Introduction

Unipolar barrier infrared detectors, such as nBn (n-type contact layer, barrier layer, and n-type absorber) and pBp (p-type contact layer, barrier layer, and p-type absorber) architectures, are emerging as promising solutions for effectively suppressing dark current and enabling high-operating-temperature (HOT)1,2,3. These structures can enhance the collection efficiency of photogenerated carriers while mitigating dark current generation, without inhibiting the flow of photocurrent. The so-called “unipolar barrier” refers to a barrier that blocks one carrier type (electrons or holes) while allowing the other type to flow unimpeded1,2. In the case of an nBn detector, its structure consists of an n-type semiconductor on one side of the barrier serving as the contact layer, and an n-type narrow-bandgap semiconductor on the other side functioning as the photon-absorbing layer, which needs to be sufficiently thick for efficient absorption4,5. The ideal design goal is to maximize the conduction band offset (ΔEc), thereby effectively suppress the majority-carrier dark current (electrons), while ensuring a zero valence band offset (ΔEv) to efficiently collect photogenerated holes1,5, as shown in the top panel of Fig. 1a. Conversely, imperfect band alignment, characterized by a small ΔEc or the presence of a ΔEv, can significantly degrade device performance, as illustrated in the bottom panel of Fig. 1a. Therefore, meticulous engineering of the barrier is crucial, with strict consideration of band alignment and lattice matching1,6. Barrier detectors based on conventional infrared materials, such as InSb and HgCdTe, face challenges in achieving minimal or zero ΔEv due to their composition-regulated type-I band alignment1,6,7. Although type-II superlattices (T2SLs) based on III-V materials have been developed for barrier detectors, the epitaxial growth of multilayer materials with varying compositions often results in lattice mismatch and interface states8,9,10. These issues can lead to problems such as Shockley-Read-Hall (SRH) current and trap-assisted tunneling (TAT) dark current, which hinder the advancement of high-performance barrier detectors.

Fig. 1: Challenges in unipolar barriers and the concept of ultra-thin polar barriers (UTPB).
figure 1

a Schematic illustration of ideal and non-ideal unipolar barriers. An ideal unipolar barrier completely blocks one type of carrier while allowing the other to flow freely. In contrast, a non-ideal barrier fails to block one carrier type completely and also impedes the flow of the other carrier. b Band diagrams of two common cases of non-ideal unipolar nBn structures. Case 1: type-II aligned barrier, where a small conduction band offset (ΔEc) causes leakage of dark current despite a zero valence band offset (ΔEv). Case 2: wide bandgap barrier, where a large ΔEc is achieved, but the presence of ΔEv hinders the flow of photogenerated carriers. c Band diagram illustrating the UTPB concept. A polar ultra-thin barrier, with a controllable polarization direction, depletes electrons in the n-type contact layer and facilitates the tunneling of photogenerated holes from the absorption layer, thereby effectively suppressing dark current and improving the collection of photogenerated carriers. d Design of the polarized water-intercalated WSe2/H2O/PdSe2 structure as a UTPB detector. In this design, the n-type WSe2 serves as the contact layer, the n-type narrow-bandgap PdSe2 as the absorption layer, and the strongly polarized water layer acts as the barrier. e, f STEM characterizations of the WSe2/H2O/PdSe2 heterostructure show that few-layer WSe2 and thick PdSe2 are used for better electrical control and optical absorption. The 0.75 nm water layer is uniformly distributed over a large area, with atomically sharp and clean interfaces. B, barrier; Ec, conduction band edge; Ev, valence band edge; ΔEc, conduction band offset; ΔEv, valence band offset; Ef, Fermi level; P, polarization electric field. STEM, scanning transmission electron microscopy. Scale bars: (e), 10 nm; top panel of (f), 20 nm; bottom panel of (f), 2 nm.

Full size image

Two-dimensional (2D) materials, with atomically flat surfaces and free of dangling bonds, can be physically assembled into diverse van der Waals (vdW) heterostructures without the restrictions of lattice matching11,12,13,14. These heterostructures offer atomically sharp and clean interfaces, and their band alignment can be finely tuned to meet specific requirements12,13. While photoconductive and photovoltaic 2D material detectors have received extensive attention, advancements in barrier detectors have been limited. The primary obstacle lies in the difficulty of fabricating an ideal unipolar barrier solely by stacking different 2D layers. Specifically, the bandgap of common transition metal dichalcogenides (TMDs, MX2) generally ranges between 1 and 2 eV due to the similar d-p orbital hybridization between transition metals (M) and chalcogen elements (X)15,16,17. Consequently, stacking these materials with narrow-bandgap 2D semiconductors typically results in a type-II barrier with a small ΔEc, which is inadequate for suppressing majority-carrier dark current, leading to leakage (case 1 in Fig. 1b). Although using wide-bandgap insulating 2D materials, such as hexagonal boron nitride (hBN), as barrier layers can generate a larger ΔEc, this often results in the formation of a non-negligible ΔEv, impeding the collection of photogenerated carriers (case 2 in Fig. 1b). Thus, constructing an ideal unipolar barrier solely through 2D layer stacking remains challenging. While preliminary concepts of barrier detectors based on 2D materials have been demonstrated8, their performance has yet to fulfill the expectations for ideal barrier detectors. There is an urgent need for innovative technical approaches to realize high-performance barrier detectors.

Here, we introduce a concept utilizing ultra-thin polar barriers (UTPB) for the construction of ideal barrier detectors, and demonstrate a polarized water-intercalated WSe2/H2O/PdSe2 structure functioning as a nBn detector. In this design, the nanoconfined water layer, with a thickness of merely 0.75 nm, serves as an ultra-thin, highly polarized barrier. Using atomically resolved imaging and microscopy, we demonstrate that the water-intercalated interface is well-ordered and atomically sharp, devoid of defects or chemical disorder. This carefully designed ultra-thin water layer, with its specific polarized field, not only depletes electrons in the WSe2 contact layer, significantly suppressing dark current (Idark), but also facilitates direct tunneling of photogenerated holes, leading to high sensitivity and ultrafast response. As a result, the device exhibits sub-pA current in the absence of light, regardless of the bias voltage, and can be “activated” by a minimal number of incident photons, showcasing significant photoresponse characteristics. Importantly, the narrow-bandgap PdSe2 absorber (0.16 eV) enables efficient detection across a wide spectral range from ultraviolet (UV) to mid-wave infrared (MWIR), along with broadband polarization resolution. At room temperature, the average specific detectivity (D*) in the MWIR range is greater than 1010 cm Hz1/2 W−1. Even at 4.6 μm, D* can still reach 7.8 × 109 cm Hz1/2 W−1. Furthermore, the device exhibits an ultrafast response, with rise/decay times of approximately 3 μs, consistently across different wavelengths. Both theoretical and experimental analyses elucidate the functional role of the polarized water layer and the underlying working principle of the device, paving the way for high-performance barrier infrared detectors.

Results and discussion

Designing ultra-thin polar barrier structures

As illustrated in Fig. 1a, b, realizing an ideal unipolar barrier nBn structure solely by engineering the band structure via 2D layer stacking presents a significant challenge. The core difficulty lies in identifying materials that can simultaneously achieve a sufficiently large ΔEc and a negligible ΔEv. In the context of an nBn structure, dark current originates from electrons within the n-type contact layer, necessitating a barrier that completely blocks their flow into the narrow-bandgap light-absorbing layer, while ensuring a zero (ΔEv) to efficiently collect photogenerated holes. Therefore, both conditions must be satisfied simultaneously to achieve optimal device performance. Conventionally, this nBn device architecture is implemented by employing wide, high barriers to reduce the thermal emission and tunneling probability of these electrons. However, this strategy always inevitably leads to a non-zero ΔEv, but when adjusting ΔEv to zero through band engineering, the ΔEc is not large enough. Therefore, it is challenging for 2D materials to satisfy both of the above conditions at the same time. In addition to this strategy, it may be possible to consider using a UTPB layer with a strong polarization electric field to deplete electrons in the n-type contact layer. In this case, ultralow dark current can be achieved without the need for a wide and high barrier to block electrons. If the polar barrier layer is sufficiently thin, photogenerated holes can still tunnel directly through it with the assistance of a specific polarization field, thereby creating the potential for high-performance barrier detectors that overcome the limitations inherent in traditional band-structure engineering approaches.

Based on this proposed mechanism, we introduce the concept of UTPB, as illustrated in Fig. 1c. When a UTPB layer with controllable polarization direction is incorporated into a heterostructure, the electrons in the n-type contact layer are depleted by a specific polarization field, effectively blocking their flow to the other side, thus resulting in sub-pA dark current. Simultaneously, photogenerated holes originating from the absorption layer can directly tunnel through the ultra-thin barrier, facilitated by the polarization electric field. This process achieves both suppression of dark current and enhancement of photocurrent. Therefore, the UTPB should have polarization properties and be nanometer-thin, the former should deplete carriers through the interface field, and the latter should remain transport transparent to photogenerated carriers.

In our previous work18, we demonstrated precise control of angstrom-scale water dielectrics in 2D interlayers through water adsorption and vdW assembly. Water molecules have a unique “V” spatial configuration with a bond angle of 104.5°, and the electronegativity of the oxygen atom (3.44) is significantly higher than that of the hydrogen atom (2.20), rendering water a strongly polarized molecule19,20 (Supplementary Fig. 1). Our earlier studies also revealed that due to the existence of periodic potential wells on the surface of 2D TMDs, spontaneous polarization of water molecules occurs readily on their surface without the need for external stimulation, leading to the H+ ion being proximal to the material and the O2- ion being distant from it. For example, in PdSe2, the Se atom sites exhibit lower potential compared to the Pd atoms, as shown in our DFT calculation results (Supplementary Fig. 2a). In this case, when an H₂O molecule approaches the surface of PdSe2, the lower potential energy at the Se atom induces the orientation of the water molecule, where the H+ ions of H2O prefer to point to the Se atom with the electric dipole moment vector pointing out of the plane (as shown in the cross-sectional schematic in the bottom panel of Supplementary Fig. 2a). In order to describe this mechanism more intuitively, we draw a 2D plane diagram to illustrate this process. As shown in Supplementary Fig. 2b, the potential fluctuation of the PdSe2 material causes the surface water molecules to orient, and the subsequent water molecules will follow the previous ones to order polarization, and then collectively exhibit polarization potential. Building upon these foundations, we here constructed an ultrathin water-intercalated WSe2/H2O/PdSe2 structure designed to function as a typical UTPB detector, as schematically illustrated in Fig. 1d. In this configuration, the ultrathin polarized water layer acts as a polar barrier, modulating carrier transport across the heterogeneous interface. For the detailed experimental process and structure, refer to Supplementary Fig. 3a, b and Methods. It is noteworthy that the water is pre-adsorbed onto the PdSe2 material and subsequently assembled with WSe2, inducing spontaneous polarization of the water on the PdSe2 surface. This results in H+ atoms positioning themselves close to the PdSe2, while O2- atoms orient away with the electric dipole moment vector pointing out of the plane, thus generating a polarization electric field directed towards the WSe2. The collective orientation behavior of water molecules induces the formation of a polarization potential on the material surface, thereby regulating the electrical transport of the material. Scanning transmission electron microscopy (STEM) analysis reveals that the thicknesses of the n-type contact layer (WSe2) and the n-type absorber (PdSe2) are approximately 4.5 nm (7 layers) and 43 nm (~ 110 layers), respectively (Fig. 1e). The thin WSe2 layer facilitates better regulation of electron concentration under polarization field, while the ~ 110 L PdSe2 is theoretically thick enough to fully absorb incident light, thereby enhancing quantum efficiency of the device. Low-magnification STEM images reveal that the water layer maintains a uniform thickness across a large area (Fig. 1f), and the construction process does not introduce any damage, strain, or interfacial defects to the materials. High-magnification images confirm that the water layer is only 0.75 nm thick. In addition, PdSe2 exhibits a distinct zigzag atomic arrangement, characteristic of its anisotropic crystal structure, which contributes to the polarization resolution capabilities21,22,23. Elemental mappings confirm that the composition and distribution within the heterojunction align with our expectations. Clean and uniform optical microscopy images further confirm the successful formation of the WSe2/H2O/PdSe2 structure (Supplementary Fig. 3c). Furthermore, Electron energy loss spectroscopy (EELS) and Raman measurements were combined to demonstrate the presence of interlayer water (see Supplementary Fig. 4 for details).

Operation principles and analysis

We then evaluated the optoelectronic performance of the WSe2/H2O/PdSe2 UTPB device. As shown in Fig. 2a, the current-voltage (Ids-Vds) curves are measured under three conditions: dark, weak light irradiation (0.87 μW cm−2), and strong light irradiation (1.65 W cm−2). In the dark state, the device exhibits sub-pA current irrespective of the Vds applied. This observation implies that the electrons in the WSe2 contact layer are entirely depleted, preventing any electron flow. Upon exposure to weak light intensity, the device promptly “turned on” with a high signal-to-noise ratio. This behavior suggests that the polarized-water barrier can act as a “valve”, wherein a small number of photons function as a “trigger” to activate this “valve”. To further highlight the advantages of the polarized water barrier, we compared the dark-state and light-state currents of devices with different barrier types, as shown in Fig. 2b (see Supplementary Fig. 5 for details). The devices include WSe2/PdSe2 (no additional barrier layer), WSe2/H2O/PdSe2 (H2O), and WSe2/2L–hBN/PdSe2 (2L–hBN), WSe2/6L–hBN/PdSe2 (6L–hBN), and WSe2/30L–hBN/PdSe2 (30L–hBN). Cross-sectional STEM images of the hBN devices with different thicknesses revealed flat and clean stacking interfaces, confirming that the devices are well-prepared (see Supplementary Fig. 6 for details). Here, we observe that devices incorporating H2O barrier exhibit the lowest Idark and significant light currents (Ilight), particularly under weak light conditions. In contrast, the pure WSe2/PdSe2 device without any intercalation barrier shows a high Idark due to a small interface energy barrier and junction-related generation-recombination current. The 2L-hBN barrier, with a thickness comparable to the H2O barrier, results in a lower Idark compared to the WSe2/PdSe2 device, but it still yields a higher Idark than the WSe2/H2O/PdSe2 device. This is attributed to the near-transparency of the thin hBN layer to majority carrier electrons, facilitating direct tunneling. While thicker hBN barriers effectively reduce Idark, they concurrently cause a pronounced decrease in Ilight due to the increased energy barrier and widened barrier width, which limits the flow of photogenerated carriers. Notably, the 6L-hBN device fails to generate any detectable photocurrent (Iph = Ilight - Idark) under weak light, and the 30L-hBN barrier completely suppresses photocurrent generation even under strong light conditions. Furthermore, within the hBN series, we observed that increasing the barrier width leads to a reduction in dark current, but this reduction is accompanied by a corresponding decrease in photocurrent. This demonstrates that in conventional unipolar barrier structures, efforts to enhance dark current suppression inevitably reduce the efficiency of photogenerated carrier collection (Supplementary Fig. 7). Collectively, these results demonstrate that the H2O barrier, through its polarization-induced dark current suppression and photocurrent enhancement, offers performance superior to that achievable with conventional detector architecture.

Fig. 2: Basic optoelectronic properties and comparison of different structures.
figure 2

a Ids-Vds curves of the device under dark, weak light, and strong light irradiation conditions. The intensity of the weak light is 0.87 μW cm−2, and the strong light is 1.65 W cm−2. The device exhibits sub-pA current in the dark state, regardless of bias voltage, and can be “turned on” by a very weak light with high signal-to-noise ratios. b Comparison of dark- and light-state currents for devices with different barrier types: None (no additional barrier), 0.75 nm H2O, 2L-hBN, 6L-hBN, and 30L-hBN intercalated between WSe2/PdSe2 heterojunctions. All measurements were made at Vds = − 2 V. The water-intercalated device shows the lowest dark current and significant light current. Vds, drain-to-source voltage; Ids, drain-to-source current; 2 L, 2 layers; 6 L, 6 layers; 30 L, 30 layers.

Full size image

Using density functional theory (DFT) calculations, we determined the positions of the Ec and Ev of 7L-WSe2 and 110L-PdSe2 and plot the band diagram of these materials under various bias voltage conditions (see Fig. 3a and Supplementary Fig. 8). The calculated bandgaps of WSe2 and PdSe2 are 1.27 eV and 0.16 eV, respectively. The narrow-bandgap of PdSe2, serving as the absorber, theoretically enables the device to respond across a broad spectral range from UV to MWIR spectrum. When the drain-source voltage is set to − 2 V, with the drain electrode on WSe2, the energy bands of the heterojunction align in a type-III configuration, which allows the free flow of photogenerated holes in PdSe2. Although water typically exhibits a large bandgap (~ 7.3 eV) and is highly insulating24, the ultra-thin water layer (~ 0.75 nm thick) is insufficient to impede the collection of photogenerated carriers. Furthermore, the polarization field directed towards WSe2 additionally enhances the movement of these carriers.

Fig. 3: Principle analysis of WSe2/H2O/PdSe2 UTPB devices.
figure 3

a Simulated energy band diagram of the device under different Vds conditions, where WSe2, polarized water, and PdSe2 serve as the contact layer, UTPB, and absorber, respectively. The horizontal dashed line represents the heterojunction Fermi level. The black arrows represent the direction of the polarization field of interlayer water. b Arrhenius fitting of the dark current for WSe2/H2O/PdSe2 devices. Inset: Arrhenius fitting of the dark current for WSe2/PdSe2 devices. The dark current of the H2O device remains low and temperature-independent, demonstrating high-operating-temperature features at room temperature and even higher temperatures. The vertical dashed line represents the location of 300 K (room temperature). These data are extracted from Supplementary Fig. 9. c Arrhenius fitting of the photocurrent for WSe2/H2O/PdSe2 and WSe2/6L-hBN/PdSe2 devices. The WSe2/6L-hBN/PdSe2 device follows a thermal emission model, as expressed in the equation of the inset, where J is the current density, A is the Richardson constant, T is the temperature, \({\varphi }_{b}\) is the barrier height, and kB is the Boltzmann constant. The photocurrent of the H2O device remains stable across temperatures, indicating compliance with the direct tunneling (DT) mode. The vertical dashed line represents the location of 300 K (room temperature). This data is extracted from Supplementary Fig. 10. d, e DFT-calculated interlayer potential difference and charge transfer efficiency for WSe2/PdSe2 (d) and WSe2/H2O/PdSe2 (e) structures. Polarized water increases the interlayer potential difference and enhances electron transfer within the heterojunction. The arrows represent the direction of electron transfer within the heterojunction. f Schematic band diagram illustrating the interlayer potential difference and WSe2 depletion state with/without polarized water. Polarized water generates a larger interlayer potential difference, completely depleting electrons in WSe2, with the Fermi level approaching the valence band edge. IDiff., diffusion current; IG-R, generation-recombination current; IBack., background current; DT model, direct tunneling mode; TE model, thermal emission mode. DFT, density functional theory.

Full size image

To gain deeper insights into the mechanisms underlying dark current suppression and photocurrent enhancement, we analyzed the temperature-dependent output characteristics of the devices. Arrhenius fits of the dark currents for the WSe2/H2O/PdSe2 and WSe2/PdSe2 devices at Vds = − 2 V are shown in Fig. 3b (see Supplementary Fig. 9 for detailed data). The dark current of the H2O barrier device remains ultra-low and stable across the entire temperature range (70–350 K), primarily influenced by the background current (IBack.) from the external environment or circuitry. This behavior is attributed to the effective depletion of electrons in WSe2 by the polarized water layer. In contrast, the WSe2/PdSe2 heterojunction device exhibits a large diffusion current (IDiff.) at low temperatures, and a significant generation-recombination current (IG-R) related to SRH traps at high temperatures. These results demonstrate that the H2O barrier device can operate effectively at room temperature and even in high-temperature environments, while the WSe2/PdSe2 device requires refrigeration for low-temperature operation. We also performed Arrhenius fittings for the photocurrents of WSe2/H2O/PdSe2 and WSe2/6L-hBN/PdSe2 devices (Fig. 3c, see Supplementary Fig. 10 for detailed data). The photocurrent of the H2O barrier device remains temperature-independent, suggesting that the collection of photogenerated carriers proceeds via a direct tunneling (DT) mechanism. In contrast, the photocurrent of the 6L-hBN device decreases significantly with decreasing temperature, indicating that carrier collection in this device follows a thermal emission (TE) mode, which can be described by the formula:

$$J=A{T}^{2}{e}^{-\frac{{\varphi }_{b}}{{k}_{B}T}}$$
(1)

where J is the current density, A is the Richardson constant, T is the temperature, \({\varphi }_{b}\) is the barrier height, and kB is the Boltzmann constant. At high temperatures, the photocurrent is proportional to T2, while at low temperatures, it follows an exponential dependence \({e}^{-\frac{{\varphi }_{b}}{{k}_{B}T}}\). These findings consistently indicate that the H2O barrier, with its unique polarization characteristics, enables simultaneous dark current suppression and photocurrent enhancement, offering promising potential for high-performance and HOT photodetectors.

Furthermore, we employed DFT modeling to investigate the unique role of intercalated water. To streamline the system for analysis, we used PdSe2 (2 × 2 × 2 unit cells), WSe2 (2 × 4 × 1 unit cells), and 24 H2O molecules to construct models for WSe2/PdSe2 and WSe2/H2O/PdSe2 (Fig. 3d, e). Based on these models, we calculated the interlayer potential difference and the efficiency of charge transfer. The results reveal that the WSe2/H2O/PdSe2 structure, with water molecules spontaneously polarized on the PdSe2 surface, generates a larger interlayer potential difference. This potential difference is 0.38 V in our simplified model, significantly higher than the 0.28 V observed in the WSe2/PdSe2 structure (Fig. 3f). Moreover, the introduction of polarized water enhances electron transfer efficiency from WSe2 to PdSe2 by approximately 1.6 times. This increased interlayer potential difference and improved charge transfer efficiency effectively deplete the electrons in WSe2, as shown by the density of states (DOS) of WSe2 in the two structures (see Supplementary Fig. 11). These findings are consistent with our electrical measurements, where the complete depletion of WSe2 ensures that the device maintains ultra-low current, regardless of bias voltage or temperature. In addition, through DFT calculations, we also demonstrate that the polarization of water molecules is highly robust and remains stable under external bias (see Supplementary Fig. 12 for details).

Sub-pA dark current photodetection

By engineering the polarized water barrier, the WSe2/H2O/PdSe2 device demonstrates significantly enhanced photodetection performance. We first evaluated its basic performance using a 532 nm laser, and the Ids–Vds curves under various light power densities (Plight) are shown in Fig. 4a. Several notable observations are as follows: 1) Non-conducting dark state: In the dark state, the device maintains sub-pA level current, with an ultralow Idark of ~ 20 fA, remaining stable over a wide range of bias voltages. 2) Light-induced conductive state: The initially non-conducting device becomes activated upon exposure to even ultra-weak light signals. Irradiation with light as low as 1.63 μW cm−2 rapidly transitions the device into a conductive state, with a current increase of roughly 2 orders of magnitude. Meanwhile, the photocurrent increases with increasing Plight, reaching a maximum value of 3.7 × 10⁻⁷ A (Supplementary Fig. 13a). This results in a corresponding increase in the Ilight/Idark ratio, with a maximum ratio of 2 × 107 (Fig. 4b). Here, in the low laser power range, our device exhibits a quasi-linear relationship between light intensity and photocurrent, suggesting the device’s potential for detecting light radiation dose. The estimated linear dynamic range (LDR) of our device is about 53.6 dB. In addition, under 532 nm laser illumination, we note that the open-circuit voltage (Voc) and short-circuit current (Isc) of the device increases steadily with increasing Plight, and they can reach maximum values of 1.2 × 10−7 A and 0.35 V, respectively (Supplementary Fig. 13b). We next quantify the photoresponse by calculating the responsivity (R) (Fig. 4c, see Methods for the calculation formula). Note that the active area of our device for responsivity calculations is estimated by the photocurrent mapping test (see Supplementary Fig. 14 for details). Here, under 532 nm laser illumination, our device exhibits a maximum R of 0.307 A W−1 for Vds = 2 V and 0.202 A W−1 for Vds = − 2 V. Notably, in our device architecture, the core purpose is to suppress dark current, achieve high operating temperature, and improve detectivity, and there is no multiplication effect and related gain.

Fig. 4: Photodetection performance.
figure 4

a Ids-Vds curves of the WSe2/H2O/PdSe2 device at different power densities of 532 nm laser, here the device exhibits non-conducting behavior in the dark state, and a very weak light illumination can “turn on” the device. The black dashed line shows the average dark-state current of the device, which is ~ 20 fA. The blue dashed line shows the change in open-circuit voltage with increasing laser power. b Extracted switching ratios at different light power densities. c 2D plot of extracted responsivity as a function of bias voltage and light power. Here, the wavelength used is 532 nm, and the data are extracted from Fig. 4a. d Time-dependent photoresponse using a modulated pulsed laser with wavelengths ranging from 375 nm to 4600 nm (UV to MWIR). Here, the response curves at different wavelengths are plotted individually, irrespective of the initial time. The bias voltage used is Vds = − 2 V. e Frequency response of the device, exhibiting a -3 dB frequency of 0.2 MHz (MWIR, λ = 4600 um). Inset: response speed of the device, showing a 3.2/3.0 μs rise/decay time, respectively. The bias voltage used is Vds = − 2 V. f Extracted rise/decay time at different wavelengths, from UV to MWIR. The bias voltage used is Vds = − 2 V. The error bars represent the standard deviations of the same sets. g Polarization photoresponse of the device (λ = 1500 nm and λ = 450 nm), showing a statistical photoresponse characteristic at different polarization angles. h Endurance test under 5000 consecutive laser pulses (532 nm), showing that the dark current remains ultralow and the photocurrent stable throughout. i Atmospheric stability test, where the device maintains stability for up to two years. λ, wavelength; UV, ultraviolet; MWIR, mid-wave infrared.

Full size image

The utilization of narrow-bandgap PdSe2 as the absorption layer makes the infrared photodetection performance of our device particularly noteworthy. Fourier-transform infrared spectroscopy (FTIR) tests demonstrate that the PdSe2 layer, with its specified thickness, effectively covers the near-infrared (NIR), short-wave infrared (SWIR) and MWIR ranges (Supplementary Fig. 15). At room temperature, we assessed the time-dependent photoresponse of the device across a wavelength range from 375 nm to 4600 nm (UV to MWIR), as shown in Fig. 4d. The device exhibits a rapid photoresponse across all examined wavelengths, with the largest photocurrent observed at 375 nm, corresponding to the higher photon energy. Even at 4600 nm, the device maintains a switching ratio of > 103 and a rapid photoresponse, highlighting its capability for highly sensitive infrared detection. We further evaluated the response speed of the device (Inset of Fig. 4e). Under Vds = − 2 V and 4600 nm laser illumination, the device exhibits rise/decay times of 3.2/3.0 μs, respectively. These times were extracted from the photocurrent transitions between 10–90% (rise) and 90% to 10% (decay). Moreover, the temporal response bandwidth in the MWIR range (4600 nm) was measured, revealing an ultrahigh -3 dB cutoff frequency of 0.2 MHz (Fig. 4e). This performance exceeds that of most previously reported 2D heterojunction photodetectors25,26,27,28. Importantly, the device exhibits consistent rise/decay times of approximately 3 μs across all tested wavelengths (UV to MWIR) (Fig. 4f and Supplementary Fig. 16a, b), demonstrating a fast, wavelength-independent response speed. This behavior is attributed to the large interlayer potential difference created by the polarized water and the direct tunneling transport of photogenerated carriers from PdSe2. Notably, in the zero-bias state (Vds = 0 V), the response speed remains essentially unaltered (Supplementary Fig. 16c), confirming that the transport of photogenerated carriers is predominantly governed by the polarized water.

In addition to the highly sensitive infrared response, our device is capable of distinguishing broadband polarized light due to the intrinsic anisotropy of PdSe2. We first employed Raman spectroscopy to analyze the angle-dependent structural features of PdSe2 (Supplementary Fig. 17). Using pulsed lasers with rapidly switched polarization angles, including 1550 nm and 450 nm wavelengths, we assessed the time-dependent photoresponse of the device (Supplementary Fig. 18). The extracted polarized photocurrent further demonstrates the device’s ability to accurately resolve the full polarization angle (Fig. 4g). Here, the polarization ratio of our device is approximately 2, which is mainly dominated by the intrinsic anisotropy of PdSe2 and is consistent with the previously reported results of PdSe222. To evaluate the durability of the device, we subjected the device to 5000 consecutive laser pulses (Fig. 4h). Throughout this rigorous testing, the dark current remained ultra-low, and the photocurrent showed negligible variation, indicating that the device can operate under high-intensity conditions for extended periods without degradation. We also conducted atmospheric stability tests over a 2-year period (Fig. 4i and Supplementary Fig. 19). Over this extended duration, both the dark current and the photocurrent remained stable, further confirming the long-term stability of the device. The strong stability can be attributed to both the materials used and the intercalation structure. Regarding the materials, previous reports have confirmed the strong air stability of both WSe2 and PdSe221,29,30,31. Here, we experimentally confirm that the water intercalation structure also exhibits long-term stability (see Supplementary Fig. 20 for details). In addition, we investigate the thermal stability of the dark current of the device (see Supplementary Fig. 21 for details). The device demonstrates a maximum temperature limit of 80 °C, beyond which the non-conductive dark state is disrupted. Structural characterization further demonstrates that high temperatures will cause edge collapse of the intercalation structure (see Supplementary Fig. 22 for details).

Infrared detection metrics and benchmarks

To quantitatively assess the performance of infrared photodetection, we calculated the R and D* across a wide range of wavelength bands. The WSe2/H2O/PdSe2 device exhibits an ultra-low noise current at room temperature, with minimal variation observed at both high (350 K) and low (250 K) temperatures (Fig. 5a). In the noise current density spectrum, the dominant contribution in the low-frequency region is 1/f noise. At high frequencies, shot noise predominates. Our device exhibits remarkably low 1/f noise of 5.6 × 10−31 A2 Hz−1 at 1 Hz, significantly lower than most previously reported 2D heterojunctions and 2D transistors32,33,34,35. This improvement is attributed to the low dark current of the water-intercalated heterostructures, which feature highly ordered atomic arrangements with minimal defects and chemical disorder at the heterointerface. In addition, the device shows reduced shot noise of 6.9 × 10−33 A2 Hz−1 at 1 kHz, which likely originates from the external environment and test equipment. We carefully measured the Ids-Vds characteristics of the device at various wavelengths, including 375, 450, 660, 808, 914, 1064, 1550, 2240, and 4600 nm (Supplementary Figs. 23, 24, 25), and extracted the R, switching ratio, and D* for these wavelengths (see Methods for the calculation formula). As shown in Fig. 5b, the wavelength-dependent responsivity indicates that the device maintains a responsivity of about 0.3 A W−1 across the entire visible to MWIR range. Especially, in the MWIR range, the Ilight/Idark value can still maintain ~ 103 with a significant signal-to-noise ratio (Inset of Fig. 5b). We also calculate the noise equivalent power (NEP) (Inset of Fig. 5c) and the D* (Fig. 5c). The device exhibits an average NEP of about 10−15 W Hz−1/2 in the visible to NIR region and about 1014 W Hz−1/2 in the MWIR region. Meanwhile, in the visible and NIR regions, the average D* is greater than 1011 cm Hz1/2 W−1, and in the MWIR region, the D* remains above 1010 cm Hz1/2 W−1. All measurements were conducted at room temperature.

Fig. 5: Performance metrics and benchmark comparisons.
figure 5

a Measured noise current of the WSe2/H2O/PdSe2 device at temperatures ranging from 250 K to 350 K. The device maintains low noise current across this temperature range. The solid red line shows a linear relationship between frequency and current noise, indicating a 1/f noise characteristic. b Extracted responsivity of the device at different wavelengths. The bias voltage used is Vds = − 2 V, and the power densities used are their minimum power densities at different wavelengths. Inset: Calculated Ilight/Idark values at various wavelengths. The solid line represents a polynomial fit to these data, showing the responsivity trend with wavelength. c Specific detectivity of the device across the UV to MWIR range. Inset: Calculated noise equivalent power (NEP) at various wavelengths. The bias voltage used is Vds = − 2 V, and the power densities used are their minimum power densities at different wavelengths. The solid line represents a polynomial fit to these data, showing the responsivity trend with wavelength. d Comparison of dark current densities at room temperature for our water-intercalated device and various commercial infrared detectors, including HgCdTe, colloidal quantum dots (CQD), T2SL, and typical 2D materials. The water-intercalated device exhibits low dark current. Lines of different colors correspond to Rule 07, Rule 22, and Low 19 of the same color. Rule 07, an empirical formula for predicting the dark current density of high-performance HgCdTe detectors. Rule 22 and Low 19 are the updated versions of Rule 07. The gray shaded area represents the performance region of current 2D material-based infrared detectors. The dashed line corresponding to our device represents an extrapolation based on the characteristics of our device. e Comparative detectivity of our device with that of commercial MWIR detectors, demonstrating superior performance with detectivity values that exceed those of most commercial MWIR detectors. BILP, background-limited infrared performance.

Full size image

We compare the infrared detection metrics of our water-intercalated devices with those of conventional infrared detectors and 2D material-based infrared detectors. As shown in Fig. 5d, we present a comprehensive comparison of the room-temperature dark current of our water-intercalated device with established infrared detector technologies, including HgCdTe, colloidal quantum dots (CQD), and T2SL detectors. Our device demonstrates dark current densities as low as 1 × 10−2 A cm−2, significantly outperforming conventional HgCdTe infrared detectors, including those governed by Rule 07, Rule 22, and Law 19 (empirical relationships that guide the design and application of infrared detectors). The dark current value is 3-4 orders of magnitude lower than that of conventional materials detectors and is also substantially lower than that of typical 2D material-based infrared detectors36,37,38,39,40,41,42,43,44,45. Although our experimental tests were confined to the MWIR range, the dark current of our device is theoretically expected to remain low level into the LWIR range, as extrapolated by the dotted line in the figure. We further compared the room temperature detectivity of our device with state-of-the-art room temperature infrared detectors, many of which are commercially available (Fig. 5e). At room temperature, our device shows an average detectivity exceeding 1010 cm Hz1/2 W−1 in the NIR to MWIR range, which is comparable to commercial mid-infrared detectors such as HgCdTe, PbSe, and InAsSb detectors. This value is also comparable or exceeds that of the reported 2D/3D structure HgCdTe/black phosphorus heterojunction46, and several times higher than 2D black arsenic phosphorus-based infrared detectors47, black phosphorus/molybdenum heterojunction detectors48, black phosphorus alloys-based detectors43, and PdSe2/MoS2 heterojunction44, etc. At 4.6 μm, the performance of our detectors can still reach 7.8 × 109 cm Hz1/2 W−1, where most of the MWIR detectors do not work well due to exceeding their cut-off wavelength. These comparisons highlight the performance of the polarized water-intercalated barrier devices in terms of dark current and detectivity, emphasizing their potential for high-sensitivity and low-noise infrared detection across a broad spectral range.

We have demonstrated the unique advantages of ultra-thin polar barriers for the development of high-performance barrier detectors. By employing a polarized water-intercalated WSe2/H2O/PdSe2 structure, we have illustrated that the polarized water barrier, driven by its specific polarized field, effectively depletes electrons in WSe2, resulting in sub-pA current in the dark state regardless of the applied bias. Meanwhile, this water barrier facilitates the direct tunneling of photogenerated carriers, greatly enhancing the photoresponse performance. The narrow-bandgap PdSe2 enables highly efficient detection across a broad spectral ranging from UV to MWIR with polarization sensitivity. Benchmarking analysis reveals that our device exhibits a dark current that is 3-4 orders of magnitude lower than that of commercial infrared detectors at room temperature. In the MWIR range, the detectivity is greater than 1010 cm Hz1/2 W−1, comparable to commercial HgCdTe, PbSe, and InAsSb detectors. Our work introduces a simple device architecture and operational principle for barrier detectors, overcoming traditional challenges such as band engineering and lattice matching. This approach paves the way for next-generation infrared detectors using much-anticipated 2D semiconductors, which are expected to simultaneously achieve small pixel size, high operating temperature, CMOS compatibility, and high-performance metrics. Moreover, the nano-water intercalation concept and its underlying principles could be extended to other 2D materials and more complex stacks, providing an optional direction of electrical modulation for vdW electronics beyond conventional layer stacking, thus opening up promising avenues for creating various 2D electronic and optoelectronic devices.

Methods

Device fabrication

WSe2 and PdSe2 flakes were obtained by mechanically exfoliating bulk single crystals (purchased from Shanghai Onway Technology Co., Ltd). To fabricate the water-intercalated WSe2/H2O/PdSe2 heterostructure, we first adsorbed water molecules onto PdSe2 and then assembled it with WSe2, forming a heterostructure with water confined between the two materials. Specifically, exfoliated PdSe2 flakes of appropriate thickness were transferred onto a SiO2/P⁺Si substrate for water adsorption. Here, the water adsorption process is controlled and completed in a glove box by a homemade sealed container that is heated on a hot plate (purchased from CIF (Beijing) Tech Co., Ltd). The sealed container contains a small amount of ultrapure water, and the heating is to form a saturated vapor pressure to facilitate the control of the adsorption level. The PdSe2 flakes are adhered to the top of the sealed container, and the adsorption level is controlled by controlling the heating temperature and time. For the ~ 0.75 nm water layer in the WSe2/H2O/PdSe2 structure, adsorption was performed at 30 °C for 10 min. During the adsorption process, a suitable WSe2 flake was exfoliated and transferred onto polydimethylsiloxane (PDMS). Once the water adsorption was complete, the PDMS/WSe2 stack was quickly aligned and laminated onto the PdSe2 surface using a transfer platform with a microscope, ensuring rapid lamination to minimize water molecule volatilization. The WSe2 was subsequently released from PDMS by slight heating, forming the WSe2/H2O/PdSe2 heterostructure with a uniform water layer sandwiched between the two materials. All processes were carried out in a N2-filled glove box.

The fabricated heterostructures were patterned using standard electron beam lithography to define the electrode layout. Cr/Au electrodes (10/50 nm) were deposited by electron beam evaporation, followed by a standard lift-off process to complete the device fabrication.

Characterizations

Raman spectroscopy was performed using a confocal Raman microscope (Witec alpha300R) with a 488 nm laser as the excitation source. Atomic-resolution STEM images were obtained using a Thermo Scientific Helios 5 CX focused ion beam system to prepare cross-sectional slices, which were then analyzed by STEM equipment (from the Analytical Instrumentation Center of Hunan University). Prior to the FIB milling, a sufficiently thick carbon protective layer was deposited, and the sample was thinned using an accelerating voltage of 30 kV and a gallium ion beam with currents ranging from 2.8-47 nA to 24-790 pA. STEM analysis, including high-angle annular dark field (HAADF) imaging, elemental mapping, and EELS, was conducted on a Thermo Scientific Themis Z (3.2) microscope, operated at 300 kV and equipped with a probe-forming aberration corrector.

Electrical and optoelectronic measurements

All electrical and visible optoelectronic measurements were performed at a Lakeshore Probe Station in a vacuum environment (10−5–10−6 Torr). Infrared photoelectric testing was conducted in an ambient environment using an MStarter 200 probe station (Metatest Corporation). The semiconductor parameter analyzers used included a Keysight B1500A, an Agilent B2912, and a PDA FS-Pro. Single-wavelength lasers, delivered to the probe station via optical fiber, were spatially irradiated onto the device. The laser wavelengths used were 375, 450, 532, 660, 808, 914, 1064, 1550, 2240, and 4600 nm. Polarization light response measurements were conducted by inserting a polarizer plate and a 1/2 wave plate into the optical path. Response speed analysis was performed using a Signal Recovery Model 7270 DSP lock-in amplifier and an SRS-SR570 preamplifier. For the long-term stability test, the devices were stored in a conventional nitrogen dry cabinet with humidity less than 10% RH.

For all visible and near-infrared measurements (375–1550 nm), a free-space laser combined with a Lakeshore vacuum probe station is used. Specifically, a laser beam first passes through an attenuator (to control laser power), then a computer-controlled chopper (to modulate the pulses), then a beam expander (to obtain a uniform spot), then an aperture (to standardize the spot shape and size), and finally passes through a reflector and through the probe station window to irradiate the sample. The laser power was measured using a Thorlabs laser power meter. The long-wave measurements (2240 nm, 4600 nm) were performed by a built-up fiber-coupled laser equipped with an MStarter 200 atmospheric probe station.

For the polarization response testing (Supplementary Fig. 18), a laser beam first passes through a linear polarizer to make it standard linearly polarization light, then passes through a 1/2 wave plate, and further passes through an optical chopper. The continuous change of polarization photoresponse can be achieved by manually rotating the 1/2 wave plate and coordinating the pause interval of the chopper.

Responsivity and detectivity calculations

The responsivity R is calculated using the following formula:

$$R=\frac{{I}_{ph}}{{P}_{{lig}ht}S}$$
(2)

where Iph is the photocurrent (Iph = Ilight-Idark), Plight is the power density of the incident laser, and S is the active area of the device.

The specific detectivity D* is calculated by the following equation

$${D}^{*}=\frac{{(SB)}^{1/2}}{N{EP}}$$
(3)

where B is the bandwidth and NEP is the noise equivalent power, given by:

$$N{EP}=\frac{{\overline{{i}_{n}^{2}}}^{1/2}}{R}$$
(4)

Here, \({\overline{{i}_{n}^{2}}}^{1/2}\) represents the root-mean-square value of the noise current at a 1 Hz bandwidth, and R is the responsivity.

DFT calculations

The DFT theory calculations were carried out by an open-source code, Quantum ESPRESSO49. The Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional was used and the ultrasoft pseudopotentials were from the PSlibrary.1.0.0. The kinetic energy cut-offs for the plane-wave basis and charge density were 55 and 448 Ry. The PdSe2 2 × 2 × 2 unit cells, WSe2 2 × 4 × 1 unit cells, and 24 H2O molecules were used to build a computational model. In the calculations, Fermi-Dirac electron smearing was taken into account. The room temperature parameters were assumed to capture the interlayer charge transfer behavior. Crystal optimization was used to perform ground-state calculations. Force and energy criteria were 1 × 10−4 Ry bohr−1 and 1 × 10−4 Ry, respectively. Dipole correction was employed to suppress the charge exchange between adjacent picture elements50. This was achieved by creating virtual dipoles at the out-of-plane dimension boundaries, thereby keeping the electrostatic potential flat. The Brillouin zone was sampled using a 4 × 4×1 k-point mesh for 7L-WSe2, while a 4 × 4 × 4 k-point mesh was employed for bulk PdSe2. Here, the 110L-PdSe2 was considered as a bulk for the calculation. A vacuum space of 60 Å along the z-direction was applied for the layered WSe2 structure to avoid interactions between periodic images. The band structure calculations for both 7L-WSe2 and bulk PdSe2 were performed with the same convergence criteria and computational settings as described, ensuring consistency across the calculations.

For the DFT calculation of the interlayer potential difference, in order to avoid the influence of the mirror structure (in the direction perpendicular to the heterojunction) on the entire potential field, we added a dipole correction at the appropriate position of the vacuum layer so that this region, which is far away from both sides of the heterojunction, exhibits vacuum-like characteristics, rather than the potential interference caused by the additional mirror structure. For the dipole correction region, we chose the region from 0.8·ac to 0.87·ac (24 Å - 26.1 Å), which is far away from the atomic structure and has enough length to avoid steep potential wells from causing wave function interference. In this case, if there is a potential difference at the heterojunction, the potential difference will appear at 24 Å - 26.1 Å region.