An ultra-sensitive neuromorphic vision sensor based on Janus structure

Abstract

The application discloses an ultra-sensitive neuromorphic visual sensor based on a Janus structure and belongs to the technical field of semiconductor neuromorphic vision. p The visual sensor of the application takes heavily doped silicon as a substrate, adopts hafnium-zirconium-oxygen material as a ferroelectric gate dielectric electric control layer, lays a few-layer two-dimensional material as a channel layer on the ferroelectric gate dielectric electric control layer, transfers an ultrathin perovskite nanosheet as a light control layer on the top, and inserts gold electrodes on both sides of the channel for signal input and output, thereby forming a Janus structure device composed of a substrate layer, a ferroelectric gate dielectric electric control layer, a two-dimensional semiconductor channel layer, an ultrathin perovskite light control layer and an electrode layer. The Janus structure device controls the energy band matching and the polarization field flipping characteristics of the upper and lower interfaces of the two-dimensional material, establishes a double-field synergistic carrier control mechanism of a light field + a ferroelectric polarization field, breaks through the performance limit of a single material, and constructs a neuromorphic visual sensor with high signal-to-noise ratio and high detection sensitivity in an extremely weak light environment.

Description

Technical Field

[0001] This invention belongs to the field of semiconductor optoelectronic device technology, specifically relating to an ultrasensitive neuromorphic vision sensor based on a two-dimensional material with a Janus structure, which is particularly suitable for ultrasensitive detection and adaptive recognition of target object features in extremely dark environments. Background Technology

[0002] Vision is a core function for intelligent organisms and smart agents to perceive the external environment. Over 80% of environmental information in the human perception system originates from the visual system, which possesses characteristics such as low redundancy, low power consumption, and high dynamic range. It can efficiently and adaptively process static and dynamic information and has excellent small-sample generalization ability and complex scene perception capabilities. As a core component of the visual perception system, visual sensors undertake the crucial task of capturing and converting light signals, and are key foundational devices in fields such as machine vision, artificial intelligence, optical imaging, national defense and security, environmental monitoring, and biomedical detection. With the rapid development of next-generation intelligent technologies, the industry is placing increasingly higher demands on the core performance of visual sensors, such as detection sensitivity, spectral response range, device integration, and operating power consumption. However, current research on visual sensors still faces many technical challenges: the difficulty of balancing high responsivity and low noise equivalent power, the inherent trade-off between sensitivity and response speed, difficulties in dark current and noise suppression, and limited spectral response range. Especially in extremely dim scenes, the contradiction between high sensitivity and low noise limits the accurate detection capability of weak light signals, becoming a core bottleneck restricting the application of photoelectric detection technology in visual sensors.

[0003] Two-dimensional materials, with their atomically thin thickness, high mobility, tunable band structure, and heterogeneous integration capabilities, can achieve excellent optoelectronic, storage, and computing performance, providing a new material foundation for the development of intelligent vision systems. In recent years, significant progress has been made in the research of neuromorphic vision sensors based on two-dimensional materials, successfully achieving preprocessing functions such as adaptive imaging, contrast enhancement, and dynamic detection. However, achieving or surpassing the perception capabilities of the human visual system in complex interactive environments requires extensive exploration. Especially under extremely dark conditions, neuromorphic vision devices are limited by the inherent limitations of photodetectors, exhibiting problems such as low sensitivity, high dark current noise, limited sensing range, and poor image quality. This significantly reduces their visual adaptability to changes in optical information in the external environment, hindering the application and development of visual sensing technology in extreme lighting scenarios. Therefore, overcoming the performance bottlenecks of photodetectors and developing novel neuromorphic vision sensors with integrated sensing, storage, and computing functions has become a key path to achieving human-like visual perception capabilities.

[0004] In summary, although some progress has been made in the research of visual sensors based on two-dimensional materials, the core technological bottlenecks in this field have not yet been overcome, and a series of technical defects and shortcomings urgently need to be addressed:

[0005] (1) High responsivity is incompatible with low noise equivalent power.

[0006] There is an inherent trade-off between high responsivity and low noise equivalent power (NEP) in visual sensors. Improving responsivity requires increasing internal gain, extending carrier lifetime, or optimizing the light absorption structure, but this amplifies shot noise, thermal noise, and 1 / f noise, leading to increased NEP and a deterioration of the detection limit. Conversely, reducing gain, shrinking the active region, or optimizing impedance matching can suppress noise, but this weakens responsivity and low-light detection capabilities. This contradiction makes it difficult for existing devices to balance high signal-to-noise ratio efficiency and low detection limit, severely limiting their performance breakthroughs in high signal-to-noise ratio applications such as low-light imaging.

[0007] (2) Insufficient detection capability in extremely dark environments

[0008] The fundamental reason for the insufficient photoelectric detection capability in extremely dark environments lies in the imbalance between the magnitude of the weak light signal and the device's inherent noise. Dark current, generated by thermal excitation, material defects, and surface states under no-light conditions, causes shot noise that significantly reduces the signal-to-noise ratio, drowning out the extremely weak photogenerated signal in the noise background. Although dark current can be suppressed through techniques such as low-temperature operation, structural optimization, or lock-in amplification, these methods often come with additional problems such as increased system complexity, power consumption, and cost. Therefore, how to achieve synergistic optimization of low dark current and low noise at room temperature, and overcome the intrinsic limitations of materials and the bottlenecks in signal extraction, remains a core challenge in the field of visual sensor detection, and is also the key to improving the detection capability in extremely dark environments.

[0009] (3) Limited spectral response range and simple device control mechanism

[0010] Existing photodetectors are mostly optimized for specific wavelengths, making it difficult to cover the entire ultraviolet-visible-infrared spectrum. The intrinsic absorption characteristics of materials limit the spectral response range, forcing multi-band detection systems to integrate multiple detectors, significantly increasing system complexity and cost. Simultaneously, the carrier control mechanisms of these devices are relatively simple, mainly relying on a single interface or external field to control carrier transport, failing to achieve coordinated control of the entire process of photogenerated carrier generation, separation, transport, and capture. This makes it difficult for a single device to simultaneously integrate high-sensitivity photosensing, low-noise signal output, and controllable electrical control functions, failing to meet the demands of next-generation intelligent vision sensing systems that are developing towards multi-functionality, high integration, and low power consumption. Summary of the Invention

[0011] In view of the shortcomings of the existing technology, the core technical problems to be solved by this invention, in order of importance, are as follows:

[0012] (1) Existing two-dimensional semiconductor-based vision sensors cannot achieve a balance between high responsivity and low noise equivalent power: Two-dimensional materials are limited by atomic-level thickness, resulting in low intrinsic light absorption efficiency. They typically rely on mechanisms such as photoconductivity gain or avalanche gain to improve responsivity, but this introduces significant additional noise. Furthermore, methods to suppress dark current through ferroelectric dielectrics and surface passivation often simultaneously reduce the device's light response efficiency. The two cannot be optimized in synergy, which fundamentally limits the upper limit of the device's detection sensitivity.

[0013] (2) Existing two-dimensional semiconductor-based vision sensors have insufficient detection capabilities and poor imaging in extremely dark environments: Existing two-dimensional material vision sensors cannot achieve effective and distinguishable signal detection in extremely dark / weak light scenarios at the nanowatt level. On the one hand, two-dimensional materials have weak intrinsic light absorption capabilities, and cannot generate sufficient photogenerated carriers to form a detectable photocurrent under extremely weak light detection conditions; on the other hand, the device's dark current and noise floor are too high, and the weak light signal generated by extremely weak light will be completely submerged by the background noise, which cannot meet the high-precision photoelectric detection requirements in extremely dark scenarios.

[0014] (3) Existing two-dimensional semiconductor-based vision sensors have limited spectral response range and simple control mechanism: Existing two-dimensional photoelectric sensors mostly rely on a single interface or a single external field to achieve carrier transport control, and cannot achieve coordinated control of the entire process of photogenerated carrier generation, separation, transport and capture; at the same time, it is difficult to achieve high-sensitivity sensing, low-noise output and controllable electrical control of optical signals in the same device, which cannot meet the needs of new photoelectric detection applications that integrate sensing, storage and computing and adapt to multiple scenarios.

[0015] To address the aforementioned technical shortcomings of existing two-dimensional semiconductor-based neuromorphic vision sensors, a universal high-sensitivity neuromorphic vision sensor based on a Janus dual-interface is provided. The "Janus-interface" is a design concept that achieves complementary and synergistic effects by combining materials with different functions at the upper and lower interfaces of a system. Its core feature is the spatial separation of heterogeneous components or functions to achieve coupled control of multiple interfaces. Through the complementary synergistic design of the upper and lower interfaces, it overcomes the industry bottleneck of incompatibility between high responsivity and low dark current / low noise. Simultaneously, it constructs a universal device architecture that can be adapted to various two-dimensional semiconductor channel materials, ultimately achieving high signal-to-noise ratio and high-precision photoelectric detection in extremely weak light environments, thus breaking through the performance limits of single materials.

[0016] The solution of the present invention is as follows:

[0017] The visual sensor of the present invention comprises, from bottom to top, a substrate layer, a ferroelectric gate dielectric control layer, a two-dimensional semiconductor channel layer, and an ultrathin perovskite optical control layer; the ferroelectric gate dielectric control layer is disposed on the lower surface of the two-dimensional semiconductor channel layer; the ultrathin perovskite optical control layer is disposed on the upper surface of the two-dimensional semiconductor channel layer; the ferroelectric gate dielectric control layer, the two-dimensional semiconductor channel layer, and the ultrathin perovskite optical control layer are stacked vertically to form a Janus dual-interface heterostructure; the material, structural parameters, spatial relationship, and connection relationship of each layer are as follows:

[0018] The substrate layer is heavily doped. p Type-1 silicon substrate; specifically, the substrate layer has a (100) crystal orientation. p The device consists of a doped low-resistivity single-crystal silicon substrate with no native oxide layer, serving as the mechanical support substrate and back gate. The ferroelectric gate dielectric control layer is constructed from hafnium zirconium oxide (HZO) material with a high dielectric constant, deposited on the upper surface of the substrate using atomic layer deposition (ALD). The two-dimensional semiconductor channel layer is constructed from few-layer MoTe2 nanosheets and other two-dimensional layered materials, laid flat on the upper surface of the oxide layer. The ultrathin perovskite optical control layer is prepared using solution recrystallization, followed by mechanical exfoliation to obtain thin nanosheets, which are then placed on the upper surface of the two-dimensional semiconductor channel layer using dry spot transfer technology. The electrode layer comprises a source electrode and a drain electrode, both made of gold. The source and drain electrodes are fabricated at opposite ends of the two-dimensional semiconductor channel layer, directly contacting the channel layer to form an electrical connection for electrical signal input and output.

[0019] Furthermore, the thickness of the ferroelectric gate dielectric electro-regulating layer is 10-20 nm; wherein, the ferroelectric gate dielectric electro-regulating layer is a hafnium zirconium oxide (HZO) based ferroelectric thin film, which is formed by alternating growth of HfO2 and ZrO2 in a 1:1 cyclic ratio through atomic layer deposition.

[0020] Furthermore, the two-dimensional semiconductor channel layer is composed of MoTe2 nanosheets prepared by mechanical exfoliation, which have bipolar transport properties and a layer thickness of 5-10 nm.

[0021] Furthermore, the ultrathin perovskite photocontrol layer is a perovskite nanosheet with a thickness of 10-20 nm; the ultrathin perovskite nanosheet is transferred to the upper surface of the two-dimensional semiconductor channel layer by PDMS-assisted dry transfer technology.

[0022] The method for fabricating the ultrasensitive neuromorphic visual sensor based on the Janus structure of this invention is as follows:

[0023] Step 1: pPreparation and pretreatment of heavily doped silicon substrates: selection p The material is a heavily doped silicon material as a substrate. p ++ type low-resistivity single crystal silicon wafer, crystal orientation (100), no natural oxide layer on the surface;

[0024] Step 2: Atomic layer deposition of ferroelectric oxide gate dielectric layer: A hafnium zirconium oxide thin film is grown on a pretreated silicon substrate using atomic layer deposition technology;

[0025] Step 3: Preparation and positioning transfer of MoTe2 nanosheets: Molybdenum telluride nanosheets were obtained on PDMS using mechanical exfoliation technology, and then transferred to the top of the hafnium zirconium oxide layer using a transfer device;

[0026] Step 4: Mechanical peeling and positioning transfer of the ultrathin perovskite photocontrol layer: The ultrathin perovskite layer is obtained on PDMS using mechanical peeling technology, and then transferred to the top of WSe2 using a transfer device;

[0027] Step 5: Fabrication of source and drain electrodes and post-processing of the device; using non-destructive electrode transfer technology, gold electrodes are fabricated at both ends of the ultrathin perovskite layer to form source and drain electrodes respectively, ultimately obtaining an ultrasensitive neuromorphic visual sensor based on the Janus structure.

[0028] Furthermore, step one specifically involves the following substrate cleaning process: First, the silicon wafer is cut into 0.8×0.8cm dimensions, and then ultrasonically cleaned in distilled water, propanol, and ethanol liquids for 3 minutes each time, repeating this process twice. After completion, the substrate surface is blown away with a dry nitrogen stream. Next, the substrate is rinsed with a 4% HF / H2O mixture to thoroughly remove the oxide film generated by natural oxidation. Then, the substrate is treated with an ultraviolet ozone device for 1 minute to fully hydroxylate the substrate surface, thereby creating ideal substrate conditions for subsequent thin film growth.

[0029] Further, step two specifically involves: using a four-channel atomic layer deposition (ALD) system to grow an HZO film; for the HfO2 portion, tetrakis(dimethylamino)hafnium (TDMA-Hf) precursor is used in combination with deionized water (H2O). Solid hafnium is selected to ensure the stability of the raw materials during growth and to prevent the liquid TDMA-Hf from easily decomposing during long-term use due to insufficient thermal stability; the hafnium source is heated to 90°C and the water-oxygen source is heated to 50°C. For the ZrO2 portion, tetrakis(diethylamino)zirconium (TDEA-Zr) precursor is used in combination with deionized water (H2O). The zirconium source is heated to 90°C and the water-oxygen source is heated to 50°C; high-purity nitrogen is used as the transport gas to maintain a high vacuum and impurity-free environment in the deposition chamber. Alternating layer-by-layer deposition is performed according to a 1:1 ratio of HfO2 to ZrO2 in a single cycle to ensure uniform distribution of film composition; the total number of 1:1 alternating depositions is precisely controlled to finally obtain an HZO film with a thickness of 15-20 nm.

[0030] Furthermore, step three specifically involves: using 3M tape to adhere a high-quality MoTe2 single crystal, and mechanically exfoliating the material to transfer it to a PDMS support; selecting thin-layer MoTe2 nanosheets of suitable thickness using an optical microscope, and using computer software to select a few-layer sample with a suitable size and thickness of 5-10 nm for later use; flipping and fixing the glass slide carrying PDMS, placing the substrate on which hafnium zirconium oxide is grown on the transfer stage and locking it with vacuum adsorption; adjusting the lifting device of the transfer device to precisely align and adhere the PDMS to the substrate, heating it at 110 ℃ for 6 minutes, and then adjusting the lifting device to separate the PDMS from the substrate, thus completing the positioning and transfer of MoTe2 nanosheets on the hafnium zirconium oxide layer.

[0031] Furthermore, step four specifically involves: using 3M tape to adhere bulk perovskite single crystals, and mechanically peeling off ultrathin perovskite nanosheets to a PDMS carrier; selecting transparent thin layers using an optical microscope, and using computer software to select films of suitable size and thickness of 10-20 nm for later use; flipping and fixing the glass slide carrying PDMS, placing the silicon wafer on the transfer stage and locking it with vacuum adsorption; adjusting the lifting device of the transfer apparatus to precisely align and adhere the PDMS to the substrate, heating at 90 ℃ for 10 minutes, and then adjusting the lifting device to separate the PDMS from the substrate, thus completing the positioning and transfer of ultrathin perovskite on the MoTe2 surface.

[0032] Furthermore, step five specifically involves: The device fabrication employs a non-destructive electrode transfer technique to avoid damage to the perovskite photocontrol layer caused by traditional micro / nano fabrication processes; a metal thin film is pre-deposited on the substrate and shaped into a rectangular form as electrodes. Using a micromanipulator on a microscope stage, the shaped rectangular electrodes are lifted and transferred to both ends of the channel layer. The probe tip is then lowered to allow the electrodes to adhere to the heterojunction sample, thus fabricating the source and drain electrodes. Repeating the above operations completes the fabrication of the ultrasensitive neuromorphic visual sensor based on the Janus structure.

[0033] In this invention, "Janus-interface" is a design concept that achieves complementary and synergistic effects by combining materials with different functions on the upper and lower interfaces of a system (molybdenum ditelluride is selected as the channel material, and the upper and lower surfaces of the channel layer are respectively: an ultra-thin perovskite photocontrol layer and a ferroelectric polarization gate dielectric control layer). Its core feature is to achieve multi-interface coupling and control through the spatial separation of heterogeneous components or functions. Through the synergistic design of complementary functions of the upper and lower interfaces, it breaks through the industry bottleneck of the incompatibility between high responsivity and low dark current and low noise. At the same time, it builds a universal device architecture that can be adapted to a variety of two-dimensional semiconductor channel materials, and finally achieves high signal-to-noise ratio and high-precision photoelectric detection in extremely weak light environments, thereby breaking through the performance limits of a single material.

[0034] The advantages of this invention compared to the prior art are as follows:

[0035] This invention addresses the core problem of the inability to synergistically optimize high responsivity and low noise in existing devices. It employs a Janus dual-interface vertical structure design, utilizing a ferroelectric gate dielectric control layer, a two-dimensional semiconductor channel, and an ultrathin perovskite photocontrol layer. By leveraging the excellent photosensitive properties of the top-layer ultrathin perovskite, the light absorption efficiency and photogenerated carrier injection efficiency are significantly enhanced, thereby improving the device's photoresponsivity. Simultaneously, by utilizing the spontaneous polarization effect and polarization-assisted carrier trapping effect of the bottom HZO ferroelectric gate dielectric, the free carriers in the channel are significantly depleted, and dark current and background noise are suppressed. This achieves synergistic optimization of high responsivity and low noise equivalent power in a single device.

[0036] This invention addresses the problem of insufficient detection capability in extremely weak light in existing devices by employing a dual-interface complementary and synergistic approach: (1) The top photosensitive perovskite and the two-dimensional channel form a Type-II band arrangement, which spatially forces photogenerated carriers to be separated into two different material layers. This separation process driven by the built-in electric field is very rapid and efficient, greatly suppressing the radiative recombination of electrons and holes and significantly reducing dark current; (2) The bottom ferroelectric gate dielectric layer utilizes the polarization-assisted interface effect to suppress the generation of dark current, significantly reducing the device's noise equivalent power and improving the specific detectivity, achieving nanowatt-level resolvable detection of ultra-weak light signals, and meeting the high-sensitivity photoelectric detection requirements in extremely dim scenes.

[0037] This invention addresses the limitations of existing two-dimensional semiconductor-based visual sensors, which suffer from limited spectral response ranges and singular control mechanisms. The effective bandgap of a Type-II heterojunction is determined by the band edge difference between the two materials. By selecting combinations of perovskite and two-dimensional materials with different bandgap values, the operating wavelength range of the device can be precisely controlled without altering the chemical composition of the materials. A multi-field synergistic carrier control mechanism is established, combining optical field control (top photosensitive perovskite layer) and ferroelectric polarization control (bottom ferroelectric gate dielectric layer). Reversible photocontrolled doping of the two-dimensional channel is achieved through illumination, while non-volatile control of channel carriers is achieved through ferroelectric polarization. This enables precise control of the entire process of photogenerated carrier generation, separation, transport, and capture. High-sensitivity photosensing, low-noise signal output, and controllable electrical control functions are integrated into a single device, providing a hardware foundation for the multifunctional and integrated application of the device.

[0038] Compared with existing two-dimensional semiconductor-based vision sensor technologies, this invention, based on an innovative Janus dual-interface vertical stacking universal architecture, overcomes the inherent bottlenecks of existing technologies in terms of material matching, structural design, and carrier modulation mechanisms. It exhibits significant technical advantages and excellent technical effects in terms of detection performance, structural versatility, operational stability, and industrial applicability. Through the complementary functional design of the upper and lower Janus interfaces, this invention fundamentally solves the core industry contradiction of the incompatibility between high responsivity and low dark current / low noise in existing technologies, achieving a leapfrog improvement in extremely weak light detection performance. It addresses the core pain points of existing technologies, such as insufficient detection sensitivity in extremely dark environments and the easy submersion of effective signals by background noise.

[0039] This invention constructs a universal device architecture, overcoming the limitations of existing high-performance 2D vision sensors that require customized design for a single channel material, and possesses excellent material adaptability and wide spectral response capabilities. The device architecture of this invention uses replaceable 2D semiconductor materials as the device channel, rather than being customized for a single material. It is adaptable to 2D semiconductor materials with different band gaps and transport properties, such as WSe2, MoS2, and Bi2O2Se. Performance verification was completed using a 5-10nm thick multilayer MoTe2 as a typical embodiment. The channel material can be flexibly changed according to the target detection band without significant adjustments to the main device structure, greatly expanding the applicable scenarios of the device and solving the problems of poor universality and narrow spectral response range of existing devices.

[0040] This invention addresses the inherent defects of traditional perovskite-two-dimensional material composite devices at the underlying structural level, significantly improving the signal-to-noise ratio, operational stability, and batch repeatability. This invention replaces traditional SiO2 with a HZO ferroelectric thin film as the gate dielectric layer. HZO possesses a much higher dielectric constant and denser film properties than SiO2, effectively suppressing potential carrier leakage paths in the perovskite-two-dimensional material heterojunction. Simultaneously, it greatly enhances the gate electric field's ability to regulate carriers in the two-dimensional channel, avoiding the disordered carrier transport problem caused by interface defects in traditional SiO2-based devices. This significantly reduces the defect state density at the heterojunction interface, allowing the device to operate stably for extended periods in air, and significantly reducing batch-to-batch performance deviations. This solves the industry pain points of low signal-to-noise ratio, poor operational stability, and poor batch repeatability in existing composite devices.

[0041] The device fabrication of this invention employs a non-destructive electrode transfer technique, which avoids damage to the perovskite material during device fabrication, ensures the intrinsic properties of the sample, and provides a new approach for perovskite device fabrication. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating the fabrication process of the ultrasensitive neuromorphic visual sensor based on the Janus structure prepared in this invention.

[0043] Figure 2 A cross-sectional view of the structure of the ultrasensitive neuromorphic visual sensor based on the Janus structure prepared in an embodiment of the present invention;

[0044] Figure 3 This is a graph showing the extremely weak light time-resolved response of the device in an embodiment of the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the following examples provide a more detailed description of the invention. It should be noted that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0046] See Appendix for Example 1 Figure 1 This is a schematic diagram of the fabrication process of the ultrasensitive neuromorphic visual sensor based on the Janus structure provided in this embodiment;

[0047] S1: Select heavily doped p Type silicon is used as the substrate, with a crystal orientation of (100) and no native oxide layer on the surface; S2: In heavily doped... pOn a silicon substrate, a high dielectric constant HZO ferroelectric gate dielectric layer is prepared by atomic layer deposition; S3: Two-dimensional layered materials such as few-layer molybdenum telluride nanosheets are prepared by mechanical exfoliation and laid flat on the upper surface of the HZO ferroelectric gate dielectric layer as the semiconductor channel layer of the device; S4: Few-layer ultrathin perovskite nanosheets are prepared by mechanical exfoliation as the photosensitive control layer and laid flat on the two-dimensional semiconductor channel layer; S5: Source and drain electrode layers are prepared on both sides of the semiconductor channel using non-destructive electrode transfer technology.

[0048] See Appendix for Example 2 Figure 2 This is a cross-sectional schematic diagram of the ultrasensitive neuromorphic vision sensor based on the Janus structure provided in this embodiment:

[0049] The Janus dual-interface vertical structure design, consisting of a ferroelectric gate dielectric control layer, a two-dimensional semiconductor channel, and an ultrathin perovskite photocontrol layer, leverages the excellent light absorption characteristics of the top ultrathin perovskite photosensitive layer to significantly enhance light absorption efficiency and photogenerated carrier injection efficiency, thereby improving the device's photoresponsivity. Simultaneously, by utilizing the spontaneous polarization effect and polarization-assisted carrier trapping effect of the bottom HZO ferroelectric gate dielectric, it significantly depletes free carriers in the channel, suppresses dark current and background noise, ultimately overcoming the bottleneck of the incompatibility between high responsivity and low dark current in traditional two-dimensional vision sensors within a single device. The extremely weak light detection response capability of the device fabricated in this invention was tested.

[0050] The device exhibits excellent resolution of extremely weak optical signals and stable cyclic operation at 1.8 nW / cm². 2 It can generate a reversible photoelectric response with a high signal-to-noise ratio under extremely weak light irradiation, which can meet the needs of high-sensitivity photoelectric detection and neuromorphic visual sensing in extremely dark environments.

[0051] Test scheme: Testing the device's extremely weak light time-resolved response characteristics at room temperature and 10°C. -6 The Torr vacuum level test is performed within a closed-loop probe station, completely isolating the system from stray light and electromagnetic interference. During the test, the source is grounded as a potential reference point, and the source-drain bias voltage is... V ds Fixed at 0.1V, gate voltage V g Fixed at 0, heavily doped p A silicon substrate was used as the bottom gate electrode; a 1.8 nW / cm² low-light laser, precisely calibrated by an optical power meter, was used as the incident light source. The source and drain currents of the device were recorded in real time using a semiconductor parameter analyzer over five consecutive cycles. I ds The relationship between the device and the test time was used to obtain the time-resolved response curve for extremely weak light.

[0052] Test results are as follows Figure 3As shown, the horizontal axis of the curve obtained from the test represents the test time in seconds, and the vertical axis represents the source and leakage current of the device. I ds The unit is ampere (A). The curve shows that the device exhibits stable photoelectric response behavior in five consecutive optical switching cycles: when the light is on, the device... I ds From approximately 6×10 -10 The dark baseline of A rapidly rose to a maximum of 1×10⁻⁶. -9 Peak photocurrent of A; device after illumination is turned off. I ds It can smoothly return to the initial dark state baseline without significant performance degradation or baseline drift. The response trend and peak current of the five cycles remain almost identical to the dark state baseline.

[0053] Results Analysis: Figure 3 It can be seen that this device is effective at 1.8 nW / cm². 2 Extremely weak light can produce a clear, distinguishable, high signal-to-noise ratio photoelectric switching response, consistent with the operating characteristics of high-sensitivity phototransistors. Even under extremely weak light illumination, the device can still generate an amplitude of 4 × 10⁻⁶. -10 The photocurrent response of the A-type device exhibits excellent extremely weak light detection capability. This characteristic stems from the synergistic control mechanism of the Janus dual interfaces: the ultrathin perovskite photosensitive layer on the upper interface can efficiently absorb extremely weak incident light photons and inject photogenerated carriers into the two-dimensional channel through the Type-II bandgap arrangement, compensating for the weak intrinsic light absorption of two-dimensional materials; the HZO ferroelectric gate dielectric layer on the lower interface can suppress the device's dark current to an extremely low level, providing an ultra-low noise substrate for resolving extremely weak light signals, fully demonstrating the device's excellent extremely weak light detection performance and cycle stability.

[0054] The core innovation of this invention lies in proposing a general-purpose, high-sensitivity neuromorphic visual sensor architecture based on Janus dual-interface engineering. Through the design of a heterogeneous interface with complementary upper and lower functions, it breaks through the industry bottleneck of the incompatibility between high responsivity and low dark current in existing two-dimensional visual sensors, and achieves high-sensitivity detection in extremely low light environments. The key technical points and protection points of this invention are described below in order of importance.

[0055] The core technical challenge and primary protection point of this invention is the overall architecture of a general-purpose vertically stacked neuromorphic vision sensor based on Janus dual-interface engineering. This invention abandons the existing design approach of single-interface control and single-material customization, and innovatively designs a bottom-up, heavily doped structure. pA vertically stacked architecture consisting of a silicon substrate, an HZO ferroelectric gate dielectric layer, a two-dimensional semiconductor channel layer, and an ultrathin perovskite photosensitive layer is proposed. Completely complementary dual heterojunctions are constructed on the upper and lower sides of the two-dimensional semiconductor channel, achieving synergistic control of optical gain enhancement and dark current suppression from the underlying architecture. This fundamentally solves the core contradiction in existing technologies where high responsivity and low noise cannot be simultaneously achieved. The corresponding core protection scope covers the overall structure of the Janus dual-interface vertically stacked device based on the HZO ferroelectric gate dielectric control layer-general-purpose two-dimensional semiconductor channel-ultrathin perovskite optical control layer, as well as all visual sensor applications based on this structure.

[0056] The second priority technical key point and protection point of this invention is the carrier full-process control mechanism with dual-interface functional synergy. This invention establishes a carrier full-process control mechanism with synergy between the optical field and the ferroelectric polarization field through band matching and polarization field control of the upper and lower interfaces. The upper interface, through the Type-II band arrangement formed by the ultrathin perovskite and two-dimensional semiconductor channel, achieves efficient absorption of incident light, efficient spatial separation and directional injection of photogenerated carriers, compensating for the inherent weak light absorption of two-dimensional materials and significantly improving the photoresponse gain. The lower interface, through the spontaneous polarization effect of the HZO ferroelectric layer and the ferroelectric interface-assisted carrier trapping effect, effectively suppresses dark current and background noise. The synergy of these two aspects achieves a breakthrough in both high responsivity and low dark current performance. The corresponding protection scope includes the optical gain control method based on the Type-II band arrangement designed with this dual interface, the dark current suppression method for ferroelectric polarization-assisted carrier trapping, and the application of the synergistic carrier control mechanism in visual sensors.

[0057] The third priority technical key point and protection point of this invention is the universal channel design adaptable to multiple types of two-dimensional semiconductors. The device architecture of this invention is not customized for a single two-dimensional semiconductor material. Using a 5-10 nm thick multilayer MoTe2 as a typical verification example, it can adapt to two-dimensional semiconductor materials with different band gaps and transport properties, such as WSe2, MoS2, Bi2O2Se, and black phosphorus, as the channel layer. Only the corresponding ultrathin perovskite optical control layer needs to be matched according to the band structure of the channel material; no adjustment to the main device architecture is required. This overcomes the limitations of customized design in existing high-performance two-dimensional neuromorphic vision sensors, significantly improving the universality and scalability of the device architecture. The corresponding protection scope includes the universal adaptation design of this architecture to different types of two-dimensional semiconductor channel materials, as well as variations of the same architecture vision sensor based on different two-dimensional channel materials.

[0058] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.

Claims

1. A highly sensitive neuromorphic visual sensor based on the Janus structure, characterized in that, The sensor, from bottom to top, comprises a substrate layer, a ferroelectric gate dielectric control layer, a two-dimensional semiconductor channel layer, and an ultrathin perovskite optical control layer; The ferroelectric gate dielectric control layer is disposed on the lower surface of the two-dimensional semiconductor channel layer; The ultrathin perovskite photocontrol layer is disposed on the upper surface of the two-dimensional semiconductor channel layer; The ferroelectric gate dielectric control layer, the two-dimensional semiconductor channel layer, and the ultrathin perovskite optical control layer are stacked in the vertical direction to form a Janus dual-interface heterostructure. The materials, structural parameters, spatial relationships, and connections of each layer are as follows: The substrate layer is heavily doped. p Type-1 silicon substrate; specifically, the substrate layer has a (100) crystal orientation. p ++Doped low-resistivity single-crystal silicon substrate, without native surface oxide layer, serves as the mechanical support substrate and back gate of the device; The ferroelectric gate dielectric control layer is constructed by selecting hafnium zirconium oxide (HZO) material with a high dielectric constant as the gate dielectric layer of the device, and depositing it on the upper surface of the substrate layer using atomic layer deposition technology. The two-dimensional semiconductor channel layer is constructed by selecting two-dimensional layered materials such as few-layer molybdenum telluride (MoTe2) nanosheets and laying them flat on the upper surface of the oxide layer as the semiconductor channel layer of the device. The ultrathin perovskite photocontrol layer is prepared by solution recrystallization and then mechanically exfoliated to obtain thin nanosheets, which are then placed on the upper surface of the two-dimensional semiconductor channel layer using a dry spot transfer technique. The electrode layer includes a source electrode and a drain electrode, both of which are made of gold. The source electrode and the drain electrode are respectively fabricated at both ends of the two-dimensional semiconductor channel layer and are in direct contact with the two-dimensional semiconductor channel layer to form an electrical connection for the electrical signal input and output of the device.

2. The ultrasensitive neuromorphic visual sensor based on the Janus structure according to claim 1, characterized in that, The thickness of the ferroelectric gate dielectric control layer is 10-20 nm; wherein, the ferroelectric gate dielectric control layer is a hafnium zirconium oxide (HZO) based ferroelectric thin film, which is formed by alternating growth of HfO2 and ZrO2 in a 1:1 cyclic ratio through atomic layer deposition.

3. The ultrasensitive neuromorphic visual sensor based on the Janus structure according to claim 1, characterized in that, The two-dimensional semiconductor channel layer is composed of few-layer MoTe2 nanosheets prepared by mechanical exfoliation. The MoTe2 nanosheets have bipolar transport properties and a layer thickness of 5-10 nm.

4. The ultrasensitive neuromorphic visual sensor based on the Janus structure according to claim 1, characterized in that, The ultrathin perovskite photocontrol layer is a perovskite nanosheet with a thickness of 10-20 nm; the ultrathin perovskite nanosheet is transferred to the upper surface of the two-dimensional semiconductor channel layer by PDMS-assisted dry transfer technology.

5. The method for fabricating a highly sensitive neuromorphic visual sensor based on a Janus structure according to any one of claims 1 to 4, characterized in that, The method is as follows: Step 1: p Preparation and pretreatment of heavily doped silicon substrates: selection p The material is a heavily doped silicon material as a substrate. p ++ type low-resistivity single crystal silicon wafer, crystal orientation (100), no natural oxide layer on the surface; Step 2: Atomic layer deposition of ferroelectric oxide gate dielectric layer: A hafnium zirconium oxide thin film is grown on a pretreated silicon substrate using atomic layer deposition technology; Step 3: Preparation and positioning transfer of MoTe2 nanosheets: Molybdenum telluride nanosheets were obtained on PDMS using mechanical exfoliation technology, and then transferred to the top of the hafnium zirconium oxide layer using a transfer device; Step 4: Mechanical exfoliation and positioning transfer of the ultrathin perovskite photocontrol layer: An ultrathin perovskite layer is obtained on PDMS using mechanical exfoliation technology, and then transferred to the top of the MoTe2 nanosheet using a transfer device; Step 5: Fabrication of source and drain electrodes and post-processing of the device; using non-destructive electrode transfer technology, gold electrodes are fabricated at both ends of the ultrathin perovskite layer to form source and drain electrodes respectively, ultimately obtaining an ultrasensitive neuromorphic visual sensor based on the Janus structure.

6. The method for fabricating a highly sensitive neuromorphic visual sensor based on the Janus structure according to claim 5, characterized in that, Step one is specifically as follows: Substrate cleaning process: First, the silicon wafer is cut into 0.8×0.8 cm specifications, and then placed in distilled water, propanol and ethanol liquid for ultrasonic cleaning for 3 minutes each time. This process is performed twice. After completion, the substrate surface is blown with dry nitrogen gas. Then, the substrate is rinsed with 4% HF / H2O mixture to thoroughly remove the oxide film generated by natural oxidation. Afterward, the substrate is treated with ultraviolet ozone equipment for 1 minute to fully hydroxylate the substrate surface, thereby creating ideal substrate conditions for subsequent thin film growth.

7. The method for fabricating a highly sensitive neuromorphic visual sensor based on the Janus structure according to claim 5, characterized in that, Step two specifically involves: using a four-channel atomic layer deposition (ALD) system to grow HZO films; for the HfO2 portion, tetrakis(dimethylamino)hafnium (TDMA-Hf) precursor is used in combination with deionized water (H2O). Solid hafnium is selected to ensure the stability of the raw materials during growth and to prevent the liquid TDMA-Hf from easily decomposing during long-term use due to insufficient thermal stability; the hafnium source is heated to 90°C and the water-oxygen source is heated to 50°C. For the ZrO2 portion, tetrakis(diethylamino)zirconium (TDEA-Zr) precursor is used in combination with deionized water (H2O). The zirconium source is heated to 90°C and the water-oxygen source is heated to 50°C; high-purity nitrogen is used as the transport gas to maintain a high vacuum and impurity-free environment in the deposition chamber. Alternating layer-by-layer deposition is performed according to a 1:1 ratio of HfO2 to ZrO2 in a single cycle to ensure uniform distribution of film composition; the total number of 1:1 alternating depositions is precisely controlled to finally obtain an HZO film with a thickness of 15-20 nm.

8. The method for fabricating a highly sensitive neuromorphic visual sensor based on the Janus structure according to claim 5, characterized in that, Step three specifically involves: using 3M tape to adhere a high-quality MoTe2 single crystal, and mechanically peeling it to transfer the material to a PDMS support; selecting a thin layer of MoTe2 nanosheets of suitable thickness using an optical microscope, and using computer software to select a few-layer sample of suitable size and thickness of 5-10 nm for later use; flipping and fixing the glass slide carrying PDMS, placing the substrate on which hafnium zirconium oxide is grown on the transfer stage and locking it with vacuum adsorption; adjusting the lifting device of the transfer device to precisely align and adhere the PDMS to the substrate, heating it at 110°C for 6 minutes, and then adjusting the lifting device to separate the PDMS from the substrate, thus completing the positioning and transfer of MoTe2 nanosheets on the hafnium zirconium oxide layer.

9. The method for fabricating a highly sensitive neuromorphic visual sensor based on the Janus structure according to claim 5, characterized in that, Step four specifically involves: using 3M tape to adhere bulk perovskite single crystals, and transferring ultrathin perovskite nanosheets to a PDMS carrier via mechanical exfoliation; selecting transparent thin layers using an optical microscope, and using computer software to select perovskite nanosheets of suitable size and thickness (10-20 nm) for later use; flipping and fixing the glass slide carrying PDMS, placing the silicon wafer on the transfer stage and locking it with vacuum adsorption; adjusting the lifting device of the transfer apparatus to precisely align and adhere the PDMS to the substrate, heating at 90 ℃ for 10 minutes, and then adjusting the lifting device to separate the PDMS from the substrate, thus completing the positioning and transfer of ultrathin perovskite on the MoTe2 surface.

10. The method for fabricating a highly sensitive neuromorphic visual sensor based on the Janus structure according to claim 5, characterized in that, Step five specifically involves: The device fabrication employs a non-destructive electrode transfer technique to avoid damage to the perovskite photocontrol layer caused by traditional micro / nano fabrication processes; a metal thin film is pre-deposited on the substrate and shaped into a rectangular form as electrodes. On a microscope stage, a micromanipulator is used to lift the shaped rectangular electrodes and transfer them to both ends of the channel layer. The probe tip is then lowered to allow the electrodes to adhere to the heterojunction sample, thus fabricating the source and drain electrodes. By repeating the above steps, the fabrication of an ultrasensitive neuromorphic visual sensor based on the Janus structure can be completed.

Images