavity photodetectors
Full size table

In terms of response speed, our SnSe2/MoS2 photodetector is faster than other 2D-material-based cavity photodetectors owing to efficient charge carrier extraction at the type II heterostructure with a large band offset. However, its speed is slower than that of commercial silicon photodetectors, which typically operate in the nanosecond range48. The measured response speed of 150 ms reflects a balance between ultrahigh responsivity and carrier lifetime. While the long carrier lifetime contributes significantly to the photoconductive gain and hence the extremely high EQE observed, it also limits the ultimate response speed. In addition to carrier lifetime, transit time effects, which depend on charge carrier mobility, channel length, and applied bias, can influence the speed. EQE exceeding 100% in our device is primarily a result of this photoconductive gain mechanism, combined with enhanced optical absorption from the cavity. These factors collectively explain why the photodetector maintains millisecond-scale response while achieving record-high responsivity and detectivity. To improve response time, interdigitated electrodes could be employed to shorten carrier transport paths, though this would reduce the responsivity due to the inherent trade-off between responsivity and speed. Ultimately, photodetector design could be tailored to its target application. This underscores the significance of asymmetric F-P cavity combined with strategic band alignment in 2D heterostructures as a powerful and industry-compatible method to enhance photodetection performance. This approach supports large-area fabrication with uniform absorption enhancement and simple processing, making it suitable for practical implementation.

Discussion

In summary, we have successfully demonstrated a high-performance MoS2/SnSe2 photodetector by strategically integrating a bandgap-engineered 2D heterostructure with an asymmetric Fabry–Pérot (F-P) cavity substrate to overcome the intrinsic absorption limitations of atomically thin materials. The photodetector achieved unprecedented photoresponse enhancement, enabled by type II band alignment along with large band offset, which facilitates efficient charge separation and photocarrier extraction, and cavity-assisted absorption, which extends the optical path length of incident light and boosts light-matter interaction. The photodetector achieves high performance, with a responsivity of 4.7 × 103 AW−1, external quantum efficiency of 1.3 × 106 %, and detectivity of 1.8 × 1014 Jones, surpassing previously reported values. It also maintains a fast response speed of 150 ms, striking a favorable balance between speed and responsivity, and operates efficiently under zero bias due to the built-in potential at the heterojunction. Our approach is scalable and industry-compatible, offers high uniformity and integration potential with existing photonic platforms, paving the way for next-generation optoelectronic devices.

Methods

Finite Difference Time Domain (FDTD) simulations

To evaluate the optimized back Al reflector / Al2O3 dielectric system integrated with the MoS2/SnSe2 heterostructure, numerical simulations were performed using the Lumerical FDTD method. The Al/Al2O3 system with 50 nm Al2O3 was determined to be the most effective in enhancing absorption over a broader range. To eliminate undesired reflection from the boundaries, 12 perfect matching layers were utilized, and periodic conditions were employed in the x and y directions. The contact area between the film and substrate was simulated by using the light source at the substrate, resulting in the exclusion of reflection from the bottom surface of the substrate in our simulation. Therefore, the simulated reflection was the ratio of the reflected power recorded by the detector to the total power of the incident light source. To ensure that no evanescent wave was involved in the calculated spectra, the detector was positioned 0.5 μm away from the top surface of the substrate in our model. A fine grid of 0.25 nm was utilized along the z-direction within the 2D films to minimize simulation errors.

Cavity fabrication

Firstly, the Al mirror (5 nm Cr/100 nm Al) was deposited on a pre-cleaned sapphire substrate using an electron beam evaporator (Denton Vacuum Explorer). Next, atomic layer deposition (ALD) using a Beneq TFS 200 system with H2O and trimethylaluminum (TMA) as precursors was utilized for the 50 nm Al2O3 layer deposition.

Optical characterization

The PL spectra of the samples were measured using Alpha 300S (WITec Gmbh) with 50x objective under laser excitation wavelength of 532 nm. The reflectance measurements at room temperature were performed using CRAIC 20/30PV. For each measurement, 50 spectra were averaged. When measuring the reflectance of the heterostructure on cavity substrate, the bare cavity substrate was used as the reference. While for samples on SiO2 substrate, the reflectance spectra were normalized against those of bare SiO2 substrate with an Al mirror serving as the reference. The differential reflectance (\(\frac{\varDelta R}{R}\)) was calculated according to the following equation49,50:

$$\frac{\varDelta R}{R}=\frac{{R}_{f}\,-\,{R}_{s}}{{R}_{s}}=\frac{4n}{{n}_{0}^{2}-1}\alpha (\lambda )$$
(1)

where Rf represents the reflectance spectrum of the sample on the substrate, and Rs is the reflectance spectrum of the bare substrate. The n and n0 are the refractive indices of the 2D flakes and substrate, respectively.

Device fabrication

The SnSe2, and MoS2 were exfoliated using polydimethylsiloxane (PDMS) from bulk crystals purchased from HQ Graphene. The number of layers for MoS2 and SnSe2 was identified by optical contrast using a calibrated optical microscope. The SnSe2 flakes used for cavity-integrated devices correspond to 3–4 layers (~2.5–3.5 nm), while the MoS2 layer was monolayer in all cases. For the heterostructure on the SiO2 substrate, the SnSe2 thickness was measured by atomic-force microscopy to be ~21 nm, as shown in Supplementary Fig. S1 of the Supporting Information. The dry transfer was employed to assemble the heterostructures. The standard photolithography process was first used to write the metal contact pads, followed by evaporation of Cr/Au (10/100 nm in thickness) and lift-off. The assembled heterostructures were annealed for 3 h in a quartz chamber at 150 °C under a flow of Ar gas (200 sccm) to improve the interlayer coupling. Afterward, another photolithography process is performed for electrode patterning, followed by Au deposition (100 nm thick) and lift-off. Finally, the devices were fixed on chip carriers and wire bonded for photoresponse measurements. To minimize any potential environmental effect, the fabricated heterostructure devices were stored in a glove box and only exposed to air during optical and electrical measurements. Under this protocol, no performance degradation was observed during experiments.

Photoresponse measurements

To measure the broadband photoresponse ranging from λ = 450–900 nm, a supercontinuum laser source (SuperK EXTREME, NKT Photonics) and multi-wavelength filters (SuperK SELECT, NKT Photonics) were used. The laser spot diameter illuminating the devices is ~150 µm for the wavelength range of 450–650 nm and ~300 µm for the range of 700–900 nm. The overlapping areas of the heterojunctions are 44 µm² for device 1 with 4 L SnSe2 and 70.11 µm² for device 2 with 3 L SnSe2. To avoid overestimation of the responsivity for the heterojunction photodetectors, the calculations account for the entire area of the 2D semiconductor materials (∼80 µm2 for device 1 and 192 µm2 for device 2), rather than just the overlapping junction area. A Keithley 2450 source meter and an optical shutter were used to collect the current and modulate the laser beam, respectively. The responsivity (R), external quantum efficiency (EQE), and detectivity (D*) were calculated as:

$$R=\frac{{I}_{{ph}}}{{P}_{{device}}}$$
(2)
$${P}_{{device}}=\frac{{P}_{{in}}{A}_{{device}}}{{A}_{{laser}}}$$
(3)
$${EQE}=\frac{h\,c\,R}{e\lambda }$$
(4)

where the photocurrent (Iph) = Ilight - Idark, Pdevice is the effective laser power on the device area. Pin, Adevice, and Alaser are the incident laser power, the effective sample area, and the laser spot area, respectively.

Assuming that the shot noise is the dominant noise of the photodetector, the detectivity can be calculated as20,36:

$${D}^{* }\approx \frac{R\sqrt{{A}_{{device}}}}{\sqrt{{2{eI}}_{{dark}}}}$$
(5)

It can also be derived from noise spectrum measurements as51:

$${D}^{* }=\frac{R\,\sqrt{{B\,A}_{{device}}}}{\sqrt{{\left\langle {i}_{n}\right\rangle }^{2}}}$$
(6)

where, \({\left\langle {i}_{n}\right\rangle }^{2}\) is the mean square noise current and B is the bandwidth (15 KHz in our setup). The estimated detectivity values of our photodetector using both Eqs. (5) and (6) are comparable, indicating that the shot noise is dominated. The response speed was determined by fitting the rise and fall times under modulated illumination to an exponential function (time required for the signal to rise from 10% to 90% and fall from 90% to 10%, respectively).