Introduction

Symmetry breaking plays a crucial role in the fascinating physical properties of low-dimensional layered materials. It offers opportunities to explore exotic physical phenomena, including but not limited to, valleytronics, twistronics, interfacial ferroelectricity, and nonlinear electrical transport1,2,3,4,5,6. For instance, the bulk photovoltaic effect (BPVE) has been widely investigated in non-centrosymmetric bulk materials. This intriguing nonlinear photoresponse phenomenon has regained attention for its potential in high energy harvesting efficiency and polarization detection (linear or circular)7,8,9.

In recent years, engineering symmetry breaking in low-dimensional layered materials has created extraordinary opportunities to manipulate their electrical, optical, magnetic, and topological properties10,11. Devices based on BPVE in non-centrosymmetric materials exhibit spontaneous photocurrent characteristics, such as shift current, injection current, and drift current driven by the depolarization field12,13,14. The shift current is a nonlinear photocurrent dependent on the polarization of the excitation light, generally expressed as a shift vector in real-space15. Polarization information of light holds immense value in medical applications, remote sensing, and modern military technology, especially in infrared polarization photodetection. Low-dimensional materials exhibit tremendous potential in high-performance polarization photodetectors, aligning with the ongoing trend of miniaturization and high integration for next-generation on-chip optoelectronic devices16.

The polarization ratio (PR), calculated using PR = Imax/Imin (or PR = Vmax/Vmin), describes the polarization sensitivity of photodetectors. The Imax and Imin (or Vmax and Vmin) refer to the maximal and minimal photocurrent (or photovoltage) response with different polarization angle lights. Generally, the performance of polarization devices based on anisotropic materials is often constrained by anisotropic light absorption. Some strategies to enhance the polarization detection ability of devices have been present and achieved remarkable results in the last few years, such as the homojunctions17,18, heterojunctions19,20,21,22, and metal-nanostructure-mediated materials23,24,25. Among them, the utilization of metal nanostructures enables PRs to reach values exceeding 104 and even exhibit unipolar (PR ≥ 1) and bipolar (PR ≤ −1) characteristics. Notably, the photoresponse in metasurface-mediated graphene photodetector is referred to as an artificial BPVE23.

In addition to conventional BPVE materials (such as ferroelectrics), photodetectors based on other artificial materials, such as van der Waals interfaces, also show polarization-sensitive photoresponse. For example, a linear polarization photocurrent was observed at the WSe2/BP26 heterointerface, and a direction-selective spin photocurrent occurred at the WSe2/SiP heterointerface27. Extensive investigations have been conducted to explore the physical mechanism between the spontaneous photocurrent and crystal symmetry. An obvious conclusion is that polarization photodetection can be realized by design symmetry of materials. However, the majority of BPVE devices show unsatisfactory performance, with polarization ratio values smaller than 2. Although twisted double bilayer graphene has successfully demonstrated tunable mid-infrared polarization detection28, the generation of Berry curvature dipole in this type of device is still constrained by angle or heterointerface-symmetry misalignment29. Therefore, it is essential to explore a more controllable method to realize symmetry-breaking engineering for high-performance polarization photodetector.

Here, we introduce a convenient and efficient strategy for symmetry-breaking engineering of 2H-MoTe2 by ferroelectric-domain-pattern doping. The interfacial symmetry of P(VDF-TrFE)/2H-MoTe2 is broken by the design of non-centrosymmetric ferroelectric-domain patterns. The electrical and photoelectric behavior similar to BPVE is observed, and infrared polarization-dependent photocurrent is examined in doped 2H-MoTe2. When illuminated, the device shows spontaneous photocurrent and linear polarization sensitivity. In photovoltaic performance, the device shows a considerable short-circuit photocurrent intensity (Jsc) of 29.9 A/cm2 and an open-circuit voltage (Voc) of 3 × 105 V/cm. Additionally, the device demonstrates enhanced polarization sensitivity, with the PR values showing a unipolar (PR ≥ 1) state and bipolar (PR ≤ −1) state at a specific bias voltage. These experimental observations reveal the successful demonstration of a general strategy for fabricating high-performance polarization photodetector, which frees symmetry-breaking engineering from angle misalignment and interface confinement in artificial stacks.

Results

Engineering symmetry-breaking by patterned ferroelectric domain

The external electric field is an effective tool to break the symmetry of low-dimensional semiconductors, enabling switchable Berry curvature dipole and tunable nonlinear optical effect28,29. Ferroelectric is a kind of dielectric material, which own electrically switchable and non-volatile polarization. Many physical properties of low-dimensional materials, such as band structure, carriers, optical conductivity, etc., are easily tuned by the ferroelectric gate. We selected ambipolar low-dimensional semiconductor 2H-MoTe2 and organic film poly(vinylidene fluoride-trifluoroethylene) (P(VDF-TrFE)) to fabricate devices. 2H-MoTe2 is one of the TMDs with a bandgap of ~1 eV at room temperature, the ambipolarity of which makes 2H-MoTe2 an interesting candidate for electrostatically tunable optoelectronic devices. Each monolayer MoTe2 is composed of a layer of tellurium (Te) atoms arranged hexagonally, which is sandwiched between two layers of tellurium (Te) atoms. Figure 1a shows a schematic of the MoTe2 structure with a threefold (C3v) rotational symmetry, meaning that the crystal appears identical after partial rotation around a rotational axis. P(VDF-TrFE) is a kind of high-quality ferroelectric material, known for its large remanent polarization (7 μC/cm2), transparency, excellent retention, and endurance properties. Piezoresponse force microscopy (PFM) allows easy manipulation of the ferroelectric domains (polarized direction and region) at the nanoscale30,31. Figure 1b shows a T-shape ferroelectric domain via ferroelectric-domain nano-patterning by PFM. The T-shape ferroelectric domain exhibits C1v symmetry, which means no rotational symmetry. When polarized up (Pup) or down (Pdown) state, P(VDF-TrFE) can provide a negative or positive ferroelectric field to tune 2H-MoTe2 into p-type or n-type. By intentionally combining the T-shape ferroelectric domain and MoTe2, holes accumulate in the T-shape region while electrons accumulate in the rest part. As a result, a non-centrosymmetric carrier distribution is built in 2H-MoTe2, thus generating a symmetry-broken heterointerface, as shown in Fig. 1c.

Fig. 1: Schematic of symmetry-breaking at ferroelectric/MoTe2 heterointerface.
figure 1

a Schematic figure of pristine 2H-MoTe2 with C3v rotational symmetry. b T-shape ferroelectric domain with C1v rotational symmetry. The symbols in the dark blue (Pup) and light blue (Pdown) refer to the direction of the ferroelectric field. c Ferroelectric/MoTe2 heterointerface with C1v rotational symmetry. d, e, f The laser polarization dependence of second harmonic generation (SHG) of pristine 2H-MoTe2, T-shape ferroelectric domain, and ferroelectric/MoTe2 heterointerface. 0° is defined as parallel to the horizontal direction, while 90° is defined as to vertical to the horizontal direction. g Schematic of the output properties of photoconductive effect (PC effect) in the dark (black) and under illumination (red). h Schematic of the ferroelectric hysteresis loop. The top panel shows the T-shape ferroelectric with opposite polarization direction can induce corresponding electrons and holes in the semiconductor below. i Schematic of the output properties of artificial bulk photovoltaic effect (BPVE) in the dark (black) and under illumination (red).

Full size image

The polarization-dependent second harmonic generation (SHG) measurements were performed to verify the symmetry analysis above, as shown in Fig. 1d–f. Pristine 2H-MoTe2 has a point group of D 3 h and a rotational symmetry of C3v. So, the polarization-dependent SHG of pristine 2H-MoTe2 shows an expected six-fold homogeneous pattern and indicates its crystal structure, as Fig. 1d shows. The SHG of the T-shape ferroelectric domain is highly in-plane anisotropic (Fig. 1e). In theory, the SHG performance in the polymeric film is decided by the electrically-induced dipole alignment32,33. This is due to the aligned orientation of dipoles, which leads to a higher second-order susceptibility when excited by light with a specific polarization angle34. The SHG signal of the ferroelectric/MoTe2 heterointerface exhibits a distorted pattern. Significantly, the tunable SHG can be attributed to not only the doping effect of the external ferroelectric field but also the ferroelectric domains with equivalent size to the incident light (Fig. 1f). This result confirms that the ferroelectric domains can break the rotational symmetry and induce in-plane polarization in 2H-MoTe235,36. Theoretically, the C3v rotational symmetry typically ensures isotropic properties37,38. According to theoretical results calculated by the Floquet formalism15,39, the inversion symmetry of the material ensures that the shift vector is an odd function and the net current is zero. When the C3v rotational symmetry is broken, 2H-MoTe2 can exhibit anisotropic vibrational, optical, and electrical properties27,35. Figure 1g exhibits the photoelectric conversion mechanism as a photoconductive effect (PC effect) in pristine 2H-MoTe2. As shown in Fig. 1h, the T-shape ferroelectric domain offers a T-shape remnant ferroelectric field and induces the corresponding distribution of electrons (Pdown) and holes (Pup) in 2H-MoTe2, resulting in non-centrosymmetric semiconductor (charge-induced electrical nonlinearity). In non-centrosymmetric materials, the spontaneous current, originating from the difference of the Berry phases interband transitions, is hardly non-negligible15,40. Thus, a kind of artificial BPVE light-to-current conversion mechanism occurs in the symmetry-breaking ferroelectric/MoTe2 heterointerface naturally (Fig. 1i).

More interestingly, the shift current is a nonlinear phenomenon depending on the polarization of excitation light. So, we can fabricate a polarization photodetector based on such a symmetry-breaking ferroelectric/MoTe2 heterointerface. Figure 2a shows the device structure. The 2H-MoTe2 was exfoliated and transferred onto a SiO2/Si substrate and then coated by P(VDF-TrFE). As shown in Fig. 2a, the arrayed T-shape ferroelectric domains were fabricated by PFM conductive tip. More details about T-shape array ferroelectric-domain fabrication and information can be found in Supplementary Figs. 1 and  2. In order to experimentally confirm the ferroelectric domains breaking the rotational symmetry of 2H-MoTe2 without phase transition, we performed the Raman spectroscopy measurements. Supplementary Fig. 3 shows the Raman spectrum of pristine 2H-MoTe2, ferroelectric/2H-MoTe2 heterointerface before and after the SHG test. The Raman peaks display similar intensity and position, indicating no structural phase transition happens41,42,43.

Fig. 2: Photoresponse obtained in symmetry-breaking 2H-MoTe2.
figure 2

a Schematic description of the device structure. The P(VDF-TrFE) above channel is polarized into T-shape arrayed ferroelectric domains by applying bias through the PFM tip. θ is the polarization angle of the incident light. b The output characteristics (Isd-Vsd) of the device in Fig. 2c in the dark and under the illumination of 940 nm laser. c The spatial photocurrent mapping at 0 V under the illumination of 520 nm. The top panels show the optical image of the device (left) and patterned ferroelectric domains via PFM lithography mode (right). The yellow and red dashed boxes show the ferroelectric domains region and photocurrent region, respectively. The white T-shape dashed box in the photocurrent mapping corresponds to one T-shape ferroelectric domain in the PFM phase. The scale bar is 5 μm. d The linear polarization dependence of the photocurrent of pristine 2H-MoTe2 (top) and 2H-MoTe2 tuned by arrayed T-shaped ferroelectric domains (bottom). The excitation laser wavelength was 940 nm. The laser power is about 5 μW.

Full size image

Figure 2b shows the linear output characteristics in the dark and under illumination, showing a non-negligible short-circuit photocurrent Isc of 2.48 nA under illumination (Fig. 2b). The cross point of output curve in the dark and under illumination offers experimental evidence for the dominant mechanism in this device is the photovoltaic effect rather than photo-thermal-electric effect23. It should be noted that the photoresponse completely originates from the 2H-MoTe2, since P(VDF-TrFE) exhibits high transmittance across the visible to infrared spectrum. The photocurrent collected by electrodes can be considered as the sum of the photocurrent generated from each T-shape micro-region. The spatial photocurrent map enables a diagnostic of the photoresponse generation mechanism. The photocurrent mapping in Fig. 2c exhibits the photocurrent flows along the p-n region (consistent with the simulation results in Supplementary Fig. 4). The white T-shape dashed line in Fig. 2c illustrates the correspondence between the PFM phase and the photocurrent mapping. The reason why the photocurrent mapping image shows a latticed shape rather than a T-shape array is that the gaps between the T-shape ferroelectric domain are indistinguishable from the incident light with a focus spot size of 1 μm. The result obviously proves the photoelectric conversion process occurred in the channel part rather than the Schottky junction. Note that the polarization-sensitive photocurrent in this device is associated with the direction of T-shape ferroelectric domains (Supplementary Fig. 5). The light-polarization dependence of the photocurrent was measured as shown in Fig. 2d. Generally, the photocurrent in pristine 2H-MoTe2 shows light-polarization independence because of the isotropic crystal structure. However, the symmetry-breaking ferroelectric/2H-MoTe2 heterointerface exhibits obvious light-polarization dependence at zero bias.

Design and optimization for higher polarization sensitivity

The nonlinear photoresponse in our device can be described by a phenomenological model. In the symmetry-breaking ferroelectric/2H-MoTe2 heterointerface, the polarization-dependent nonlinear responses should be considered, and a time-dependent photocurrent can be expressed as44

$${j}_{i}(t)={\sum}_{{{{\rm{j}}}}}{\sigma }_{ij}^{(1)}{E}_{j}(t)+{\sum}_{{{{\rm{jk}}}}}{\sigma }_{ijk}^{(2)}{E}_{j}{E}_{k}(t)+{\sum}_{jkl}{\sigma }_{ijkl}^{(3)}{E}_{j}{E}_{k}{E}_{l}(t)+{{\mathrm{..}}}.$$
(1)

Where the second-order term σ(2) is the second-order nonlinear coefficient, and the E(t) is the incident light electric field.

Apparently, the photocurrent is proportional to the even powers of the incident light electric field in this symmetry-breaking ferroelectric/2H-MoTe2 heterointerface. For periodically oscillating fields with a frequency ω, the two Fourier components ±ω have a non-trivial probability of either constructively interfering to generate the SHG signal or destructively interfering to generate the intrinsic photogalvanic effect (that is BPVE)2. The intrinsic spontaneous photocurrent is decided by the interface symmetry of the heterointerface. Therefore, it provides a chance to engineer the symmetry of the ferroelectric/MoTe2 heterointerface and enhance infrared polarization-sensitivity by designing ferroelectric-domain patterning. The next section will delve into designing geometric patterns, aiming at enhancing device performance.

We observed that the photocurrent response reaches its maximum value when the light polarization angle is 0°. Based on this finding, we proceeded to calculate and design three devices, each featuring a unique nanodomain geometry pattern. As shown in Fig. 3a–c, three different devices made of MoTe2 are marked as device 1, device 2, and device 3. Device 1 is tuned by arrayed T-shape ferroelectric domains, device 2 is doped into an in-plane PN junction, and device 3 is doped with a combination of device 1 and device 2. In the device fabrication process, PFM technology was still used to write domain patterns and realize patterning doping. A color scale of a black and white photo was used to control the bias voltage of the PFM tip and then re-imaged in PFM mode, as shown in Supplementary Fig. 6.

Fig. 3: Design and optimization for higher polarization sensitivity.
figure 3

a, b, c Schematic of the device structure of device 1, device 2, and device 3. The MoTe2 doped by T-shape array ferroelectric domains is marked as device 1, the MoTe2 doped into in-plane PN junction is marked as device 2, and the MoTe2 doped by a combination nano-pattern of T-shape ferroelectric domains array and PN junction is marked as device 3. d, e, f Calculated surface potential maps of 2H-MoTe2 in device 1, device 2 and device 3. g, h, i Schematic of local photocurrent generated in 2H-MoTe2 around p-n interfaces of device 1, device 2 and device 3. The red arrow (I) represents the photocurrent collected at the metal electrodes. j, k, l Simulated (Sim.) polarization photocurrent of device 1, device 2, and device 3 at 0 V bias under illumination of 1550 nm incident light. Device 3 shows an obvious improvement in polarization photocurrent.

Full size image

The mechanism of the ferroelectric-patterning-doping MoTe2 is further investigated from a microscopic view. Broadly, both in-plane PN junctions and non-centrosymmetric materials exhibit a nonlinear transport process for their inversion asymmetric charge distributions45. Thus, the PN junction region is considered the fundamental unit for exploring the generation process of polarization photocurrent. Therefore, we fabricated an in-plane MoTe2 PN junction device and measured its polarization performance. The device structure and experiment results are shown in Supplementary Fig. 7. The maximum photocurrent response occurs for the light polarization direction parallel to the PN junction, suggesting that the photocurrent is more beneficial generated when the preferred momentum direction of photogenerated carriers is parallel to the built-in electric field. This phenomenon can be attributed to the momentum-matching relationship between the photons and the photogenerated carriers46. Based on the physical understanding of the artificial BPVE in the ferroelectric-domain-tuned MoTe2, we simulate the photoresponse of this device in the next step. The surface potential of MoTe2 was first calculated, as shown in Fig. 3d–f. The surface potential difference between P-type MoTe2 and N-type MoTe2 is approximately 0.78 V, indicating an efficient photon-generated carrier separation process in the junction region. To simulate the global photoresponse, a Shockley-Ramo-like formalism is utilized as follows47,48. Figure 3g–i describe the photocurrent flows along the p-n region in three different ferroelectric-doping-patterned devices. The photocurrent in linear response is a superposition (i.e., a weighted volume integral):

$$J=\int j({{{\bf{r}}}})r({{{\bf{r}}}})d{{{\bf{r}}}}$$
(2)

Where j(r) represents the local photocurrent density and r (r) represents the local collection efficiency.

Based on this model, we investigated the photocurrent values of three kinds of devices by sweeping polarization angles from 0° to 180°. Figure 3j–l depicts the linear polarization photoresponse of device 1, device 2, and device 3. Note that the device 3 shows a significant polarization photocurrent compared to device 1 and device 2. The improvement of the polarization ratio in device 3 can be attributed to additive effects resulting from the combination of the T-shape array ferroelectric domain and PN junctions. The detailed calculation process is shown in Supplementary Note 1, Note 2, and Note 3. Unless stated otherwise, the following discussion of the electrical and optoelectronic characteristics is based on the device 3 in Fig. 3c.

A colossal artificial BPVE in the optimized device

Based on the simulated and experimental results, we fabricated the optimized device that exhibits the best polarization photoresponse performance. The schematic of the device structure is shown in Fig. 4a. To clarify the intrinsic nature of photocurrent generation in our device, we measured the output curves under both dark and illuminated conditions. As shown in Fig. 4b, the 520 nm excitation power increases from dark to 120 μW. The inset displays the output curves under 1550 nm illumination, with power increasing from 180 nW to 3.4 mW. Typical spontaneous photocurrent behaviors are observed under laser illumination with a wavelength of both 520 nm and 1550 nm. For 520 nm, the responsivity (R) calculated by R = Iph/P reaches the largest value of 20 mA/W at the incident power density of 3 μW/cm2. The detectivity calculated by D* = RA1/2(2eIdark)1/2 is 2.4 × 109 Jones. The laser-power dependences of Jsc for both 520 nm and 1550 nm are shown in Fig. 4c. In nonlinear photoresponse, the photocurrent I can be proportional to the exciting light electric field intensity E (\(I\propto {E}^{2}\)). As shown in Fig. 4c, the photocurrent exhibits a linear increase with laser power and demonstrates a trend of square-root power dependence in the saturation regime. The crossover observed in the laser-power dependence indicates that the photocurrent in this device shows similar features with the shift current mechanism26. Furthermore, the maximum values of Jsc and Voc under 520 nm are 29.9 A/cm2 and 0.12 V (3 × 105 V/cm). Under 1550 nm illumination, the maximum values of Jsc and Voc are 6.2 mA/cm2 and 0.02 V (5 × 104 V/cm), respectively. The large Jsc and Voc can be attributed to the combined effect from every ferroelectric domain. 14 of the same devices in Supplementary Fig. 11 prove that our method can stably fabricate spontaneous photovoltaic effect devices with wonderful performance. Considering the low light absorptivity of MoTe2 in the near-infrared band, these performance indexes can be seen as satisfying achievements. Figure 4d and Supplementary Fig. 12 exhibit the spatial distribution of the photocurrent under 520 nm and 1550 nm, indicating that the polarization photocurrent originates from the channel rather than from the metal/semiconductor heterointerface49,50.

Fig. 4: A colossal artificial BPVE in the optimized device.
figure 4

a Schematic and measurement configuration of a MoTe2 device. 0° (90°) indicates the polarization angle of incident light is parallel (vertical) to electrodes. b The output characteristics of the device under different excitation powers (from dark to 120 μW) at 520 nm. The inset shows the output characteristics under different excitation powers (from 1.5 μW to 150 μW) of 1550 nm. c Laser power dependence of the photocurrent of the device under illumination of 520 nm (red dots) and 1550 nm (blue dots). In the low-power region, Jsc is proportional to P while Jsc is proportional to P0.5 in the high-power region. The low-power and high-power regions are relative definitions because of the different light absorptivity for 520 nm and 1550 nm. d The spatial photocurrent mapping at 0 V under the illumination of 520 nm. The top panels show the optical image of the device (left) and patterned ferroelectric domains via PFM lithography mode (right). The red dashed box shows the ferroelectric domains region and photocurrent region The scale bar is 5 μm.

Full size image

Enhanced near-infrared polarization photodetector with configurable polarity transition

Furthermore, we investigated the potential of this device for high-quality polarization detection in near-infrared wavebands. All the polarization measurements were based on device 3 in Fig. 4a and carried out at room temperature. Figure 5a exhibits the polarization photocurrent increases step-by-step with the light power increasing (The laser wavelength is 1270 nm.). For longer wavelength detection, Fig. 5b shows the photocurrent as a function of the linear polarization angle with different excitation power (The laser wavelength is 1550 nm). The photocurrent shows no direction change when the polarization angle changes. This is because the patterned ferroelectric domains acting on MoTe2 break the rotational symmetry of MoTe2, create a potential difference between the two electrodes, and drive photo-excited carriers to transport to be collected by the electrodes. Figure 5c shows the polarization ratio value decreases monotonically as the excitation power increases, which is defined as a ratio of photocurrent at 0° and 90° and extracted in Fig. 5b. In the low-power region, the PR value decreases rapidly and remains a value of ~2 gradually, indicating nonlinear photoresponse of the device21. In principle, the photocurrent in a few-layer 2H-MoTe2 is independent of the angle of light polarization. Therefore, the present results can be actual experimental evidence to support the BPVE behavior in our device. Figure 5d shows the output curves of the device in both dark and illumination. The incident light has the same power of 180 μW, with the polarization angle changing from 90° to 0°. The polarization photocurrent ratios, which are defined as Iθ/Iθ = 90°, at different bias voltages extracted from Fig. 5d are displayed in Fig. 5e. The polarization photocurrent ratio curve, aiming to find the optimized operating bias voltage for maximum PR, as a function of bias can be effectively fitted with a hyperbola. Consequently, the PR values of our device can vary from positive to negative values, as shown in Fig. 5e. At the polarity-transition point, the polarization value increases infinitely. However, in practice measurements, we hardly obtain an infinitely great polarization ratio value. The photocurrent at different polarization angles for the as large as possible PR was measured at the bias voltage close to the polarity-transition point. Thus, the PR value in our device actually reaches a maximum value of 310.23, as shown in Fig. 5f.

Fig. 5: Enhanced near-infrared polarization photodetector with configurable polarity transition.
figure 5

a The polarization photocurrent of 0° and 90° increase with the light power increasing step-by-step from zero to 0.37 W/cm2. The incident light is 1270 nm. b Anisotropic photocurrent response under different excitation powers at 1550 nm at 0 V bias shown in the polar plots. The incident power from P1 to P7 is 180 nW, 1.8 μW, 18 μW, 180 μW, 1.8 mW, and 3.4 mW. c Excitation power dependence of the polarization ratio extracted from Fig. 5b. In the low-power region, the PR value decreases sharply. d The output characteristics of devices under different polarization angle lights with the same excitation power (180 μW). e Polarization photocurrent ratio (symbols), which is defined as Iθ/Iθ=90°, extracted from Fig. 5d. The black solid line is fitted with a hyperbola model. The blue and yellow regions indicate the unipolar (PR  >  0) and bipolar (PR  <  0) PR values. f Polarization photocurrent ratio (Iθ/Iθ=90°) with a bias of −37 mV under illumination of 1550 nm. The laser power is about 180 μW. g, h, i Measured (symbols) and fitted (lines) photoresponse of the device at Vbias1 (Vbias1 = -37 mV), Vbias2 (Vbias2 = −49 mV) and Vbias3 (Vbias3 = −59 mV).

Full size image

Additionally, Fig. 5g–i shows the direction of photocurrent changes with the polarization angle at different biases. The device shows a high PR value close to +∞ when a bias of −37  mV (Vbias1) is applied. The photocurrent keeps a positive value and flows from source to drain at 0°. When a bias of −49 mV (Vbias2) is applied, the photocurrent shows a positive maximum value at 0° and a negative minimum value at 90°. When the polarization angle is 45°, no net current is generated. When a bias of −59 mV (Vbias3) is applied, the photocurrent changes direction totally, flowing from drain to source at 90°. The photoresponse with configurable polarity showcases the potential of our device used as a polarization-resolved and bias-selectable photodetector. Furthermore, the comparison of the related photodetectors based on various materials is shown in Supplementary Fig. 13. It is clear that the Jsc of our device is one order of magnitude higher than that in perovskites and ferroelectrics. In a two-dimensional symmetry-broken system, our device shows top-level Jsc and Voc values. The versatility of symmetry-breaking by patterned ferroelectric domains is not limited to 2H-MoTe2. Black phosphorus (BP) is a potential candidate for infrared polarization detection because of its narrow bandgap and intrinsic linear dichroism. Similar artificial BPVE could be induced in BP, as shown in Supplementary Fig. 14. The enhanced polarization photoresponse demonstrates the great potential improvement in the infrared polarization detection via fabricating devices proposed in this work.

Discussion

In summary, we have demonstrated a reliable strategy to manipulate the symmetry of low-dimensional materials via ferroelectric-doping patterns. The symmetry-breaking ferroelectric/MoTe2 heterointerface exhibits nonlinear optoelectronic phenomena, such as distorted SHG signals and photoelectric behavior similar to BPVE, which can be considered an artificial BPVE. A photodetector with both spontaneous photocurrent and infrared polarization detection capabilities has been achieved by elaborately simulation and fabrication. These devices in this work exhibit impressive short-circuit photocurrent intensity (Jsc = 29.9 A/cm2) and open-circuit voltage (~3 × 105 V/cm). Moreover, they show configurable polarity transition in infrared polarization detection with the PR tunable to unipolar (PR ≥ 1) or bipolar (PR ≤ −1), enabling direct measurement of the full-Stokes parameters in a single device. This work enriches the understanding of the photovoltaic conversion mechanism and highlights the immense potential of low-dimensional materials in infrared polarization detection. More importantly, symmetry-breaking engineering via ferroelectric nano-patterning has been established, providing a broad platform to manipulate the symmetry of low-dimensional materials and fabricate a series of polarization optoelectronic devices.

Methods

Device fabrication

Few-layer MoTe2 was mechanically exfoliated from the bulk single crystal and transferred onto 285 nm thick SiO2/Si substrates. The source and drain electrodes (chromium/gold (Cr/Au), 15/35 nm) were deposited in a vacuum chamber on the MoTe2 through standard electron-beam lithography and metal deposition techniques, followed by the lift-off process in acetone at room temperature. The 100 nm-thick P(VDF-TrFE) film was prepared by the spin-coating method. All the flake’s thickness is identified by AC mode (imaging in air) of atomic force microscope (AFM).

Piezoresponse force microscopy (PFM) measurement

The patterned ferroelectric domains are fabricated by lithography and vertical PFM imaging mode using a commercial Asylum Research system. Firstly, the mechanical response of P(VDF-TrFE) was measured by applying a voltage on a conductive tip (ASYELEC.01-R2-10). The ferroelectric polarization of the sample was modified by controlling the bias of the tip as it was rastered over the surface and then re-imaged in PFM mode. For P(VDF-TrFE) with a thickness of 50 nm, a voltage of about 10 V was applied on the PFM tip to ‘write’ the domains. The PFM phase images show the polarization state of P(VDF-TrFE) after such a piezoelectric lithography process.

Raman and second harmonic generation (SHG) measurement

The Raman measurements were illuminated with a 520 nm wavelength laser of 0.47 mW power under ×100 objective. The SHG and its polarization dependence measurements were performed using a 1064 nm picosecond laser and the SHG signal (532 nm) was collect by a photomultiplier tube. The incident light (1064 nm) power is about 10 mW under ×100 objective and the integral time is 1 s. The polarization angle of incident light is changed by rotating a half wavelength plate mechanically.

Electrical and optoelectronic characterization

All the electronic and optoelectronic measurements were performed at room temperature and under ambient conditions. A 520 nm and 1550 laser are normally incident on the sample area, and the diffraction-limited laser focus spot is estimated to be 1 μm and 3 μm, respectively. The linear polarization light is obtained by rotating a linear polarizer (Thorlabs). The polarizer uses a fixed polarizer and a rotating half-wave plate, and the laser energy does not change during the process of changing the polarization direction. The photocurrent of the device was recorded during the rotation process of the half-plate. The incident power of the laser is measured by a light power meter. The optoelectronic results were recorded by a picoammeter. The measurements were taken by an MStarter 200 optoelectronic measurement system from Maita Optoelectronic Technology Co., LTD.

Reporting summary

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