Introduction

Sliding ferroelectricity often occurs in the manually stacked bilayer configurations of two-dimensional (2D) materials. The out-of-plane polarization is flipped by in-plane sliding one layer by a fraction of the unit cell1,2,3. This endows sliding ferroelectrics with many advantages, including not requiring polar crystals4, ultrafast and fatigue-free switching5,6, and low polarization reversal barrier6. Initially, Wu et al. proposed the sliding ferroelectric mechanism and predicted that BN and MoS2 can generate out-of-plane polarization by sliding engineering7. This theoretical prediction was subsequently verified through extensive experiments, such as topological semimetal WTe24,8, insulator BN1,2, semiconductors (MoS2, WSe2, WS2, InSe)9,10,11,12,13,14,15, and single crystal16. These pioneering progresses open up a research field called slidetronics1,2. Noteworthy, the sliding ferroelectric semiconductors with moderate band gap and the efficient gate tunability exhibit significant application potentials in multifunctional information storage12,17,18, ferroelectric photovoltaics13,15,19, and neuromorphic computing15,20. Therefore, they have great potential to advance the development of slidetronics.

As a typical sliding ferroelectric semiconductor, 3R-stacked bilayer MoS2 is only C3-symmetry, having neither an inversion center nor a mirror symmetry plane. This symmetry breaking allows for charge transfer between adjacent atomic layers, resulting in spontaneous polarization along the out-of-plane direction10. However, constructing such systems typically relies on complex manual stacking or specific growth techniques. Notably, an important work has reported the enhanced sliding switchable polarization by modifying the microstructure of InSe system through Y doping strategy21, which avoids the complex artificial stacking. To further broaden the selection scope of material, it is crucial to explore intrinsic sliding ferroelectric semiconductor systems without artificial stacking and chemical modification. ReX2 (X = S, Se) exhibits a twisted octahedral structure (1T’ phase) with different natural stacking configuration (parallel and antiparallel)22,23. From the perspective of point groups, monolayer ReX2 belongs to the Ci point group and only contains one symmetry operation24, namely inversion of the central symmetry, which does not produce out-of-plane polarization. For ReX2 of antiparallel stacking with layer number N ≥ 2, the out-of-plane polarization was observed, and the electric polarization can be switched by interlayer sliding25,26. However, in parallel-stacked ReX2 systems, whether vertical polarization can be induced by sliding engineering remains unexplored, such studies would be useful in understanding the physical origin and spurring further research into sliding ferroelectric semiconductor system. Therefore, it is crucial to explore in depth the sliding approach for realizing ferroelectricity in naturally parallel-stacked multilayer ReX2 systems and to elucidate the mechanism on the atomic scale.

Here, we demonstrate sliding ferroelectricity in parallel-stacked ReSe2 semiconductor with layer number N ≥ 3, which is different from the conventional stacking structure. The first-principles calculations uncover the origin of ferroelectricity that the sliding of arbitrary middle atomic layer breaks the spatial inversion symmetry, and the displacements of microscopic atomic configuration are directly visualized using scanning transmission electron microscopy (STEM). Meanwhile, piezoresponse force microscopy (PFM) measurements demonstrate the visualized ferroelectric domains and ferroelectric switching property. As the direct observation of sliding ferroelectricity, the ferroelectric hysteresis loop shows an experimental polarization value of ~0.67 μC/cm2. In addition, the switchable ferroelectric diode effect and the programmable ferroelectric photovoltaic effect are successfully demonstrated in the ReSe2 based sandwiched device, further confirming sliding ferroelectricity. This work avoids the rigorous artificially stacking design required for conventional sliding ferroelectrics and expands the family of sliding ferroelectric semiconductors into the naturally parallel-stacked multilayer ReX2.

Results

ReSe2 is a high ambient stable van der Waals material where each layer is composed of a Re atoms plane sandwiched between two Se atoms planes27. Different from the most common group-VI transition metal dichalcogenides (e.g., MoS2 and WS2), there are unbound valence electrons outside the nucleus of the Re atom in ReSe2. Due to the Jahn-Teller effect (or Peierls distortion), the adjacent Re atoms are bonded in the form of zigzag Re4 clusters and align along b-axis direction to form Re4 chains, forming a distorted octahedral (1T′) structure and reducing the symmetry of structures (Fig. 1a)28,29,30. In addition, the Peierls distortion greatly weakens the van der Walls interaction in the interlayer31. The low-symmetry structure and weak van der Waals interlayer interactions of ReSe2 endow it the unique ability of interlayer sliding32, which facilitates the realization of low-potential-barrier interlayer sliding. ReSe2 crystal was synthesized using the chemical vapor transport method, and ReSe2 flakes of different thicknesses were mechanically exfoliated using polydimethylsiloxane (PDMS) films to further verify their atomic structure and physical properties (More details can be found in “Methods”). Figure 1b, c shows atomic-level high-angle annular dark-field (HAADF) image and selected area electron diffraction (SAED) pattern of ReSe2 flake along the [120] zone axis. The results exhibit high-quality layered parallel stacking structure. The ultraviolet-visible-near infrared (UV-Vis-NIR) absorption spectra of samples with different thicknesses show the sharp absorption edge near 950 nm (Fig. 1d). From the Tauc plot, the band gap was calculated to be ~1.1–1.2 eV, which is very close to the reported value27,33. The transfer characteristic curve indicates that ReSe2 is an ambipolar semiconductor material (Supplementary Fig. 1). From Raman results of ReSe2 with different thicknesses (Fig. 1e and Supplementary Fig. 2), the peaks at 124 cm−1 and 173 cm−1 correspond to the in-plane Eg-like mode and the out-of-plane Ag-like mode, respectively, and the around 13 modes were observed between 100 cm−1 and 300 cm−1, which is much more than that of other transition metal dichalcogenides34. These Raman peaks are mainly attributed to the low crystal symmetry of ReSe2 and related to fundamental in-plane and out-of-plane Raman modes27. The anisotropy of ReSe2 flake, caused by its low crystal symmetry, was confirmed by the excitation angle-resolved Raman spectra (Supplementary Fig. 3a–d). Subsequently, second harmonic generation (SHG) technique, a reliable tool for demonstrating the non-centrosymmetric structure35,36, was adopted to characterize the structural symmetry of ReSe2 flakes. The schematic diagram of the SHG was shown in Supplementary Fig. 4a. Figure 1f illustrates the SHG intensity of samples with different thicknesses. A significant SHG signal at 532 nm was observed under the excitation wavelength of 1064 nm, revealing ReSe2 flakes have a non-centrosymmetric structure. The power-dependent SHG intensity of ReSe2 flake (~2.5 nm) was investigated (Supplementary Fig. 4b), and the SHG intensity increased monotonically with the power intensity. We further investigated the effect of crystal structure on signal intensity by polarization-resolved SHG. A clear four-lobed pattern was observed (Supplementary Fig. 4c), showing the variation of SHG intensity with azimuthal angle. The above results confirm that ReSe2 is a parallel-stacked semiconductor material with a non-centrosymmetric structure.

Fig. 1: Crystal structures and spectral characteristics of ReSe2 flakes.
figure 1

a Top-view of the atomic arrangements in monolayer ReSe2. The four covalently bonded Re atoms in the unit cell arrange in a parallelogram and the adjacent parallelograms connect by covalent bonding along the b-axis forming a Re-chain. Purple and yellow spheres represent Re and Se atoms, respectively. b, c HAADF image of mechanically exfoliated ReSe2 flake along the [120] direction and the corresponding SAED pattern. UV-Vis-NIR absorption spectra (d), Raman spectra (e), and SHG signals (f) of ReSe2 with different thicknesses.

Full size image

Next, we studied theoretically using first-principles calculations whether ferroelectric polarization can be induced through sliding mechanism in naturally parallel-stacked ReSe2. Since ReSe2 exhibits a distorted octahedral structure (1T’ phase), the monolayer belongs to the Ci point group with central inversion symmetry and no vertical polarization occurs. For parallel-stacked bilayer ReSe2 crystal, a 1 × 1 × 2 supercell containing two layers is insufficient to induce polarization by interlayer sliding because of its inversion symmetry preserve (Supplementary Fig. 5). Therefore, a large 1 × 1 × 6 supercell is needed as shown in Fig. 2a. Starting from the large supercell of parallel stacking ReSe2 with inversion symmetry, sliding the middle two layers of six layers can break its inversion symmetry. There are two low symmetry stacking modes (state 1 and state 2) appear when the sliding of middle two layers, corresponding to geometric structures in Fig. 2b, c. The atomic arrangement of the intermediate state was shown in Supplementary Fig. 6. The calculation of polarization based on Berry-phase method confirmed the symmetry breaking of state 1 and state 2 stacked configurations. The state 1 and state 2 structures have spontaneous vertical polarization with opposite polarization directions, and the polarization values are calculated to be 4.37 and −4.37 μC/cm2, respectively. Therefore, the polarization of ReSe2 can be switched from upward to downward when the middle two layers sliding, and the first-principles density functional theory calculated transition barrier is ~12.8 meV/f.u. (Fig. 2d). In order to reveal the origin of the out-of-plane polarization, we calculated the charge density difference between the top, middle, and bottom layers in state 1 stacking configuration. As shown in Fig. 2e, it is found that an interlayer net charge transfers after sliding transition, resulting in the accumulation and depletion of electrons in the van der Waals gap.

Fig. 2: Theoretical analysis of ferroelectricity in ReSe2 flakes.
figure 2

a Schematic illustration of the atomic arrangement for parallel stacking. b, c Schematic illustration of the atomic arrangement for different polarization directions (state 1 and 2), and the state 2 can be obtained by mirroring the state 1 with respect to the central horizontal plane. d Ferroelectric switching pathway for multilayer ReSe2 from polar state 1 to polar state 2. e The charge density difference between the layers in polar state 1 produces a vertical polarization. The arrows indicate the polarization direction. Purple and yellow spheres represent Re and Se atoms, respectively.

Full size image

Furthermore, the influence of layer numbers of ReSe2 and sliding layer numbers on polarization were studied. The ferroelectric polarization occurs in the parallel-stacked multilayer (N ≥ 3) ReSe2 (Fig. 3a and Supplementary Fig. 7). Figure 3b shows the ferroelectric polarization values of sliding different layers in a fixed six-layer system. It can be found that no matter how many middle layers undergo sliding, polarization is always generated. These theoretical results indicate that sliding ferroelectricity can be easily obtained through sliding mechanism in naturally parallel-stacked ReSe2. Excitedly, the sliding of middle layers of ReSe2 flake was clearly observed from high-resolution STEM-HAADF images. The relative sliding of the middle two layers with left sliding (Fig. 3c) and right sliding (Fig. 3d), corresponding to two opposite ferroelectric polarization states. Atomic configurations of the two polarization states, i.e., state 1 (downward polarization) and state 2 (upward polarization) were placed on top of the HAADF images of Fig. 3c, d. Additionally, we also observed the sliding of the middle three layers (Fig. 3e) and five layers (Fig. 3f). The sliding layers were marked by a shaded box.

Fig. 3: Theoretical and STEM-HAADF microscopic analyses of ReSe2 sliding ferroelectricity.
figure 3

a Ferroelectric polarization values of sliding one layer in different layer systems. b Ferroelectric polarization values of sliding different layers in a fixed six-layer system. c, d HAADF images recorded along the [100] direction for the ferroelectric polarization state 1 and state 2. In the ReSe2 structural model, purple and yellow spheres represent Re and Se atomic columns, respectively. e, f HAADF images of sliding the middle three layers and five layers.

Full size image

Based on the above theoretical prediction of sliding ferroelectricity in parallel-stacked ReSe2 and direct observation of sliding layers, we experimentally investigated the ferroelectricity of ReSe2 by PFM technique. PFM is a powerful tool for demonstrating the switchable ferroelectric polarization, enabling non-destructive visualization and modulation of ferroelectric domains37,38,39,40,41. The ReSe2 flakes of different thicknesses used for PFM measurements were obtained through mechanically exfoliation and then transfer onto the conducting Pt substrate (Hefei Kejing Material Technology Co., Ltd supplies) which has good smoothness and conductivity. Figure 4a, d presents the topography images and the height profiles of ~10 L (~6.3 nm) and ~3 L (~1.8 nm) samples. The corresponding out-of-plane PFM phase images at room temperature are shown in Supplementary Fig. 8a, b, revealing a clear phase contrast with multidomain structure. Figure 4b, e shows the PFM phase images of 10 L, 3 L samples after writing box-in-box patterns by reversed DC bias (±4 V), (±3 V). Clear reversal of phase contrast demonstrates the ferroelectric switching in naturally parallel-stacked ReSe2 flakes. Furthermore, the corresponding local out-of-plane PFM hysteretic loops in phase and amplitude were also recorded. The PFM phase response exhibits a hysteresis loop with a phase contrast of ~180°, while the amplitude response shows a butterfly-shaped curve (Fig. 4c, f), both of which further confirm the robust ferroelectric polarization. However, the ferroelectric polarization was not observed in bilayer ReSe2 (Supplementary Fig. 9), which is consistent with the theoretical results. Then, we further evaluated the piezoelectric coefficient d33 of an exfoliated ReSe2 flake with a thickness of ~8.5 nm using PFM. When the AC voltage (VAC) is scanned from 1 V to 5 V, the out-of-plane (OOP) amplitude signal of the sample significantly increases (Supplementary Fig. 10). By fitting the amplitude of the sample to the driving VAC, the effective d33 of the 1T’ ReSe2 flake was determined to be ~3.83 pm V−1. Additionally, the temperature-dependent SHG intensity reveals the Curie temperature of 1T’ ReSe2 to be ~490 K (Supplementary Fig. 11).

Fig. 4: Ferroelectric polarization switching by PFM for multilayer (10 L and 3 L) ReSe2.
figure 4

a, d Topography, b, e Out-of-phase images after writing box-in-box patterns by applying reverse voltage bias. c, f Local phase (red) and amplitude (blue) hysteresis loops.

Full size image

Polarization-electric field (P-E) hysteresis loops are regarded as direct evidence for proving ferroelectricity16,42. Therefore, we fabricated Au/ReSe2 (~15.5 nm)/Au vertical device to measure macroscopic ferroelectricity (Supplementary Fig. 12a, d) (more specific preparation details are shown in “Methods” section). The device exhibits the standard P-E hysteresis loops, confirming ferroelectricity of ReSe2, and the spontaneous polarization (Ps) is ~0.67 μC/cm2 (Fig. 5a). The discrepancy between experimental and theoretical polarization value may be attributed to the partial switching of the ferroelectric domains and the differences in the layer number of sample and sliding. Additionally, when applying different poling voltages, switchable diode effects are observed (Supplementary Fig. 13), and the rectification ratio reaches ~33 and ~57 with applying poling bias of −5 V and 5 V (Fig. 5b). This can be explained by the single-barrier model of source43,44,45, and the positive (negative) bias produces a positive (negative) polarization charge at the source, which further causes a decrease (increase) in the potential barrier height, resulting in an increased (decreased) current (Supplementary Fig. 14). In addition, the nanoscale electrical characterization was performed by Conductive Atomic Force Microscope (C-AFM) and Kelvin Probe Force Microscope (KPFM). An open I-V loop and increasing surface potential appear, which is due to ferroelectric polarization switching (Supplementary Figs. 15, 16). We also constructed BN/Gr/ReSe2 (~17.9 nm)/BN/Au device to measure the endurance of ferroelectric switches (Supplementary Fig. 12b, e). By applying single pulses with a width of 1 s and amplitudes of Vpulse = 3 V and Vpulse = −3 V repeatedly at Vread = 0.5 V, the switching properties of ferroelectric ReSe2 remain almost unchanged over 500 switching cycles (Fig. 5c). This stable cycle demonstrates the naturally parallel-stacked multilayer ReSe2 semiconductor has the robustness of polarization switching.

Fig. 5: Switchable ferroelectric polarization in ReSe2 vertical devices.
figure 5

a P-E hysteresis loops at different scanning voltages. b Switchable ferroelectric diode effect. c Endurance of sliding ferroelectrics against switching cycles after applying ±3 V pulse voltages with pulse width of 1 s. d I-V curves under different light intensities and in the dark. e Programmable photovoltaic effect after applying polarization voltages of 0 V (initial), −2 V, −3 V, 2 V, and 3 V, respectively. f Time-dependent photocurrent at various light intensities (Vread = 0 V).

Full size image

Given the ferroelectricity and semiconductor properties of ReSe2, it is expected to generate polarization tunable photovoltaics. A Gr/ReSe2 (~15.8 nm)/Au vertical device was fabricated (Supplementary Fig. 12c, f) to investigate photovoltaic effect through measuring the vertical current (detailed fabrication process described in “Methods” section). The few-layer Gr flake was used as top electrode due to its high optical transmittance in the visible range13,19. Figure 5d depicts the I-V curves in the dark and under 457 nm illumination with different light intensity from 0.1 mW/cm2 to 10.2 mW/cm2. As expected, the photovoltaic response of the device is closely related to the light intensity. Both open-circuit voltage (Voc) and short-circuit current (Isc) increase with increasing light intensity. This is due to increasing the light intensity induces more electron-hole pairs separated by the built-in electric field, which leads to larger Voc and Isc. More interestingly, the photovoltaic effect of the ReSe2 flake is programmable by different polarized electric fields. As shown in Fig. 5e, initially, the device exhibits Voc of 0.08 V and Isc of −43 pA. Compared to widely reported 3R-MoS2, 3R-WS2, ε-InSe, and γ-InSe, larger Voc values were achieved at smaller optical power densities in ReSe2 ferroelectric photovoltaic device (Supplementary Fig. 17). By applying negative poling voltages (−2 V, −3 V), the I-V curves shift rightward and the Voc increases to 0.11 V and 0.13 V, and the Isc decreases to −72 pA and −100 pA. Subsequently, by applying positive poling voltage (+2 V, +3 V), the I-V curves shift leftward and the Voc decreases to 0.03 V and −0.01 V, and the Isc increases to −42 pA and +37 pA. These results demonstrate that the ferroelectric polarization state can enhance or suppress the photovoltaic effect of the ReSe2. This programmable photovoltaic effect is reproducible and it can also be obtained on other ReSe2 device (Supplementary Fig. 18). In addition, we found that the photocurrent can be read out without applying reading voltage, and there are no significant changes in over 20 consecutive switching cycles (Fig. 5f).

Discussion

In conclusion, we report sliding ferroelectricity in naturally parallel-stacked multilayers (N ≥ 3) ReSe2 semiconductor. This is very different from reported antiparallel-stacking ReS2 and typical rhombohedral-stacking transition metal dichalcogenides, where the sliding ferroelectricity can appear in bilayer samples. The origin of the sliding ferroelectricity in parallel-stacked ReSe2 is investigated by first-principles calculations, indicating that the intermediate-layer sliding in the multilayer system breaks the spatial inversion symmetry. The non-centrosymmetric configuration was revealed by SHG technology and the intermediate layer sliding is visually observed by STEM. The vertical ferroelectric switching can occur in ReSe2 multilayers (N ≥ 3), but is not existent in its centrosymmetric bilayers. The P-E hysteresis loops of ReSe2 flake directly show a polarization value of ~0.67 μC/cm2. In addition, the ReSe2 vertical devices exhibit switchable ferroelectric diode effects and programmable ferroelectric photovoltaic effects. We anticipate that our work will provide a perspective for the research and application of sliding ferroelectrics.

Methods

Sample preparation

The bulk single crystal ReSe2, Gr, and BN were purchased from Nanjing MKNANO Tech. Co., Ltd. (www.mukenano.com). The flakes with different thicknesses were mechanically exfoliated from the bulk crystals using polydimethylsiloxane (PDMS). Then, the exfoliated flakes were transferred to the target substrates by the dry transfer technique and heated at 80–90 °C for 10 min to release the adhesion of PDMS to ensure the successful transfer of the flakes.

Device fabrication

Au/ReSe2/BN/Au devices were prepared by transferring ReSe2 flakes onto SiO2 (300 nm)/Si substrate of pre-deposited Ti/Au (5 nm/40 nm) electrodes by dry transfer technique, and then depositing Au top electrodes on the surface of the flakes. The electrodes were defined using a standard lithography technique and deposited using electron beam evaporation. Additionally, Gr/ReSe2/BN/Au and BN/Gr/ReSe2/BN/Au devices were prepared by sequentially transferring the alternative materials onto the pre-deposited Au electrodes.

Characterization

The morphology of flakes and devices was characterized by optical microscopy (OM, AOSVI, M330). The Raman spectroscopy (Horiba LabRAM HR Evolution) was used to measure Raman spectra of ReSe2 flakes. Second-harmonic generation (SHG) measurements were conducted using Raman spectrometer equipped with a femtosecond laser (1064 nm) and high-temperature accessories (Linkam THMSG600G). The UV-Vis-NIR spectroscopy (Metatest, MStarter ABS) was used to characterize the optical bandgap. The STEM-HAADF images were acquired by an aberration-corrected JEM-ARM200F microscope operating at 200 kV. During the HAADF image acquisition, the convergence angle was 21 mrad, and the camera length was 8 cm, and the corresponding collection angle ranged from 68 to 280 mrad. The surface morphology and thickness of ReSe2 flakes were measured using tapping-mode atomic force microscopy (AFM, Bruker Dimension Icon). PFM measurements of ReSe2 flakes transferred onto conductive Pt substrates were performed using a commercial atomic force microscope (Bruker Dimension Icon) with an AC resonance mode under the tip-sample contact resonant frequency (~320 kHz). A standardized ferroelectric tester (Precision Multiferroic IcI, Radiant Technologies) was used to measure P-E hysteresis loops. The light sources of different wavelengths were provided by commercially available light-emitting diode lamps, where the power density was measured by FieldMate+PM150X photometer. The electronic and optoelectronic measurements were conducted on Lakeshore’s vacuum probe station using a Keithley 4200SCS semiconductor parameter analyzer.

Density functional theory calculation

The first-principles density functional theory (DFT) calculations were performed by the Vienna ab initio simulation package46 with projector augmented wave method and the generalized gradient approximation with Perdew–Burke–Ernzerhof exchange-correlation functional47. We adopted a plane-wave cutoff energy of 600 eV and a Monkhorst-Pack k-point mesh of 12 × 12 × 1 in the structural relaxation calculations. The optimized lattice constants of parallel-stacked multilayer ReSe2 are a = 6.63 Å, b = 6.75 Å, and c = 6.80 Å with triclinic symmetry. The kinetic of polarization reversal with interlayer sliding pathways were calculated using the climbing nudged elastic band method48. The Berry phase method was used to calculate the perpendicular crystalline polarization49. The van der Waals interactions in layered ReSe2 are described via the van der Waals correction with optB86b functional in DFT calculations. The spatial charge distribution of the systems was calculated using the Bader Charge analysis50.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.