A multi-modal sensory artificial synapse device based on a frame material and a preparation method and application thereof

Abstract

The application discloses a kind of multi-modal feeling artificial synapse devices based on framework material and preparation method, application thereof.Artificial synapse device includes: substrate, channel, the channel is arranged on the surface of the substrate;The channel is amorphous form of organic metal framework film responding to light and gas stimulation;Metal electrode, the metal electrode is arranged on the surface of the channel;The metal electrode includes source electrode and drain electrode.Compared with prior art, the framework material with excellent photoelectric response performance is used in the application, by making full use of the regular pore structure, large specific surface area and good light response characteristics of MOF material, a new artificial synapse device with multi-modal sensing capability is constructed.Two electrodes of the device are connected into electrical test equipment, under the stimulation of light pulse signal, the conductance of amorphous MOF film can be changed, and finally the simulation based on light response synaptic plasticity is realized;The device is placed in different gas environment, the capture and release of target gas molecules by the porous characteristics of amorphous MOF material itself can also change the conductance of MOF film, so as to realize the simulation based on gas response synaptic plasticity.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a multimodal sensory artificial synapse device based on a frame material, its fabrication method, and its application. Background Technology

[0002] With the rapid development of artificial intelligence technology, intelligent sensing systems face an urgent need for integrated sensing and computing. The traditional architecture of separating sensing and computing has many drawbacks: First, the physical separation of sensing and computing units leads to the frequent transmission of large amounts of data between the two, resulting in huge energy consumption; second, traditional systems mostly adopt single-modal sensing methods, making it difficult to achieve the fusion processing of multi-sensory information; in addition, the functional separation of sensing and computing also reduces the overall efficiency and response speed of the system.

[0003] In contrast, in-sensor computing based on artificial synapses has significant advantages: by directly integrating computing functions into the sensing unit, the data transmission bottleneck between sensing and computing is eliminated, and the system power consumption is greatly reduced; at the same time, the integrated sensing and computing architecture improves information processing efficiency and enhances the real-time response capability of the system.

[0004] However, existing sensing devices and systems still face numerous technical challenges. In olfactory perception, the complexity of gas molecule-material interactions makes material selection difficult for existing devices, particularly due to insufficient stability in response to gas excitation, which limits recognition accuracy and reliability. Regarding device structure, while vertically structured sensing devices are advantageous for high-density integration, their relatively small sensing area and weak reconfigurability limit their application flexibility. Furthermore, how to realize multimodal sensing (such as simultaneously responding to light and gas signals) artificial synaptic devices to better simulate visual and olfactory behaviors in biological neural networks is also a critical technical problem that urgently needs to be solved.

[0005] Therefore, the existing technology still needs further research and improvement. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a multimodal sensory artificial synapse device based on a framework material, its fabrication method, and its applications, aiming to solve the problem of low multimodal sensory integration in existing neuromorphic devices. Specifically:

[0007] In a first aspect, embodiments of the present invention provide a multimodal sensory artificial synapse device based on a frame material, comprising:

[0008] Substrate,

[0009] a channel, arranged on the surface of the substrate; the channel is an amorphous organometallic framework thin film responsive to light and gas stimuli;

[0010] metal electrodes, arranged on the surface of the channel; the metal electrodes include a source electrode and a drain electrode.

[0011] The following are preferred technical solutions of the present invention, but are not intended to limit the technical solutions provided by the present invention. Through the following preferred technical solutions, the objectives and beneficial effects of the present invention can be better achieved and realized.

[0012] As a preferred technical solution, in the framework material-based multimodal sensory artificial synaptic device, the channel length is 40-60 μm.

[0013] As a preferred technical solution, in the framework material-based multimodal sensory artificial synaptic device, the organometallic framework thin film is a Cu-HHTP thin film, a Zn-HHTP thin film or a Ni-HHTP thin film.

[0014] In a second aspect, the present invention provides a method for preparing the framework material-based multimodal sensory artificial synaptic device according to the first aspect, comprising:

[0015] providing a substrate;

[0016] preparing an amorphous organometallic framework thin film responsive to light and gas stimuli on the surface of the substrate;

[0017] preparing metal electrodes on the surface of the organometallic framework thin film.

[0018] As a preferred technical solution, in the preparation method, the step of preparing the amorphous organometallic framework thin film responsive to light and gas stimuli on the surface of the substrate comprises:

[0019] immersing the substrate in an ethanol solution of HHTP for 10-60 seconds, taking out the substrate and cleaning it with ethanol;

[0020] immersing the ethanol-cleaned substrate in an aqueous solution containing copper ions for 10-60 seconds, taking out the substrate and rinsing it with water, which is recorded as one cycle;

[0021] repeating the above cycle n times, where 10<n<50, and drying the substrate loaded with the deposited layer in an inert atmosphere to obtain the multimodal sensory artificial synaptic device.

[0022] As a preferred technical solution, in the preparation method, a silicon dioxide layer is arranged on the surface of the substrate, and the thickness of the silicon dioxide layer is 80-120 nm.

[0023] As a preferred technical solution, the preparation method, wherein preparing a metal electrode on the surface of the metal-organic framework thin film includes: attaching a mask to the surface of the metal-organic framework thin film, and depositing gold on the surface of the metal-organic framework thin film using vapor deposition, electron beam evaporation, magnetron sputtering or mechanical transfer techniques to obtain a metal electrode.

[0024] As a preferred technical solution, in the preparation method, the number of the photomasks is an array of 5×6 to 10×10.

[0025] As a preferred technical solution, the preparation method includes providing a substrate, comprising: the substrate being a single-sided polished P-type doped silicon wafer; cleaning and drying the substrate; and modifying the surface of the substrate using UVO technology after drying.

[0026] Thirdly, the application of the multimodal sensory artificial synapse device based on the framework material described in the first aspect as a neuromorphic device.

[0027] Beneficial Effects: Compared with existing technologies, this invention utilizes a framework material with excellent photoelectric response performance. By fully leveraging the regular porous structure, large specific surface area, and good photoresponse characteristics of MOF materials, a novel artificial synaptic device with multimodal sensing capabilities is constructed. Connecting the two electrodes of the device to an electrical testing device, the conductivity of the amorphous MOF film can be altered under the stimulation of a light pulse signal, ultimately achieving simulation based on photoresponsive synaptic plasticity. Placing the device in different gas environments, the trapping and release of target gas molecules by the porous properties of the amorphous MOF material itself can also alter the conductivity of the MOF film, thereby achieving simulation based on gas-responsive synaptic plasticity. This invention fills the gap in the application of MOF materials in multimodal recognition artificial synaptic devices. Furthermore, by utilizing the customizability of the MOF material's structure, it is possible to simulate visual and olfactory perception in biological neural networks, as well as their interaction, providing a framework for the fabrication of novel integrated sensing-memory-computing chips based on multimodal sensing. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A schematic diagram of a multimodal sensory artificial synapse device based on a frame material, provided as an example of the present invention.

[0030] Figure 2 A schematic diagram of the process for preparing an amorphous organometallic framework thin film responsive to light and gas stimuli, as provided in this invention example.

[0031] Figure 3 SEM cross-sectional view of a multimodal sensory artificial synapse device based on a frame material provided as an example of the present invention.

[0032] Figure 4 TEM lattice diffraction pattern of a multimodal sensory artificial synapse device based on a framework material, provided as an example of the present invention.

[0033] Figure 5 The postsynaptic current of the multimodal sensory artificial synapse device based on frame material provided in this invention is shown under stimulation by light pulses of different wavelengths.

[0034] Figure 6 A schematic diagram of static preprocessing calculation for a multimodal sensory artificial synapse device based on frame materials, provided as an example of the present invention.

[0035] Figure 7 A schematic diagram of dynamic preprocessing calculation for a multimodal sensory artificial synapse device based on a frame material, provided as an example of the present invention.

[0036] Figure 8 The multimodal sensory artificial synapse device based on frame material provided in this invention exhibits an STM to LTM transition as the number of pulses increases.

[0037] Figure 9 The PPF characteristics of a multimodal sensory artificial synapse device based on a frame material provided in this invention example.

[0038] Figure 10 The multimodal sensory artificial synapse device based on a frame material provided in this invention provides long-term retention of light pulse stimulation.

[0039] Figure 11 The current of the multimodal sensory artificial synapse device based on frame material provided in this invention under different nitrogen dioxide concentrations.

[0040] Figure 12 The light pulse response of the multimodal sensory artificial synapse device based on a frame material provided in this invention under different sulfur dioxide concentrations. Detailed Implementation

[0041] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] The inventors discovered that existing neuromorphic devices face many problems. First, multimodal sensing is difficult to achieve, especially olfactory sensing performance. While traditional neuromorphic device materials (such as oxides and perovskites) perform well in photoelectric signal response, they lack sensitivity to gas molecules, exhibit poor response stability, and are prone to signal drift in complex gas environments, making it difficult to achieve high-precision olfactory simulation. Existing devices have limitations in material selection and device structure design, resulting in low integration of multimodal sensing and an inability to effectively simulate the collaborative working mechanisms of multiple senses such as vision and olfaction in biological nervous systems. Second, the separation of sensing and computation functions limits system performance. Traditional sensing-computation separation architectures require frequent data transmission between sensing and computation units, consuming significant energy and requiring complex AD conversion processes, resulting in slow processing speeds and high system latency. Simultaneously, security risks during data transmission reduce the overall system security. This architecture struggles to meet the comprehensive requirements of modern intelligent sensing systems for high efficiency, low power consumption, and high security.

[0043] Addressing the technical challenges faced by existing neuromorphic devices, the inventors of this invention employ framework materials with excellent photoelectric response properties, particularly MOF materials. By fully utilizing the regular pore structure, large specific surface area, and good photoresponse characteristics of MOF materials, a novel artificial synaptic device with multimodal sensing capabilities is constructed. Combined with... Figure 1As shown, the artificial synaptic device includes a single-sided polished silicon dioxide substrate, source / drain electrodes, and a MOF thin film on the substrate. The MOF thin film serves as the channel material between the two electrodes. Connecting the two electrodes to an electrical testing device, the conductivity of the MOF thin film can be altered under the stimulation of a light pulse signal, ultimately achieving a simulation based on photoresponsive synaptic plasticity. Placing the device in different gas environments, utilizing the porous nature of the MOF material to trap and release target gas molecules, can also change the conductivity of the MOF thin film, thereby achieving a simulation based on gas-responsive synaptic plasticity. Simultaneously, by using either the light signal or the gas environment signal as a variable and fixing the other stimulus signal, the influence of the former on the latter's stimulus response can be explored, thus simulating bimodal recognition in biological neural networks. This invention fills the gap in the application of MOF materials in multimodal recognition artificial synaptic devices. Furthermore, by utilizing the customizable structure of MOF materials, it is possible to simulate visual perception and olfactory perception in biological neural networks, as well as their interaction, providing a framework for the fabrication of novel integrated sensing-memory-computing chips based on multimodal perception. In addition, MOF materials themselves have low cost, and the thin film and device preparation methods involved in this invention are simple in process, have low equipment requirements, and are inexpensive, and have broad application prospects in the fields of information, communication, and smart electronics.

[0044] In this invention, a multimodal sensory artificial synapse device based on a framework material uses a Cu-HHTP thin film as the channel between two electrodes, and gold is selected as the electrode material. The device is fabricated by measuring the current value between the channels using a constant reading voltage, and its synaptic plasticity simulation performance is studied. A biological synapse can be divided into three parts: the presynaptic membrane, the synaptic cleft, and the postsynaptic membrane. During operation, the presynaptic membrane releases neurotransmitters that travel through the synaptic cleft to the postsynaptic membrane, thereby causing excitatory (inhibitory) postsynaptic currents (EPSC / IPSC). Similar to biological synapses, the stimulating signal to the artificial synapse device can be considered as the neurotransmitter released from the presynaptic membrane. The interaction between the stimulating signal and the device's channel material can be considered as the process of neurotransmitters interacting with the postsynaptic membrane through the synaptic cleft. Changes in the channel current can be considered as postsynaptic currents, including both excitatory and inhibitory types. For example, when the device's channel is stimulated by a light pulse, the photocurrent generated in its channel can be considered as an excitatory postsynaptic current (EPSC). By exploring the relationship between its EPSC and light pulse characteristics and their changing trends, this artificial synaptic device can be applied to simulate visual perception in biological neural networks. Furthermore, by placing it in different gas environments, it can also be applied to simulate olfactory perception, thereby exploring its application in multimodal recognition. The Cu-HHTP thin film was prepared by direct in-situ layer-by-layer growth on the substrate, and the electrodes were prepared by vapor deposition.

[0045] In this embodiment of the invention, the fabrication method of the above-mentioned multimodal sensory artificial synapse device is as follows:

[0046] 1. Substrate selection and pretreatment

[0047] The substrate for the device is a single-sided polished P-type doped silicon wafer (with a 100nm thick silicon dioxide layer on the polished surface). Before thin film preparation, the silicon wafer surface was ultrasonically washed for 15 minutes each with isopropanol, ethanol, and ultrapure water, and then dried. After drying, it was treated with UVO for at least 15 minutes to hydrophilize the substrate surface.

[0048] 2. Preparation of Cu-HHTP thin films

[0049] The MOF film was prepared using a layer-by-layer in-situ growth method. Figure 2 Copper acetate monohydrate and HHTP were dissolved in ultrapure water and ethanol, respectively, and dissolved by ultrasonication to form solutions with concentrations adjusted to 5 mmol / L and 0.5 mmol / L, respectively. At room temperature, the pretreated silicon wafer was first immersed in the HHTP ethanol solution for 40 seconds, then rinsed with ethanol to remove any residual solution from the silicon wafer surface. After rinsing, the silicon wafer was immersed in the copper acetate solution for 30 seconds, and then rinsed with ultrapure water to remove any remaining solution from the surface. This process constituted one cycle and could be repeated to grow a multilayer MOF film on the silicon dioxide surface. When the required number of cycles was reached, the silicon wafer was placed in a nitrogen glove box and heated at 120°C for 1 hour to remove moisture from the film. This cycle was repeated 35 times.

[0050] 3. Electrode preparation

[0051] After attaching a photomask to the MOF thin film, device electrodes were fabricated using a resistive thermal evaporation deposition apparatus. The electrode material was gold, the deposition rate was 0.1 A / s, and the channel length of the device was 50 μm. The photomask was an array of 10 × 10 devices.

[0052] In this embodiment, compared to common single-crystal or polycrystalline active thin films (active thin films refer to thin films that are sensitive to gas and light stimulation, and can be semiconductors, conductors, or insulators), the MOF thin film prepared in this embodiment is amorphous, and its morphology and structure can be obtained from SEM cross-sectional images (…). Figure 3 ) and TEM lattice diffraction ( Figure 4This can be used to prove the advantages of amorphous active thin films: 1. They provide more defect sites, which is beneficial for carrier trapping. For example, under light pulse excitation, these defect vacancies can trap more photogenerated carriers, giving the thin film material more photoconductive states and a longer relaxation time, which is beneficial for the preprocessing of light (visual) sensing signals (such as contrast enhancement, noise reduction, etc.) and reduces the amount of data to be processed; 2. Compared with single crystals and polycrystalline materials, amorphous structures are relatively loose, which is conducive to the penetration and adsorption of gas molecules, thereby enhancing the gas sensing ability; 3. Combined with the inorganic-organic properties of MOFs themselves (i.e., metal nodes-organic ligands), porous MOF thin films with amorphous structures have more diverse charge binding sites and gas molecule adsorption sites than other materials, such as pure inorganic or organic materials, which is beneficial for the realization of high-performance multimodal sensing computing.

[0053] The performance of the multimodal sensory artificial synapse device based on the frame material prepared above was tested. The test process and results are as follows:

[0054] The manifestation of information computing capability (mainly low-level computing capability, i.e., preprocessing function) in inductive computing: During the testing process, the synaptic simulation function of the device is first tested (in this invention, it is mainly reflected through EPSC, i.e., the device can simulate the excitatory postsynaptic current function of biological synapses under different excitation environments); here, based on the EPSC function, taking the processing of visual (light) signals as an example, the preprocessing function of the device is described: 1. Preprocessing of static image signals (such as... Figure 6 As shown), the image to be processed is first encoded and input into the synaptic device array via light irradiation. Before the light irradiation, the device current is I0; after irradiation, the device currents at 0s, 20s, 40s, and 60s are I1, I2, I3, and I4, respectively. The currents corresponding to the image are normalized to obtain the output image. The output images at different times are the preprocessed images of the device array. This processing can achieve noise reduction and contrast enhancement of the original image. 2. Preprocessing of visual dynamic signals (such as...) Figure 7As shown in the diagram, according to EPSC, the conductivity of a device changes over time after being stimulated by light. Therefore, when different devices on the device array are stimulated by dynamic light at different time points, the conductivity of these stimulated devices at the same time point will be different. Thus, by comparing the conductivity values ​​of different devices at the same time point, the order in which the devices were stimulated by dynamic light can be determined, and the motion trajectory (i.e., position and time information) of the light source (i.e., the visual target) can be obtained. Therefore, this device can realize dynamic calculations such as trajectory tracking and path recognition under different gas atmospheres. Besides visual recognition examples, since the device also responds to different gas stimuli, it can determine the type and concentration of a gas based on the response to different gases, and determine the location of a gas source based on the gas concentration gradient. The existence of the device array also facilitates the correction of different devices' gas responses under the same gas atmosphere, and the construction of conductivity fingerprints under different gas atmospheres, thereby enhancing recognition accuracy.

[0055] Furthermore, the above examples only discussed the computational function under a single excitation condition (light or gas). It is foreseeable that the resulting device can achieve inductive computation under multiple excitation conditions. Therefore, under visual and olfactory multimodal excitation, the device can comprehensively process multimodal information, which is beneficial to the realization of complex functions. For example, combined with a robot dog entity, it is expected to realize obstacle avoidance of static obstacles and hunting of dynamic prey based on odor source tracing.

[0056] By measuring the EPSC of a device using a semiconductor analyzer and observing its changes with different light pulses and gas environments, the synaptic plasticity of the device can be investigated. Under different light pulse signals, the device exhibits different EPSC values. Figure 5 Based on the different photoconductivities generated by the devices for different optical signals, and by setting a threshold and normalizing the corresponding current at the device output, when applied to optical signals such as images, the device array can achieve static low-level calculations such as noise reduction and recognition. Figure 6 With reduced read voltage and pulse width, the power consumption of a single device can be as low as 3.8 pJ. At a wavelength of 445 nm (240 mW / cm²), this is achievable. 2 Pulse tests were conducted under conditions of 1s. The results showed that as the number and frequency of pulses increased, both the channel conductivity and recovery time of the device increased. This indicates that the device can be used as a light-controlled artificial synapse to simulate synaptic plasticity, such as the transition from STM to LTM. Figure 8 The learning and forgetting curve; as the interval between two pulses changes, the ratio of its EPSC peak value also decreases, and its behavior is similar to the PPF behavior of biological synapses. Figure 9Similar to [other characteristics]. Under prolonged illumination, the device's EPSC increases significantly. After the illumination is removed, the EPSC does not immediately recover but remains at a level higher than the original value for a long time, indicating that the device has a long-term retention capability for the results of optical pulse signals. Figure 10 It exhibits good non-volatility and can be used for storing optical information. The device displays different conductivities under different gaseous environments. Figure 11 This demonstrates that the device can be used as an artificial synapse to simulate olfactory synaptic plasticity. Furthermore, under different gaseous environments, when the same light pulse signal is applied, the device exhibits different conductivities and pulse characteristics. Figure 12 This demonstrates the device's ability to simulate the dual-modal visual and olfactory recognition characteristics of a biological neural network. Furthermore, by comparing the conductance values ​​of different devices at different times after stimulation by various light pulses, the order in which the devices were dynamically stimulated can be determined, allowing for the deduction of object position and time information. Therefore, this device can perform dynamic calculations such as trajectory tracking and path recognition under different gas atmospheres. Additionally, measuring the EPSC of different devices on the substrate under light pulse signal stimulation revealed a high yield for the device array and good pulse uniformity among the devices. This indicates that the artificial synaptic device based on Cu-HHTP MOF thin film material has good stability and can be scalably integrated. It exhibits excellent light response, gas response, and cooperative response, and shows promising application prospects in neural network computing, machine learning, and intelligent speech / image recognition.

[0057] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

[0058] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multimodal sensory artificial synapse device based on a frame material, characterized in that, comprising: a substrate, a channel arranged on the surface of the substrate; the channel is an amorphous organometallic framework thin film responsive to light and gas stimuli; metal electrodes arranged on the surface of the channel; the metal electrodes include a source electrode and a drain electrode; The organometallic framework thin film is a Cu-HHTP thin film.

2. The multimodal sensory artificial synapse device based on frame material according to claim 1, characterized in that, The channel length is 40-60 μm.

3. A method for fabricating a multimodal sensory artificial synapse device based on a frame material as described in claim 1, characterized in that, comprising: providing a substrate; preparing an amorphous organometallic framework thin film responsive to light and gas stimuli on the surface of the substrate; preparing metal electrodes on the surface of the organometallic framework thin film.

4. The preparation method according to claim 3, characterized in that, The step of preparing the amorphous organometallic framework thin film responsive to light and gas stimuli on the surface of the substrate comprises: immersing the substrate in an ethanol solution of HHTP for 10-60 s, taking it out and cleaning it with ethanol; immersing the ethanol-cleaned substrate in an aqueous solution containing copper ions for 10-60 s, taking it out and rinsing it with water, which is recorded as one cycle; repeating the above cycle n times where 10<n<50, and subjecting the substrate loaded with the deposited layer to drying treatment in an inert atmosphere to obtain the multimodal sensory artificial synaptic device.

5. The preparation method according to claim 3, characterized in that, A silicon dioxide layer is arranged on the surface of the substrate, and the thickness of the silicon dioxide layer is 80-120 nm.

6. The preparation method according to claim 3, characterized in that, The step of preparing the metal electrodes on the surface of the organometallic framework thin film comprises: attaching a mask plate to the surface of the organometallic framework thin film, and depositing gold on the surface of the organometallic framework thin film by evaporation, electron beam evaporation, magnetron sputtering or mechanical transfer technology to obtain the metal electrodes.

7. The preparation method according to claim 6, characterized in that, The number of the mask plates is an array of 5×6 to 10×10.

8. The preparation method according to claim 3, characterized in that, The step of providing a substrate comprises: the substrate is a single-side polished P-type doped silicon wafer; cleaning and drying the substrate, and modifying the surface of the substrate using UVO technology after drying.

9. An application of the framework material-based multimodal sensory artificial synaptic device according to claim 1 as a neuromorphic device.

Images