Optoelectronic synapse device for visual-olfactory fusion of gas identification and preparation method thereof

Abstract

The application discloses a gas recognition visual-olfactory fusion optoelectronic synapse device and a preparation method thereof, and belongs to the technical field of semiconductor devices. The optoelectronic synapse device comprises a substrate, a two-dimensional material floating gate heterojunction and source and drain electrodes. The heterojunction at least comprises a channel layer and a floating gate layer stacked thereon, and forms a Type-II energy band structure, and the interface has a defect state for carrier capture and release. The channel layer generates photo-generated carriers in response to light pulses, and the floating gate layer responds to gas stimulation and changes the charge state through gas adsorption / desorption; by applying a programmable light pulse stimulus to the channel layer, the adsorption or desorption process of gas molecules on the floating gate layer and the charge state thereof are dynamically modulated, a dynamic electrical response signal fusing visual and olfactory information is output, and gas recognition is realized. The application significantly improves the gas recognition accuracy from 52.24% of a single mode to 98.27% by using a dual-mode fusion feature, and realizes the synergy of miniaturization, low power consumption and high-precision perception.

Description

Technical Field

[0001] This application relates to the field of semiconductor device technology, and more specifically to a gas recognition-smell fusion optoelectronic synaptic device and its fabrication method. Background Technology

[0002] Artificial olfaction technology has broad application prospects in environmental monitoring, medical diagnosis, industrial safety, and consumer electronics, and is particularly crucial for promoting the intelligent development of chip-level applications such as microrobots and implantable devices. These emerging applications place higher demands on sensor performance, driving artificial olfaction technology towards miniaturization, low power consumption, high sensitivity, and high recognition accuracy.

[0003] Currently, mainstream artificial olfaction systems primarily rely on two technological approaches. The first involves using an array of multiple independent metal-oxide-semiconductor (MOS) sensors, such as those from Alpha MOS and Airsense. Each sensor exhibits response characteristics to different gas types. Gas identification is achieved by analyzing the overall response pattern of the array using pattern recognition algorithms. However, each sensor unit in this approach typically consumes significant power (several milliwatts), and the entire array system is bulky (often tens of square centimeters), making it difficult to integrate into miniaturized platforms with strict space and power consumption constraints.

[0004] The second approach is on-chip integrated sensing. This approach attempts to integrate thousands or even tens of thousands of sensing units within a chip area of ​​a few square centimeters, constructing hard-coded characteristic responses through material composition gradients or operating parameter gradients (such as temperature and voltage). While this reduces the size to some extent, it inevitably faces inherent problems such as complex interface circuits and cumbersome fabrication processes. More importantly, whether it's a discrete array or an on-chip integrated solution, data processing heavily relies on the traditional von Neumann computing architecture, where sensing, storage, and computing units are physically separated. Frequent data transfer between the sensor and the central processing unit leads to a severe "memory wall" problem, exacerbating the system's hardware cost and energy consumption bottlenecks, thus limiting its widespread application in resource-constrained scenarios.

[0005] Two-dimensional transition metal chalcogenides and their heterostructures represent a promising neuromorphic bimodal sensing material for addressing the aforementioned challenges. In recent years, researchers have developed two-dimensional material devices capable of simultaneously responding to optical and gaseous signals, achieving preliminary device-level simulations of visual-olfactory cross-modal interactions. However, existing research largely focuses on simple superposition responses to two physical stimuli; combining this cross-modal perception with higher-level neuromorphic cognitive functions (such as learning, memory, and decision-making) remains a challenge.

[0006] Therefore, there is an urgent need for a gas recognition-smell fusion optoelectronic synaptic device and its fabrication method to solve the technical problems in the existing technology. Summary of the Invention

[0007] The purpose of this application is to provide a visual-olfactory fusion optoelectronic synaptic device for gas recognition and its fabrication method, which can solve at least one of the aforementioned technical problems. The specific solution is as follows:

[0008] This application provides a gas recognition vision-olfaction fusion optoelectronic synaptic device, comprising:

[0009] Substrate;

[0010] A two-dimensional material floating gate heterojunction is disposed on the substrate; the two-dimensional material floating gate heterojunction includes at least: a channel layer and a floating gate layer stacked on the channel layer;

[0011] The source electrode and drain electrode are electrically connected to the channel layer;

[0012] The floating gate layer is used to respond to gas stimulation by changing the charge state through the adsorption or desorption of gas molecules; the channel layer is used to respond to light pulse stimulation to generate photogenerated carriers.

[0013] The channel layer and the floating gate layer constitute a Type-II band heterojunction, and the interface between the floating gate layer and the channel layer has defect states for carrier capture and release.

[0014] By applying programmable light pulse stimulation to the channel layer, the adsorption or desorption process of gas molecules on the floating grid layer and the charge state of the floating grid layer are dynamically modulated, and a dynamic electrical response signal that integrates visual and olfactory information is output to achieve gas recognition.

[0015] Furthermore, the channel layer and the floating gate layer are bonded together by van der Waals forces to form the two-dimensional material floating gate heterojunction.

[0016] Furthermore, the floating gate layer satisfies the following condition:

[0017] The work function is adjustable;

[0018] The bandgap width is 1.5–2.5 eV;

[0019] It has a layered structure and a specific surface area of ​​not less than 500 m² / g;

[0020] Furthermore, the surface of the floating gate layer has surface active sites for the adsorption of gas molecules.

[0021] Furthermore, the channel layer is a two-dimensional semiconductor material layer with a direct bandgap, satisfying the following conditions:

[0022] Single-layer or multi-layer two-dimensional structure;

[0023] Bandgap of 1.1-1.9 eV; high carrier mobility >50 cm² / Vs; low dark current <1 nA;

[0024] Fermi level is adjustable;

[0025] Furthermore, the surface of the channel layer has surface active sites for the adsorption of gas molecules.

[0026] Furthermore, the channel layer and the floating gate layer satisfy the following conditions:

[0027] The difference in work function between the interface between the channel layer and the floating gate layer ranges from 0.3 to 0.8 eV;

[0028] The mismatch rate between the lattice constants of the channel layer and the floating gate layer is <5%.

[0029] Furthermore, the channel layer is made of any one of WSe2, WS2, black phosphorus, or MoS2; the floating grid layer is made of any one of MoSe2, WS2, Ta2NiSe5, or MoS2.

[0030] Furthermore, the channel layer is made of MoS2, and the floating gate layer is made of SnS2.

[0031] Furthermore, the programmable light pulse-stimulated light source satisfies the following condition:

[0032] The wavelength range is 405–785 nm;

[0033] The power density range is 10–100 μW / mm².

[0034] The optical pulse width is adjustable within 0.1–1 s;

[0035] Gas pulse width: adjustable from 1 to 10 seconds;

[0036] Optical duty cycle range: continuously adjustable from 0.1 to 0.9.

[0037] Furthermore, the photosynaptic device can be switched to operate in one of the following sensing modes by adjusting the duty cycle:

[0038] Olfactory-dominated mode: duty cycle less than or equal to 0.3;

[0039] Balanced fusion mode: duty cycle 0.4-0.6;

[0040] Visual dominance mode: duty cycle greater than or equal to 0.7.

[0041] This application also provides a method for fabricating a vision-olfaction fusion-based optoelectronic synaptic device, the method comprising:

[0042] A substrate is provided, and the surface of the substrate is cleaned and activated.

[0043] Two-dimensional material sheets were prepared to serve as the channel layer and the floating gate layer, respectively. The floating gate layer sheets were then oriented and stacked on the channel layer sheets using a transfer technique to form a van der Waals heterojunction. Annealing was then performed to optimize the interface bonding.

[0044] Photoresist is spin-coated onto the heterojunction, and electrode regions are defined by photolithography patterning. Subsequently, metal is deposited to form source and drain electrodes that are electrically connected to the channel layer.

[0045] Finally, demolding and annealing are performed to complete the fabrication of the photoelectric synapse device.

[0046] Compared with the prior art, the above-described solutions of this application have at least the following beneficial effects:

[0047] 1. This application discloses a gas recognition-smell fusion optoelectronic synaptic device and its fabrication method. It constructs a van der Waals heterojunction of channel layer-floating gate layer type with two-dimensional material and a Type-II band arrangement. The sensing of light is achieved through photogenerated carriers in the channel layer, and the sensing of gas is achieved through gas adsorption and change of charge state in the floating gate layer. It also has the functions of information memory and processing. The charge capture and slow release of the interface state simulates synaptic plasticity. It is integrated into a single micron-scale device, realizing the fusion of sensing and preliminary information processing at the physical hardware level, laying the foundation for miniaturization and low power consumption.

[0048] 2. This application discloses a gas recognition optoelectronic synaptic device with visual-olfactory fusion and its fabrication method. By applying parameters to the channel layer, particularly a programmable light pulse sequence with a duty cycle, rather than simply using it as a passive detection light source, this approach directly addresses the problems of traditional sensors, such as fixed response modes, susceptibility to interference, and lack of environmental adaptability. By adjusting the light duty cycle, the charge state of the floating gate layer can be actively and in real-time changed, thereby dynamically modulating the adsorption energy barrier and desorption rate of gas molecules on its surface. This allows the device to operate in different modes, such as olfactory-dominated, visual-dominated, or balanced fusion. Attached Figure Description

[0049] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings:Figure 1 Figures a and b in the figures are schematic diagrams of the structure and optical microscope of a gas recognition visual-olfactory fusion photoelectric synaptic device provided in the embodiments of this application, respectively.

[0050] Figure 2 This is a schematic diagram of the postsynaptic current (EPSC) response induced by light pulses of different wavelengths, provided in an embodiment of this application.

[0051] Figure 3 This is a schematic diagram illustrating the dependence of EPSC on light pulse intensity in an embodiment of this application.

[0052] Figure 4 This is a schematic diagram illustrating the dependence of EPSC on the duration of the optical pulse, as provided in an embodiment of this application.

[0053] Figure 5 This is a schematic diagram illustrating the effect of the optical pulse interval time (ΔT) on the paired pulse enhancement (PPF) index provided in an embodiment of this application.

[0054] Figure 6 This is a schematic diagram illustrating the adjustment of postsynaptic current state using light pulses with different duty cycles in a NO2 environment, as provided in an embodiment of this application.

[0055] Figure 7 This is a schematic diagram illustrating the effect of the gas pulse duration on postsynaptic current, as provided in an embodiment of this application.

[0056] Figure 8 This is a schematic diagram of the structure of a fully connected neural network (FCNN) provided in an embodiment of this application.

[0057] Figures a and b in Figure 9 are schematic diagrams of the confusion matrices of classification results under single-modal and dual-modal inputs provided in the embodiments of this application.

[0058] Figures a and b in Figure 10 are schematic diagrams of the LDA distribution of gas response characteristics under single-mode and dual-mode inputs provided in the embodiments of this application.

[0059] Figure 11 This is a schematic diagram illustrating the fabrication process of a gas recognition visual-olfactory fusion opto-synaptic device provided for some embodiments of this application. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0061] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a product or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a product or device. Without further limitation, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the product or device that includes that element.

[0062] The embodiments of this application are described in detail below with reference to the accompanying drawings. Example 1:

[0063] This application provides a visual-olfactory fusion opto-synaptic device for gas recognition. The opto-synaptic device mimics the efficient collaborative perception mechanism of the human visual and olfactory systems. The human brain integrates information from both vision and smell to achieve dynamically malleable cognitive decisions. This application employs a two-dimensional material floating-gate heterojunction as a biomimetic sensing unit. A programmable light pulse sequence dynamically modulates the adsorption dynamics of gas molecules on the floating-gate layer, resulting in a dynamic electrical response signal output by the device. It fuses visual information (light pulse timing) and olfactory information (gas type / concentration), achieving high-precision and robust gas recognition capabilities on a single micro-device through dual-modal information from both vision and smell.

[0064] In this application, nitrogen dioxide (NO2) and ammonia (NH3), two important environmental and safety monitoring targets, are selected as representative gases for illustration. NO2 is a typical oxidizing gas that tends to remove electrons from material surfaces. NH3 is a typical reducing gas that tends to donate electrons to material surfaces. These gases can respectively simulate the inhibitory (NO2) and excitatory (NH3) nerve signal transmission, comprehensively illustrating the photoelectric synaptic device of the technical solution in this application, and simulating the bidirectional regulatory capability of synaptic plasticity.

[0065] like Figure 1 As shown, a gas recognition-based optoelectronic synaptic device integrating vision and olfaction includes: a substrate, a two-dimensional material floating gate heterojunction, a source electrode, and a drain electrode. In this embodiment, the channel layer is described as MoS2 and the floating gate layer as SnS2; the same applies to other materials.

[0066] In this embodiment, a p-type heavily doped silicon wafer with a surface layer of SiO2 grown to a thickness of 300 nm is used, with the SiO2 layer serving as an insulating layer.

[0067] A two-dimensional material floating gate heterojunction disposed on a substrate, the two-dimensional material floating gate heterojunction includes at least: a channel layer and a floating gate layer stacked on the channel layer; in the embodiments of this application, the channel layer is MoS2 and the floating gate layer is SnS2 for illustration, and the same applies to other materials.

[0068] The source and drain electrodes, which are electrically connected to the channel layer, are fabricated in this embodiment using electron beam lithography and electron beam evaporation processes. The electrodes employ a titanium and gold composite metal layer to form good ohmic contacts.

[0069] The floating gate layer is used to respond to gas stimulation by changing the charge state through the adsorption or desorption of gas molecules; the channel layer is used to respond to light pulse stimulation by generating photogenerated carriers.

[0070] By applying programmable light pulse stimulation to the channel layer, the adsorption or desorption process of gas molecules on the floating grid layer and the charge state of the floating grid layer are dynamically modulated, and a dynamic electrical response signal that integrates visual and olfactory information is output to achieve gas recognition.

[0071] In the technical solution of this application embodiment, the channel layer is responsible for photoelectric conversion and signal transmission, and the floating gate layer is responsible for gas sensing, charge storage and fusion control.

[0072] This application provides a preferred technical solution in which a channel layer and a floating gate layer are bonded together by van der Waals forces to form a two-dimensional material floating gate heterojunction. The interface between the floating gate layer and the channel layer contains defect states that serve as centers for carrier capture and release.

[0073] The technical solution of this application adopts a floating-gate heterojunction as the core architecture, instead of a traditional field-effect transistor or a simple heterojunction. The floating-gate structure naturally possesses non-volatile charge storage capability, which can simulate the memory characteristics of neural synapses; while the heterojunction interface provides an efficient channel for the separation and transfer of photogenerated carriers. This solves the technical problems of weak gas adsorption capacity and low gas response sensitivity in gas recognition devices with planar heterojunctions, gate-controlled field-effect transistors, and multilayer stacked structures in the prior art.

[0074] This application provides a preferred technical solution where the floating gate layer meets the following conditions: a layered structure with a specific surface area of ​​not less than 500 m² / g; and surface active sites for gas molecule adsorption. More specifically, the conditions are: electrical performance: a moderate bandgap of 1.5–2.5 eV and a tunable work function of 1.6 eV–7.5 eV; optical performance: strong light absorption above 60%, long photogenerated carrier lifetime on the order of microseconds, and high photoinduced charge transfer efficiency above 90%; structural characteristics: layered structure, high specific surface area >500 m² / g, and controllable surface defects; interface characteristics: aligned with the channel layer to form a Type-II band, providing a charge trapping potential well; high surface chemical activity, facilitating gas adsorption / desorption.

[0075] The electrical properties of a floating gate layer are crucial; a bandgap that is too narrow results in high dark current and noise, while a bandgap that is too wide makes it difficult to be excited by visible light. A suitable bandgap allows the floating gate layer to effectively absorb ultraviolet-visible light while maintaining a low dark current, providing a clear signal baseline for optical modulation. Furthermore, a suitable bandgap significantly increases the energy barrier required for valence band electrons to randomly transition to the conduction band due to thermal excitation, ensuring that the stored charge is localized within the floating gate without external excitation, achieving long-term non-volatile storage.

[0076] The high dielectric constant of the floating gate layer enhances the charge storage capacity of the floating gate, i.e., the capacitance effect, while the tunable work function ensures an ideal band offset with the channel layer, thereby achieving efficient and persistent control of the channel conductance without continuous external bias, and realizing low-power synapses.

[0077] The strong light absorption, long carrier lifetime, and high charge transfer efficiency of the floating gate layer ensure that the energy of the incident light pulse can be effectively absorbed by the floating gate layer and converted into a storable charge, i.e., photogenerated carriers. This charge state can be maintained for a sufficiently long time, thereby achieving continuous and effective dynamic modulation of the gas adsorption process, rather than transient disturbance.

[0078] The layered structure of the floating gate layer and its high specific surface area (>500 m² / g) ensure low-damage van der Waals integration with the channel layer, while the extremely high specific surface area provides adsorption sites for gas molecules, which is the physical prerequisite for achieving high sensitivity and fast response.

[0079] The surface defects of the floating gate layer are controllable. Controllable defects, such as sulfur vacancies, serve as specific adsorption and reaction sites for gas molecules. They not only enhance adsorption but also provide intrinsic molecular selectivity through their differentiated interactions with different gas molecules, while ensuring the consistency of device performance.

[0080] The highly chemically active surface of the floating grating layer significantly reduces the gas adsorption energy barrier, accelerating the adsorption / desorption process. This allows the state of the photoelectric synaptic device to change rapidly and reversibly with light pulse modulation, generating rich dynamic response characteristics. This is the foundation for achieving high-precision dynamic recognition. In other words, the floating grating layer has a high specific surface area and abundant surface active sites, which is conducive to the efficient adsorption / desorption of gas molecules, enabling olfactory response.

[0081] This application provides a preferred technical solution where the channel layer meets the following conditions: a single-layer or multi-layer two-dimensional structure; a band gap of 1.1 eV-1.9 eV; high carrier mobility >50 cm² / Vs; low dark current <1 nA; and surface active sites for gas molecule adsorption. More specifically, the conditions are: electrical performance: high carrier mobility >50 cm² / V·s, low dark current <1 nA, and a tunable Fermi level; optical performance: a strong light absorption coefficient of 10. 5 cm-1 Large-scale, broad-spectral response of 405 nm-785 nm; rapid photogenerated carrier generation and recombination dynamics, generation time <1 ps, recombination time of several ps to tens of ns. Structural characteristics: single-layer or multi-layer two-dimensional structure, high carrier mobility >50 cm⁻¹. 2 / V·s, low defect density (excluding sulfur vacancies, 10¹²cm²) -2 Interface characteristics: It forms a Type-II band alignment with the floating gate layer, which is beneficial for charge separation and capture; its high surface activity is conducive to gas adsorption.

[0082] The high carrier mobility of the channel layer is >50 cm² / Vs. This high mobility ensures that photogenerated or modulated carriers can quickly pass through the channel, forming a clear electrical signal with high response speed and low operating voltage.

[0083] The low dark current of the channel layer (<1 nA) provides a stable and clean detection baseline for weak light and gas responses, enabling the resolution of minute current changes caused by trace amounts of gas or weak light, which is key to achieving high sensitivity.

[0084] The tunable Fermi level of the channel layer can be adjusted by doping, electric field and other means to optimize the alignment with the interface band of the floating gate layer, providing freedom for constructing high-performance, designable Type-II heterojunctions.

[0085] The strong light absorption coefficient and wide spectral response of the channel layer, with a direct band gap of approximately 1.8 eV, enable the channel layer to strongly absorb visible light. The wide spectral response allows the photoelectric synaptic device to be compatible with light sources of multiple wavelengths, ensuring that light pulse stimulation can be efficiently converted into electrical signals, thus providing a foundation for visual perception.

[0086] The rapid generation and recombination dynamics of the channel layer enable photoelectric synaptic devices to closely follow the temporal changes of light pulses, thereby accurately mapping the encoded information of the light pulses, such as the duty cycle, onto the current response. This is a prerequisite for generating high-precision dynamic characteristics and solves the problems of uncontrollable light response delay and relaxation processes.

[0087] The single-layer / multi-layer two-dimensional structure of the channel layer and its high crystallinity make it easy to form a low-defect, high-quality heterojunction interface with the floating gate layer through van der Waals forces. The high crystallinity ensures the stable performance of the material's intrinsic electrical / optical properties.

[0088] The low defect density of the channel layer (excluding sulfur vacancies) can maximize carrier mobility and reduce random fluctuations in the performance of photosynaptic devices, ensuring the reliability and repeatability of the sensing response.

[0089] The channel layer and the floating gate layer form a Type-II band alignment. The built-in electric field generated by the Type-II band alignment drives photogenerated electrons to be efficiently injected from the channel layer and stored in the floating gate layer, thereby directly modulating the gas response using photocharge.

[0090] This application provides a preferred technical solution, wherein the channel layer is made of any one of WSe2, WS2, black phosphorus, or MoS2; and the floating gate layer is made of any one of MoSe2, WS2, Ta2NiSe5, or MoS2. In this embodiment, the channel layer is described as MoS2 and the floating gate layer as SnS2; the same applies to other materials.

[0091] In this embodiment, MoS2 is selected as the channel layer, primarily serving as the photoelectric conversion and signal transduction unit, simulating the process by which the biological retina converts light signals into neural electrical signals. MoS2, as the channel layer, possesses a direct bandgap and high carrier mobility. When stimulated by light pulses in the wavelength range of 405-785 nm, MoS2 efficiently absorbs photons and generates a large number of photogenerated electron-hole pairs. These photogenerated carriers move rapidly under an applied bias voltage, forming a surge in source-drain current. After the light pulse ends, the current slowly decays due to some carriers being trapped by interface states and their own recombination dynamics. This dynamic process of rapid excitation and slow relaxation perfectly simulates the postsynaptic current (EPSC) generated by a biological synapse after receiving a neurotransmitter. Therefore, the carrier behavior in the MoS2 channel can be considered as a "neurotransmitter" transmitting visual information, and its current changes directly correspond to the excitatory activity of neurons.

[0092] SnS2 was chosen as the floating gate layer, primarily serving as a chemical sensing and information storage unit, simulating the specific response and short-term memory of olfactory neurons to gas molecules. It can also form a high-quality Type-II heterojunction with MoS2. Under the Type-II bandgap arrangement, photogenerated electrons in MoS2 spontaneously transfer to and are captured in SnS2, achieving efficient conversion of optical signals into charge memory. SnS2, as the floating gate layer, possesses a high specific surface area and abundant surface active sites, such as sulfur vacancies. When gas molecules, such as NO2 or NH3, are adsorbed, charge transfer occurs: NO2 captures electrons, and NH3 donates electrons, thereby altering the charge state of the SnS2 layer. This charge change continuously modulates the conductivity of the underlying MoS2 channel through electrostatic coupling, manifesting as a stable increase or decrease in source-drain current. Different gases induce opposite conductance changes, resulting in excitation or inhibition, simulating the differentiated neural responses triggered by olfactory stimuli. The SnS2 layer captures and stores the charge introduced by gas adsorption, achieving "non-volatile memory" of odor information, similar to the formation of short-term memory in the biological olfactory system.

[0093] This application provides a preferred technical solution whereby programmable light pulse stimulation is applied to the channel layer. The core of the fusion lies in using light pulses as a programmable tool to actively modulate the olfactory gas response process. Programmable synaptic weights: The duty cycle (D) of the light pulse becomes a key control parameter. It determines the total amount of photogenerated charge injected into SnS2, thereby dynamically balancing or suppressing the charge effect generated by gas adsorption. This is equivalent to dynamically programming the relative weights of biomimetic synapses for visual and olfactory inputs at the hardware level.

[0094] This application provides a preferred technical solution where the programmable light pulse stimulation light source meets the following conditions: wavelength range: 405–785 nm; power density range: 10–100 μW / mm²; pulse width: 0.1–1 s.

[0095] The wavelength range is set according to the absorption spectra of the trench layer material and the floating gate layer material. In the embodiments of this application, 405–785 nm covers the absorption spectra of MoS2 and SnS2.

[0096] The power density range is 10–100 μW / mm², with a lower limit ensuring sufficient light intensity to generate a adequate concentration of photogenerated carriers in the MoS₂ channel. These carriers can be effectively injected into the SnS₂ floating gate layer, thereby producing observable modulation of the gas adsorption barrier. Below this value, the modulation effect may be too weak to dominate the device response. The upper limit is primarily intended to avoid thermal effects.

[0097] Gas pulse width: adjustable from 1 to 10 s. It takes time for gas molecules to adsorb onto the SnS2 surface to reach a certain coverage or steady state. A pulse width of 1 s is the basic time to trigger a detectable gas response.

[0098] The continuously adjustable duty cycle range of 0.1-0.9 is crucial for programmability and fine-grained weight control. It allows for precise adjustment of the injected charge through minute duty cycle variations, enabling smooth and continuous modulation of the gas adsorption process. D=0.1 corresponds to the extreme olfactory-dominant mode, where light interference is extremely weak, maximizing the gas signal. D=0.9 approaches the extreme visual-dominant mode, where the light effect almost completely suppresses the gas effect.

[0099] like Figure 2As shown, the wavelength range of the light pulses is 405–785 nm, the pulse duration is 200 ms, and the intensity of the light pulses at all wavelengths is fixed at 20.24 μW / mm². The instantaneous EPSC response under light pulse triggering at different wavelengths (405 nm, 532 nm, 635 nm, and 785 nm) is demonstrated. The EPSC amplitude differs with different wavelengths, corresponding to the light absorption spectrum of the channel layer material MoS2, proving its broad visible-to-near-infrared response capability. Notably, after the 405 nm light pulse ends, the EPSC decays slowly over a timescale of up to 950 seconds, rather than immediately returning to zero. This indicates that the photoelectric synaptic device of this application simulates signal retention and short-term memory in biological visual perception, i.e., short-term synaptic plasticity.

[0100] like Figure 3 As shown, the EPSC varies under different intensities of a 405nm light pulse. When the light pulse intensity increases from 20.24 μW / mm² to 46.71 μW / mm², the maximum EPSC value of the photosynaptic device also increases from 85.51 nA to 138.01 nA. This figure demonstrates that, with a fixed light pulse wavelength and duration, the peak amplitude of the excitatory postsynaptic current (EPSC) increases significantly with increasing light power density, from a lower current at 20.24 μW / mm² to a higher current at 126.26 μW / mm². This directly reflects the light absorption and carrier generation capabilities of the MoS₂ channel layer. The stronger the light intensity, the more photogenerated electron-hole pairs are generated within the channel layer, resulting in a larger initial photocurrent.

[0101] like Figure 4 As shown, the EPSC changes under different intensities of a 405nm light pulse. With increasing pulse duration, EPSC exhibits a monotonically increasing trend; notably, even with a light stimulation duration of 500s, EPSC still fails to reach saturation. At a fixed wavelength and intensity, the EPSC amplitude monotonically increases with increasing pulse duration, from 0.1 s to 3.6 s. Longer illumination time means more photogenerated carriers are continuously generated. The trend of not saturating even after several seconds demonstrates that photogenerated carriers are continuously injected and captured and stored by the floating gate layer SnS2. This reflects the synergistic effect between the photogenerated carrier supply of the channel layer and the charge storage capacity of the floating gate layer.

[0102] like Figure 5As shown, two identical light pulses (wavelength 405 nm, intensity 20.24 μW / mm², pulse width 200 ms) were applied consecutively to a photoelectric synaptic device, with the time interval ΔT between the two pulses being systematically varied. It can be seen that when the second light pulse is applied shortly after the first pulse ends, for example, when ΔT = 800 ms, the resulting EPSC increase A2 is significantly greater than the increase A1 caused by the first pulse. This means that some of the charge carriers injected by the first light pulse are trapped in the deep or interface defect states of the floating gate layer SnS2 and have not been fully released on a timescale of hundreds of milliseconds to several seconds. When the second pulse arrives, these residual charges superimpose with the newly generated charges, jointly modulating the channel conductance, thus producing an enhanced response. The paired pulse facilitator (PPF) exponent of the photoelectric synaptic device decays with increasing pulse interval ΔT, and the change conforms to the double exponential decay function describing the short-term memory dynamics of biological synapses. The expression for the double exponential decay function is:

[0103] ,

[0104] in, This represents a double exponential decay function; C1 and C2 represent the initial promoting effect, respectively. , These represent the characteristic relaxation times of the fast decay term and the slow decay term, respectively.

[0105] By accurately fitting the PPF curve, two key time constants were obtained: the characteristic time of the rapid decay term. ≈3.72 seconds, characteristic time of the slowly decaying term ≈ 1175 seconds. This reflects the rapid release process of charge in shallow defect states at the floating gate interface, corresponding to rapid adaptation on the order of milliseconds to seconds; This reflects the extremely slow release process of charge in the deep trapped state, corresponding to a long-term memory effect lasting tens of minutes. The fitting results, with the solid line in the figure closely matching the experimental data points, conclusively demonstrate that the photoelectric synaptic device provided in this application can simulate the complex multi-timescale dynamics and memory characteristics of biological synaptic signal transmission with extremely high fidelity.

[0106] This application provides a preferred technical solution for switching the sensing mode of a photoelectric synapse device by adjusting the duty cycle, including:

[0107] Olfactory-dominated mode: duty cycle less than or equal to 0.3. In olfactory-dominated mode, the light-injected charge is low, resulting in a weak weakening effect on the adsorption energy barrier. The adsorption / desorption equilibrium of gas molecules tends towards strong adsorption. The conductivity of photoelectric synaptic devices is mainly suppressed by the gas adsorption effect, and the output current shows a decreasing trend, such as... Figure 6As shown, the three sub-images from left to right represent the olfactory-dominated mode, the balanced fusion mode, and the visual-dominated mode, respectively. The olfactory-dominated mode simulates the state of organisms relying primarily on their sense of smell in dim environments and is suitable for trace gas detection where extreme sensitivity is required.

[0108] Balanced fusion mode: duty cycle 0.4-0.6. In balanced fusion mode, the amount of light-injected charge is moderate, achieving a dynamic balance between the weakening effect on the adsorption energy barrier and the adsorption effect of gas molecules. The conductivity and output current of the photoelectric synapse device remain relatively stable within a certain range, such as... Figure 6 As shown, the balanced fusion mode simulates the brain's balanced integration of visual and olfactory information, and is suitable for obtaining robust response characteristics in complex, dynamic gas environments.

[0109] Visually Dominated Mode: Duty Cycle greater than or equal to 0.7. In this mode, strong and sustained light injection significantly reduces the adsorption energy barrier, promoting gas desorption. The enhancement effect of photogenerated carriers on channel conductivity is dominant, and the output current shows an upward trend, such as... Figure 6 As shown, the visual-dominated mode simulates the state where visual information dominates in a bright environment, even "masking" or "resetting" the olfactory background. It is suitable for optical feature extraction or device state reset under a known strong interfering gas background. In a fixed gas environment, light pulse sequences with different duty cycles will cause drastically different evolution trajectories of electrical conductance: at low duty cycles, electrical conductance continuously decays, and olfactory dominance prevails; at a balanced duty cycle, electrical conductance remains dynamically stable, achieving equilibrium; at high duty cycles, electrical conductance gradually increases, and visual dominance prevails. Example 2:

[0110] In this application, two important environmental and safety monitoring targets, NO2 (nitrogen dioxide) and NH3 (ammonia), are selected as representative gases for illustration. NO2 is a typical oxidizing gas that tends to remove electrons from material surfaces. NH3 is a typical reducing gas that tends to donate electrons to material surfaces. These gases can be used to simulate the inhibitory (NO2) and excitatory (NH3) neural signal transmission, respectively, to comprehensively illustrate the gas recognition method based on a visual-olfactory fusion photoelectric synaptic device in this application, and to demonstrate the accuracy of gas recognition.

[0111] This application also provides a gas recognition method based on a visual-olfactory fusion optoelectronic synaptic device, including:

[0112] S1. Expose the photosynaptic device to an environment containing a target gas. Place the photosynaptic device in a controlled environment containing a target gas, such as NO2, NH3, or a mixture thereof. Subsequently, apply programmable light pulse stimulation to the channel layer of the photosynaptic device.

[0113] In this embodiment, 20 ppm of NH3 (reducing gas) and 20 ppm of NO2 (oxidizing gas) are introduced respectively.

[0114] S2. A programmable light pulse is applied to the channel layer of the photosynaptic device to dynamically modulate the adsorption or desorption process of target gas molecules on the floating gate layer. A fixed pulse period is set, and the duty cycle is adjusted by changing the pulse duration. The wavelength is 405 nm, the pulse width is 200 ms, and the period is 4 s. Different duty cycles are achieved by changing the ratio of pulse width to period. For example, a sequence of light pulses with D=0.25 (olfactory dominant), D=0.4-0.6 (balanced fusion), and D=0.75 (visual dominant) are applied sequentially. Photogenerated carriers actively and dynamically modulate the adsorption state and charge state of gas molecules on the SnS2 floating gate layer, causing the photosynaptic device to output different response trajectories.

[0115] S3. Acquire the dynamic electrical response signal output by the photoelectric synapse device. While applying the aforementioned programmable optical pulse sequence, use a semiconductor parameter analyzer to acquire and record the curve of the device's source-drain current changing over time at a high sampling rate. This curve represents the dynamic electrical response signal that integrates specific optical programming information and specific gas information. For example... Figure 7 As shown, it can be observed that the amplitude of EPSC (for NH3) or IPSC (for NO2) changes accordingly with the duration of the gas pulse from 5 to 120 s, reflecting the cumulative effect of gas adsorption on the charge state of the device.

[0116] S4. Identify the type or concentration of the target gas based on the dynamic electrical response signal.

[0117] The method in this application embodiment can directly identify the type or concentration of the target gas, and can also be combined with neural network models in the prior art.

[0118] In the embodiments of this application, the following is adopted: Figure 8 The fully connected neural network shown is used as the recognition model. The dynamic current response curve collected in step S3 is used as the input feature and input into a pre-trained fully connected neural network (FCNN) model. This model has been trained using massive amounts of dynamic response curve data collected under different concentrations and light duty cycles for various gases (such as NO2, NH3, H2, etc.). The neural network extracts and matches the features of the input curve and outputs the determination result of the target gas type and the estimated value of its concentration.

[0119] The technical solution of this application is to perform gas recognition through dual-modal fusion of vision and olfaction. The effectiveness of this technical solution is verified through comparative experiments with traditional single-modal olfaction methods. The responses of the same gas samples were collected under both light-free modulation (single-modal olfaction) and light pulse modulation (dual-modal fusion), and features were extracted and used to train the same FCNN model for recognition. As shown in Figure 9, the confusion matrix shows that the overall recognition accuracy under single-modal features is only 52.24%, while the recognition accuracy jumps to 98.27% based on the dual-modal fusion features obtained by the method of this application, proving that the dynamic signal extracted through light pulse modulation contains stronger discriminative information.

[0120] To illustrate the concept clearly, linear discriminant analysis (LDA) is used to reduce the dimensionality of the single-modal and dual-modal fused features in the confusion matrix for visualization. As shown in Figure 10, points of the same color represent samples belonging to the same category, and the number of samples of each color reflects the sample size of that category. In the single-modal feature space, as shown in Figure a, the distribution of single-modal features in two-dimensional space after dimensionality reduction shows severe overlap of sample points from different gas categories. However, in the dual-modal dynamic feature space obtained in this application, as shown in Figure b, samples of the same category are highly clustered, while samples of different categories are clearly separated, significantly enhancing feature separability. This geometrically and intuitively demonstrates the high cohesion and strong discriminative power of the features generated by the method in this application. Example 3:

[0121] This application provides a fabrication process for a gas recognition-based visual-olfactory fusion optoelectronic synaptic device, such as... Figure 11 As shown, it includes the following steps:

[0122] Step a: Prepare SiO2 / Si substrate

[0123] Prepare a P-type highly doped silicon wafer with 300nm thick SiO2. Clean it with acetone, anhydrous ethanol, and deionized water, respectively, and then dry the wafer with a nitrogen gun. Perform a 1-minute oxygen plasma treatment to deeply clean the SiO2.

[0124] Step b: Preparation of MoS2 / SnS2 heterojunction

[0125] MoS2 and SnS2 sheets were prepared by mechanical exfoliation, and then stacked to form a heterojunction by directional transfer technology. The MoS2 / SnS2 heterojunction was then heated at 80 degrees Celsius for 20 minutes to allow the two materials to adhere tightly and form a high-quality van der Waals interface.

[0126] Step c: Prepare metal electrode

[0127] Photoresist was spin-coated onto the MoS2 / SnS2 heterojunction, and patterns were formed at both ends of the MoS2 and SnS2 layers using ultraviolet lithography. Then, 20 nm and 80 nm Ti and Au were deposited using thermal evaporation to form the drain and source electrodes. A demolding process was then used to remove the photoresist and excess metal. Subsequently, the device was annealed at 200°C for 20 minutes in a nitrogen atmosphere to repair lattice defects between the materials and electrodes, reduce contact resistance, and improve device performance.

[0128] The technical solution of this application provides a visual-olfactory fusion opto-synaptic device for gas recognition and its fabrication method. Based on a two-dimensional material floating gate heterojunction and a Type-II band structure, it realizes a biologically inspired visual and olfactory fusion perception function. This is achieved by applying programmable light pulse stimulation to the channel layer, actively and dynamically modulating the adsorption or desorption process of gas molecules on the floating gate layer. This allows the device to generate cooperative or competitive dynamic electrical response signals when faced with light and gas stimuli, thereby extracting gas response features with significant discriminative power. The opto-synaptic device operates in a micrometer-scale size of approximately 20 μm × 32 μm, with power consumption at the microwatt level and a gas detection limit of 74 ppb. Through the above-mentioned visual-olfactory dual-mode fusion perception mechanism, the gas recognition accuracy of the device is improved from 52.24% in the single olfactory mode to 98.27%, successfully achieving synergistic optimization of device miniaturization, low power consumption, and high detection performance.

[0129] Finally, it should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

[0130] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A gas recognition-olfactory fusion opto-synaptic device, characterized in that, include: Substrate; A two-dimensional material floating gate heterojunction is disposed on the substrate; The two-dimensional material floating gate heterostructure includes at least: a channel layer and a floating gate layer stacked on the channel layer; The source electrode and drain electrode are electrically connected to the channel layer; The floating gate layer is used to respond to gas stimulation by changing the charge state through the adsorption or desorption of gas molecules; the channel layer is used to respond to light pulse stimulation to generate photogenerated carriers. The channel layer and the floating grid layer constitute In a bandgap heterojunction, the interface between the floating gate layer and the channel layer contains defect states for carrier capture and release. By applying programmable light pulse stimulation to the channel layer, the adsorption or desorption process of gas molecules on the floating grid layer and the charge state of the floating grid layer are dynamically modulated, and a dynamic electrical response signal that integrates visual and olfactory information is output to achieve gas recognition. The programmable light pulse-stimulated light source satisfies the following conditions: The wavelength range is 405–785 nm; The power density range is 10–100 μW / mm². The optical pulse width is adjustable within 0.1–1 s; Gas pulse width: adjustable from 1 to 10 seconds; Optical duty cycle range: continuously adjustable from 0.1 to 0.9; The photosynaptic device can be switched to operate in one of the following sensing modes by adjusting the duty cycle: Olfactory-dominated mode: duty cycle ≤ 0.3; Balanced fusion mode: duty cycle 0.4-0.6; Visual dominance mode: duty cycle ≥ 0.

7.

2. The photoelectric synapse device according to claim 1, characterized in that, The channel layer and the floating gate layer are bonded together by van der Waals forces to form the two-dimensional material floating gate heterojunction.

3. The photoelectric synapse device according to claim 1, characterized in that, The floating gate layer satisfies the following conditions: The work function is adjustable; The bandgap width is 1.5–2.5 eV; It has a layered structure and a specific surface area of ​​not less than 500 m² / g; Furthermore, the surface of the floating gate layer has surface active sites for the adsorption of gas molecules.

4. The photoelectric synapse device according to claim 1, characterized in that, The channel layer is a two-dimensional semiconductor material layer with a direct bandgap, and satisfies the following conditions: Single-layer or multi-layer two-dimensional structure; Bandgap of 1.1-1.9 eV; high carrier mobility >50 cm² / Vs; low dark current <1 nA; Fermi level is adjustable; Furthermore, the surface of the channel layer has surface active sites for the adsorption of gas molecules.

5. The photoelectric synapse device according to claim 4, characterized in that, The conditions that the channel layer and the floating gate layer satisfy are: The difference in work function between the interface between the channel layer and the floating gate layer ranges from 0.3 to 0.8 eV; The mismatch rate between the lattice constants of the channel layer and the floating gate layer is <5%.

6. The photoelectric synapse device according to claim 1, characterized in that, The material of the channel layer is , Black phosphorus or Any one of the following; the material of the floating gate layer is , , or Any one of them.

7. The photoelectric synapse device according to claim 6, characterized in that, The material of the channel layer is The material of the floating grid layer is .

8. A method for fabricating a visual-olfactory fusion-based opto-synaptic device, comprising fabricating a visual-olfactory fusion-based opto-synaptic device for gas recognition as described in any one of 1 to 7, characterized in that, The preparation method includes: A substrate is provided, and the surface of the substrate is cleaned and activated. Two-dimensional material sheets were prepared to serve as the channel layer and the floating gate layer, respectively. The floating gate layer sheets were then oriented and stacked on the channel layer sheets using a transfer technique to form a van der Waals heterojunction. Annealing was then performed to optimize the interface bonding. Photoresist is spin-coated onto the heterojunction, and electrode regions are defined by photolithography patterning. Subsequently, metal is deposited to form source and drain electrodes that are electrically connected to the channel layer. Finally, demolding and annealing are performed to complete the fabrication of the photoelectric synapse device.

Images