CN110443345B

CN110443345B - Method for regulating and controlling electric pulse distribution behavior of nano molecular neural network

Info

Publication number: CN110443345B
Application number: CN201910699882.8A
Authority: CN
Inventors: 周文利; 吴硕; 陈昌盛; 朱宇; 程润虹; 王耘波; 高俊雄
Original assignee: Huazhong University of Science and Technology
Current assignee: Huazhong University of Science and Technology
Priority date: 2019-07-31
Filing date: 2019-07-31
Publication date: 2021-10-08
Anticipated expiration: 2039-07-31
Also published as: CN110443345A

Abstract

The invention belongs to the technical field of artificial intelligence, and relates to a method for regulating and controlling the electric pulse distribution behavior of a polyoxometallate-semiconductor low-dimensional nano material neural network by using oxidation/reduction molecular adsorption and a novel ternary molecular neural network obtained based on the regulating and controlling method. The electric pulse distribution behavior of the polyoxometallate-semiconductor low-dimensional nano material neural network is regulated and controlled by regulating and controlling the adsorption of oxidation/reduction molecules on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure and changing the charge transfer between the polyoxometallate and the semiconductor low-dimensional nano material, so that the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is changed, and further the charge potential barrier and the corresponding charge transmission mechanism between polyoxometallate-semiconductor low-dimensional nano material composite structure units are changed, and the technical problem that the electric pulse behavior in the molecular neural network is uncontrollable in the prior art is solved.

Description

Method for regulating and controlling electric pulse distribution behavior of nano molecular neural network

Technical Field

The invention belongs to the technical field of artificial intelligence, and relates to a method for regulating and controlling electric pulse distribution behaviors of a nano molecular neural network. More particularly, relates to a method for regulating and controlling the electric pulse emission behavior of a polyoxometallate-semiconductor low-dimensional nano material neural network by using oxidation/reduction molecular adsorption.

Background

Semiconductor nanowires, nanobelts, and nanotubes such as Carbon Nanotubes (CNTs), graphene nanobelts, GaAs nanowires, and other low-dimensional materials have been widely studied due to their unique physical and chemical properties and excellent electrical properties, and are currently receiving attention from the research field of novel neuromorphic devices oriented to artificial intelligence.

The Hirofumi team at Osaka university of Japan in 2018 proposed phosphomolybdic acid molecule (H) for the first time₃PMo₁₂O₄₀) The modified single-walled carbon nanotube (SCNT) network constructs an ultra-high density nano-molecular neural network. The electronic structure and the conductivity of the CNT are directly regulated and controlled by utilizing phosphomolybdic acid molecule modification with multi-electron oxidation-reduction performance, the SCNT in the device does not need to be purely semiconductive, and different from a conventional carbon nanotube field effect tube, the electrical noise generated by the metallic SCNT can provide a rich neural network dynamic environment. The phosphomolybdic acid/CNT molecules have strong electron storage capacity and can release electrons along the CNT network to generate pulse action, the generation image of electric pulse of the experimental demonstration of the phosphomolybdic acid/CNT molecules is shown in figure 1, and a Hirofumi team provides a two-dimensional automatic cellular model for the SCNT molecular network, so that the basic learning function based on the storage pool calculation is simulated. Therefore, a new research direction based on CNT is opened for the ultra-high capacity neuromorphic hardware system.

Phosphomolybdic acid is a Polyoxometallate (POMs), a class of polyoxometallate compounds formed by linking transition metal ions through oxygen, and the interaction between metal ions in these systems through electron transfer and their interaction with end-group ligands make them exhibit many specific physical functions, chemical properties and biological activities. As shown in fig. 2(a) and 2(b), phosphomolybdic acid molecules can store or release up to 24 electrons.

However, the impulse generation mechanism of the phosphomolybdic acid/CNT molecular neural network proposed by Hirofumi team is not revealed from experimental aspects, so that the regulation of the electric impulse behaviors such as the electric impulse distribution amplitude, frequency and probability cannot be realized, and the network cannot be designed in a forward direction.

Disclosure of Invention

Aiming at the defects and the improvement requirements of the prior art of the nerve morphological device and the network of the polyoxometallate-semiconductor low-dimensional nano material, the invention provides a method for regulating and controlling the electric pulse distribution behavior of the molecular nerve network through the adsorption of oxidation/reduction molecules, it utilizes different types and concentrations of oxidation/reduction molecules to be adsorbed on the surface of polyoxometallate or semiconductor low-dimensional nano material, so that the charge transfer in the polyoxometallate and semiconductor low-dimensional nano material system is different, so that the conductance of the composite structure unit formed by the polymetallic hydrochloride and the semiconductor low-dimensional nano material, the charge barrier between the units and the charge transmission rule can be regulated and controlled, thereby regulating and controlling the electric pulse emission behavior of the polyoxometallate-semiconductor low-dimensional nano material neural network, therefore, the technical problem that the electric pulse behaviors in the molecular neural network are uncontrollable in the prior art is solved.

To achieve the above object, according to one aspect of the present invention, there is provided a method for regulating and controlling an electric pulse sending behavior of a polyoxometallate-semiconductor low-dimensional nanomaterial neural network by using an oxidation/reduction molecule, wherein the neural network is a molecular neural network constructed based on a polyoxometallate-semiconductor low-dimensional nanomaterial composite structure, and the adsorption of the oxidation/reduction molecule on the surface of the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure is regulated to change charge transfer between polyoxometallate and a semiconductor low-dimensional nanomaterial, thereby changing the conductance of the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure, and further changing a charge barrier and a corresponding charge transfer mechanism between polyoxometallate-semiconductor low-dimensional nanomaterial composite structure units, thereby regulating and controlling the electric pulse issuing behavior of the polyoxometallate-semiconductor low-dimensional nano material neural network; the oxidation/reduction molecules are molecules capable of generating electron gain and loss with the polyoxometallate or the semiconductor low-dimensional nano material; the semiconductor low-dimensional nano material is a semiconductor one-dimensional nano material or a semiconductor two-dimensional nano material;

when the reducing molecules are adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, part of electrons of the reducing molecules are transferred into the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, the hole concentration in the semiconductor low-dimensional nano material is reduced, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is reduced, the charge potential barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is increased, and the amplitude and the frequency of electric pulses emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are reduced.

When the oxidizing molecules are adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, electrons are obtained from the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, so that the hole concentration in the semiconductor low-dimensional nano material is increased, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is increased, the charge barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure units is reduced, and the amplitude and the frequency of electric pulses distributed in the polyoxometallate-semiconductor low-dimensional nano material neural network are increased.

Preferably, the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure is immersed in the oxidation/reduction molecule environment, and the oxidation/reduction molecule is adsorbed on the surface of the polyoxometallate or semiconductor low-dimensional nanomaterial by standing, pressurizing, heating or electrifying.

Preferably, the adsorption behavior of the oxidation/reduction molecules on the surface of the polyoxometallate or the semiconductor low-dimensional nano material is regulated by regulating the species, concentration or immersion process parameters of the oxidation/reduction molecules.

Preferably, the polyoxometallate-semiconductor low-dimensional nano material composite structure is placed in a closed container, after vacuum pumping, nitrogen or inert gas is introduced to normal pressure, and then the oxidation/reduction molecules with required types and concentrations are introduced, so that the oxidation/reduction molecules are used as a regulation source for regulating and controlling the conductance of the composite structure unit, the charge barrier between units and a charge transmission mechanism.

Preferably, the structure of the molecular neural network comprises, from bottom to top: the semiconductor low-dimensional nano material modified by polyoxometallate molecules is the polyoxometallate-semiconductor low-dimensional nano material composite structure;

the thin film electrode is a metal thin film electrode, a graphene electrode or a composite thin film electrode of graphene and transition metal; the number of the film electrodes is multiple, and a film electrode array is formed;

the semiconductor low-dimensional nano material modified by polyoxometallate molecules is a semiconductor low-dimensional nano material adsorbed with polyoxometallate molecules, and the semiconductor low-dimensional nano material modified by the polyoxometallate molecules is erected between the film electrodes.

According to another aspect of the invention, the ternary molecular neural network is obtained by regulating and controlling the method, and is constructed on the basis of a composite structure of the oxidation/reduction molecule, the polyoxometallate and the semiconductor low-dimensional nano material.

Preferably, the structure of the ternary molecular neural network comprises, from bottom to top: the device comprises a substrate, a film electrode, a semiconductor low-dimensional nano material modified by polyoxometallate molecules and oxidation/reduction molecules adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure.

Preferably, the thin film electrode is a graphene electrode, and the polyoxometalate molecule-modified semiconductor low-dimensional nanomaterial is connected with the graphene electrode through van der waals force; or the thin film electrode is a graphene/transition metal composite thin film electrode, and the two ends of the semiconductor low-dimensional nanomaterial modified by polyoxometallate molecules are connected with graphene in the graphene/transition metal composite thin film electrode through covalent bonds after being melted.

Preferably, the preparation of the ternary molecular neural network comprises the following steps:

(1) preparing a thin film electrode array for pulse signal acquisition on the surface of the substrate;

(2) ultrasonically mixing the solution of the semiconductor low-dimensional nano material with the solution of polyoxometallate to prepare a mixed solution of the semiconductor low-dimensional nano material modified by polyoxometallate molecules;

(3) by adopting a dielectrophoresis technology, dripping the mixed solution of the semiconductor low-dimensional nano material modified by polyoxometallate molecules between the film electrodes, and introducing alternating voltage to ensure that the polyoxometallate-semiconductor low-dimensional nano material composite structure is assembled between the film electrodes; then dropping an organic solvent in the middle of the electrode to remove the residual polyoxometallate solution, and obtaining a binary molecular neural network which is built by taking the polyoxometallate-semiconductor low-dimensional nano material composite structure as a basic unit and has a topological structure or random connection, wherein the organic solvent is preferably acetone solution;

(4) and immersing the polyoxometallate-semiconductor low-dimensional nano material composite structure in the binary molecular neural network in the oxidation/reduction molecular environment, so that the oxidation/reduction molecules are adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure, and obtaining the ternary molecular neural network.

(2) ultrasonically mixing the semiconductor low-dimensional nano material solution with the polyoxometallate solution to prepare a semiconductor low-dimensional nano material mixed solution modified by polyoxometallate molecules;

(3) performing membrane filtration on the mixed solution of the polyoxometallate molecule-modified semiconductor low-dimensional nano materials obtained in the step (2), covering one surface, containing the semiconductor low-dimensional nano materials/polyoxometallate, of the obtained filtration film on a multi-electrode array, and dissolving and removing the filtration film by using an organic chemical solvent to obtain a binary molecular neural network which is built by taking the polyoxometallate-semiconductor low-dimensional nano material composite structure as a basic unit and has a topological structure or random connection;

(2) assembling semiconductor low-dimensional nano materials between film electrodes by adopting a dielectrophoresis technology, then dripping polyoxometallate solution between the film electrodes, and standing for 1-2 hours;

(3) dropping an organic solvent between the film electrodes to remove residual polyoxometallate solution, and obtaining a binary molecular neural network which is built by taking the polyoxometallate-semiconductor low-dimensional nano material composite structure as a basic unit and has a topological structure or random connection;

Preferably, the polyoxometallate is a phosphomolybdic acid molecule or a phosphotungstic acid molecule, and the semiconductor low-dimensional nanomaterial is a semiconductor one-dimensional nanomaterial or a semiconductor two-dimensional nanomaterial.

Preferably, the semiconductor low-dimensional nanomaterial comprises a semiconductor nanowire, a semiconductor nanobelt, or a semiconductor nanotube.

Preferably, the oxidation/reduction molecules are redox gas molecules, redox liquid molecules or redox solid molecules.

Preferably, the redox gas molecules are one or more of ammonia, nitrogen dioxide, sulphur dioxide, ethanol and water molecules.

Preferably, the redox liquid molecule is toluene or hydrogen peroxide.

In general, the invention is directed to a molecular neural network formed by compounding two materials, namely polyoxometallate and a semiconductor low-dimensional nano material, the invention is called as the existing binary molecular neural network, proposes that the oxidation/reduction micromolecules are adsorbed on polyoxometallate and semiconductor low-dimensional nano materials, by regulating and controlling the charge transfer between the semiconductor nanowire/strip/tube and the polyoxometallate, the conductance of the semiconductor nanowire/strip/tube and the polyoxometallate composite structure unit, the charge barrier between the units and the corresponding charge transmission mechanism, so that the electric pulse distribution characteristic in the molecular network constructed by the semiconductor nano-wire/belt/tube and the polyoxometallate can be adjusted, meanwhile, the electric pulse distribution mechanism in the molecular network can be revealed by utilizing the difference between different small molecules. Compared with the prior art, the technical scheme of the invention can obtain the following beneficial effects:

(1) the invention provides a method for regulating and controlling the electric pulse distribution behavior of a polyoxometallate-semiconductor low-dimensional nano material neural network by utilizing oxidation/reduction molecule adsorption, which changes the charge transfer between polyoxometallate and a semiconductor low-dimensional nano material by regulating and controlling the adsorption of oxidation/reduction molecules on the surface of a polyoxometallate-semiconductor low-dimensional nano material composite structure, thereby changing the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure, further changing the charge potential barrier and the corresponding charge transmission mechanism between polyoxometallate-semiconductor low-dimensional nano material composite structure units, and further regulating and controlling the electric pulse distribution behavior of the polyoxometallate-semiconductor low-dimensional nano material neural network.

(2) The invention provides a method for regulating and controlling the electric pulse issuing behavior of a neural network, which provides a feasible electric pulse regulating and controlling method for a network constructed based on semiconductor nanowires/strips/tubes-polyoxometallate, so that the learning mechanism and parameters of the network are controlled when the network is applied to artificial intelligence application, and meanwhile, the method has the characteristics of strong feasibility, high controllability and simple experiment.

(3) The invention verifies the charge transfer regulation and control of a single carbon nano tube/polyoxometallate by gas molecules through calculation of a first principle, verifies the principle and the feasibility of the regulation and control method by utilizing the charge transfer regulation and control mechanism and using a network simulation result based on a cellular automaton and a partial experimental result of the conductance change of the carbon nano tube/polyoxometallate molecular network after the gas molecules are adsorbed, and reasonably expands the regulation and control method simultaneously so that the regulation and control method can be suitable for other semiconductor low-dimensional nano materials. The invention utilizes oxidation/reduction molecules to control the charge transfer and transmission mechanism in the polyoxometallate-semiconductor low-dimensional nano material composite network structure, so that the charge transmission threshold value among the single polyoxometallate-semiconductor low-dimensional nano material composite structure units can be regulated and controlled, and the distribution characteristics of electric pulses in the constructed molecular neural network, such as the distribution frequency, the amplitude and the like of the electric pulses, can be regulated and controlled, and the polyoxometallate-semiconductor low-dimensional nano material composite network with the required specific electric pulse behavior can be obtained.

(4) The invention also provides a novel ternary molecular neural network obtained based on the regulation and control method, the ternary molecular neural network is built based on a composite structure of the oxidation/reduction molecules, the polyoxometallate and the semiconductor low-dimensional nano material, and compared with the traditional binary molecular neural network formed by taking a composite structure of two materials, namely the polyoxometallate and the semiconductor low-dimensional nano material, as a basic core unit, the oxidation/reduction molecules adsorbed on the surface of the composite structure are increased, so that the novel ternary molecular neural network has stronger controllability, for example, the controllability of electric pulse behaviors is obviously enhanced and has designability.

Drawings

FIG. 1 is a current pulse amplitude versus time curve at both ends of a random network electrode constructed by the polyoxometallate and carbon nanotube composite structure of the present invention.

FIG. 2 is a schematic representation of the molecular structure of phosphomolybdic acid of the present invention, wherein (a) is [ PMo ]₁₂O₄₀]^3-The molecular model of (1). (b) Is [ PMo₁₂O₄₀]^27-The molecular model of (1).

FIG. 3 is a schematic diagram of the polyoxometallate-carbon nanotube composite structure (a) and a model thereof after different kinds of oxidation/reduction small molecules are adsorbed thereon (b, c, d, e, f).

FIG. 4 is a schematic diagram of the composite structure unit of polyoxometallate and carbon nanotube and the structure of two-end electrodes.

FIG. 5 is a schematic diagram of a thin film electrode array assembled with a polyoxometallate and carbon nanotube composite structure according to the present invention, wherein (a) is a three-dimensional schematic diagram of the array electrode, and (b) is an enlarged detail view selected from the group consisting of the diagram in a.

FIG. 6 is a three-dimensional schematic diagram of a composite structure of a plurality of polyoxometallate molecules and carbon nanotubes according to the present invention.

Fig. 7 is a three-dimensional schematic diagram of the polyoxometallate and graphene nanoribbon composite structure of the invention, wherein (a) is a front view of the three-dimensional structure, and (b) is a top view of the three-dimensional structure.

FIG. 8 is a graph of the electron state transition probability function in the simulation model of the automatic cellular automaton of the present invention.

FIG. 9 is a flow chart of cellular automata simulation in the present invention.

FIG. 10 is a schematic diagram of the cellular automaton simulation random network structure according to the present invention.

FIG. 11 is a diagram illustrating the simulation of electrical pulses across the electrodes of the random network of FIG. 3 using cellular automata in accordance with the present invention.

Fig. 12 is a graph illustrating the number of charge transfers between phosphomolybdic acid and carbon nanotubes after different redox small molecules modify phosphomolybdic acid according to the first principle.

FIG. 13 is a graph of the rate of change of resistance of polyoxometallate-carbon nanotube composite structures using different concentrations of small oxidation/reduction molecules in accordance with the present invention.

FIG. 14 is an atomic force microscope schematic view of the polyoxometallate-phosphomolybdic acid composite structure of the present invention.

FIG. 15 is an SEM image of a single carbon nanotube/phosphomolybdic acid composite structure according to the present invention.

FIG. 16 is an SEM image of a two-terminal electrode random network constructed by the carbon nanotube/phosphomolybdic acid molecular neural network composite structure of the present invention.

FIG. 17 is a schematic diagram of the simulation of the electric pulse after the regulation and control by ammonia gas according to the present invention.

FIG. 18 is a simulated image of an electrical pulse without conditioning according to the present invention.

FIG. 19 is a schematic diagram of the simulation of the electrical pulse after the nitrogen dioxide regulation.

The same reference numbers will be used throughout the drawings to refer to the same or like elements or structures, wherein:

1-carbon nanotubes; 2-polyoxometallate; 3-graphene nanoribbons; 4-a thin film electrode; 5-a gate dielectric layer; 6-substrate.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

The invention provides a method for regulating and controlling the electric pulse distribution behavior of a polyoxometallate-semiconductor low-dimensional nano material neural network by using oxidation/reduction molecules, which changes the charge transfer between polyoxometallate and a semiconductor low-dimensional nano material by regulating and controlling the adsorption of the oxidation/reduction molecules on the surface of a polyoxometallate-semiconductor low-dimensional nano material composite structure, thereby changing the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure, further changing the charge potential barrier and the corresponding charge transmission mechanism between polyoxometallate-semiconductor low-dimensional nano material composite structure units, and regulating and controlling the electric pulse distribution behavior of the polyoxometallate-semiconductor low-dimensional nano material neural network; the oxidation/reduction molecules are molecules capable of generating electron gain and loss with the polyoxometallate or the semiconductor low-dimensional nano material; the semiconductor low-dimensional nano material is a semiconductor one-dimensional nano material or a semiconductor two-dimensional nano material;

when the reducing molecules are adsorbed on the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, part of electrons are transferred to the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, the hole concentration in the semiconductor low-dimensional nano material is reduced, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is reduced, the charge potential barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is increased, and the amplitude and the frequency of electric pulses emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are reduced.

When the oxidizing molecules are adsorbed on the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, electrons are obtained from the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, so that the hole concentration in the semiconductor low-dimensional nano material is increased, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is increased, the charge potential barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is reduced, and the amplitude and the frequency of electric pulses distributed in the polyoxometallate-semiconductor low-dimensional nano material neural network are increased.

The invention provides a method for regulating and controlling the electric pulse distribution behavior of a polyoxometallate-semiconductor low-dimensional nano material neural network by using oxidation/reduction molecules. The electric pulse distribution behavior of the neural network is regulated and controlled by regulating and controlling the adsorption behavior of the oxidation/reduction molecules on the surfaces of the polyoxometallate and the semiconductor low-dimensional nano material. The oxidation/reduction molecules can be adsorbed on polyoxometallate, also can be adsorbed on the surface of a semiconductor low-dimensional nano material, or can be adsorbed on the surfaces of the polyoxometallate and the semiconductor low-dimensional nano material simultaneously, and the adsorption conditions are uniformly described as being adsorbed on the surface of a composite structure unit of the polyoxometallate-semiconductor low-dimensional nano material; because the adsorption of the oxidation/reduction molecules changes the charge distribution state in the basic unit, the regulation mechanism of the electric pulse issuing behavior of the neural network is the same, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is changed by changing the charge transfer between the polyoxometallate and the semiconductor low-dimensional nano material, and the charge potential barrier and the corresponding charge transmission mechanism between the polyoxometallate-semiconductor low-dimensional nano material composite structure units are changed, so that the electric pulse issuing behavior of the molecular neural network is influenced; therefore, the electric pulse distribution behavior in the polyoxometallate-semiconductor low-dimensional nano material neural network can be regulated and controlled by the regulation and control method.

In the oxidation/reduction molecule of the present invention, "/" is "or" means an oxidative molecule or a reductive molecule.

When the reducing molecules are adsorbed on the surface of the polyoxometallate in the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, part of electrons in the reducing molecules are transferred to the polyoxometallate, so that the number of the electrons transferred to the polyoxometallate molecules in the semiconductor low-dimensional nano material is reduced, the hole concentration in the semiconductor low-dimensional nano material is reduced, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is reduced, the charge barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure units is increased, and the amplitude and the frequency of electric pulses emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are reduced.

When the reducing molecules are adsorbed on the surface of the semiconductor low-dimensional material in the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, part of electrons in the reducing molecules are transferred to the low-dimensional nano material, so that the hole concentration in the semiconductor low-dimensional nano material is reduced, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is reduced, the charge potential barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is increased, and the amplitude and the frequency of electric pulses emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are reduced.

When the reducing molecules are simultaneously adsorbed on the surfaces of the polyoxometallate and the semiconductor low-dimensional nano material, part of electrons in the reducing molecules are transferred to the polyoxometallate molecules and the semiconductor low-dimensional nano material, the number of electrons transferred to polyoxometallate molecules in the semiconductor low-dimensional nano material is reduced, meanwhile, electrons in the reducing molecules adsorbed on the surface of the semiconductor low-dimensional nano material are also transferred into the semiconductor low-dimensional nano material, the action mechanisms of the electrons and the reducing molecules are to reduce the hole concentration in the semiconductor low-dimensional nano material, so that the electric conductance of the corresponding polyoxometallate-semiconductor low-dimensional nano material composite structure is reduced, and the charge potential barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure units is increased, so that the amplitude and the frequency of the electric pulse emitted in the polyoxometallate-semiconductor low-dimensional nano-material neural network are reduced.

When the oxidizing molecules are adsorbed on the polyoxometallate surface in the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, part of electrons in the polyoxometallate molecules are transferred into the oxidizing molecules, so that the number of the electrons transferred to the polyoxometallate molecules in the semiconductor low-dimensional nano material is increased, the hole concentration in the semiconductor low-dimensional nano material is increased, the charge barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is reduced, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is increased, and the amplitude and the frequency of electric pulses emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are increased.

When the oxidative reductive molecules are adsorbed on the surface of the semiconductor low-dimensional material in the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, part of electrons in the semiconductor low-dimensional nano material are transferred to the oxidative molecules, so that the hole concentration in the semiconductor low-dimensional nano material is increased, the charge potential barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is reduced, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is increased, and the amplitude and the frequency of electric pulses emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are increased.

When the oxidizing molecules are adsorbed on the surfaces of the polyoxometallate and the semiconductor low-dimensional nano material at the same time, partial electrons in the polyoxometallate and the semiconductor low-dimensional nano material are transferred into the oxidizing molecules, and the action mechanisms of the polyoxometallate and the semiconductor low-dimensional nano material are that the hole concentration in the semiconductor low-dimensional nano material is increased, so that the charge potential barrier between the polyoxometallate and semiconductor low-dimensional nano material composite structure units is reduced, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is increased, and the amplitude and the frequency of electric pulses emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are increased.

In some embodiments of the present invention, the polyoxometallate is a phosphomolybdic acid molecule or a phosphotungstic acid molecule, the semiconductor nanomaterial is a low-dimensional semiconductor nanomaterial, and the low-dimensional semiconductor nanomaterial of the present invention refers to a one-dimensional or two-dimensional semiconductor nanomaterial. The invention is preferably a semiconductor nanowire/strip/tube, which refers to a strip, belt or tubular material at the nanoscale, such as a carbon nanotube, a graphene nanostrip, a GaAs nanowire and the like; for example, the material may be a semiconducting carbon nanowire, a semiconducting graphene tape, a semiconducting carbon nanotube, or a metallic carbon nanotube; single-walled semiconducting carbon nanotubes are preferred.

The oxidation/reduction molecules described in the present invention may be redox gas molecules, or may be redox solid or liquid molecules. In some embodiments, the oxidation/reduction molecules are small molecules of redox gases including, but not limited to, one or more of ammonia, nitrogen dioxide, sulfur dioxide, hydrogen sulfide, ethanol, and water molecules, and fig. 3 is a schematic diagram of a structure in which various gases are adsorbed on polyoxometallate, wherein polyoxometallate is adsorbed on the surface of carbon nanotubes. Wherein fig. 3(a) corresponds to a comparison graph of adsorbed oxidation/reduction molecules, and fig. 3(b), fig. 3(c), fig. 3(d), fig. 3(e) and fig. 3(f) correspond to molecules of ammonia gas, sulfur dioxide, carbon monoxide, hydrogen gas and sulfur dioxide adsorbed on the surfaces of polyoxometalate molecules, respectively.

In some examples, the oxidation/reduction molecules are redox liquid molecules, including but not limited to toluene and hydrogen peroxide molecules.

In some embodiments, the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure is immersed in an environment containing the oxidation/reduction molecules, and the oxidation/reduction molecules are adsorbed on the surface of the polyoxometallate or semiconductor low-dimensional nanomaterial by standing, pressurizing, heating or electrifying. The adsorption can be realized by directly standing for several hours, or the adsorption of the oxidation/reduction molecules on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure is promoted by using a pressurizing, heating or electrifying mode.

The adsorption of the oxidation/reduction molecules on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure means that the oxidation/reduction molecules are in direct contact with the polyoxometallate molecules or the semiconductor low-dimensional nano materials. The balance adsorption quantity of the oxidation/reduction molecules on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure can be changed by regulating the type and concentration of the oxidation/reduction molecules adsorbed on the surface of the polyoxometallate or semiconductor low-dimensional nano material or by adjusting the process parameters of the polyoxometallate-semiconductor low-dimensional nano material composite structure immersed in the environment containing the oxidation/reduction molecules.

At present, the precise adsorption quantity of the oxidation/reduction molecules on the surface of polyoxometallate or semiconductor low-dimensional nano materials cannot be obtained in experiments, but according to the regulation and control method disclosed by the invention, the electric pulse amplitude and frequency of the molecular neural network disclosed by the invention can be increased or decreased by changing experimental conditions, such as the types or concentrations of the oxidation/reduction molecules and controlling the process parameters of the adsorption quantity of the oxidation/reduction molecules. In some embodiments, when the oxidation/reduction molecule is a gas molecule, the volume fraction of the oxidation/reduction molecule in the system is no greater than 10%. The system may contain a shielding gas, which may be nitrogen or an inert gas. The adsorption quantity of the oxidation/reduction molecules on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure is adjusted by adjusting the concentration of the oxidation-reduction gas molecules in the system or regulating the neural network immersion process of the polyoxometallate-semiconductor low-dimensional nano material composite structure, for example, the equilibrium adsorption quantity of the oxidation/reduction molecules on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure can be adjusted by adjusting the immersion time, heating temperature, electrified voltage, pressurizing pressure and other parameters of the composite structure in the environment of the oxidation/reduction molecules. Specifically, in some embodiments, the immersion time is not less than 1 hour when the composite structure is immersed in an oxidizing/reducing molecule environment to adsorb the oxidizing/reducing molecules on the surface of the polyoxometalate-semiconductor low-dimensional nanomaterial composite structure.

In some embodiments, the oxidation/reduction molecules are gas molecules, the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure, or the binary molecular neural network with a topological structure or random connection constructed by taking the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure as a basic unit is placed in a closed container, after vacuum pumping, nitrogen or inert gas is introduced to normal pressure, and the oxidation/reduction molecules with required types and concentrations are introduced to serve as the regulation source in the invention.

In some embodiments, when the oxidation/reduction molecule is a liquid molecule, the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure, or the neural network with a topological structure or random connection constructed by the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure as a basic unit is immersed in a redox liquid molecule system, and the adsorption of the oxidation/reduction molecule on the surface of the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure is promoted by standing for several hours or by adopting a pressurizing, heating or electrifying mode.

In some embodiments, the structure of the binary molecular neural network constructed by the polyoxometallate-semiconductor low-dimensional nano material composite structure as a basic unit comprises from bottom to top: the composite structure comprises a substrate, a film electrode, a semiconductor low-dimensional nano material modified by polyoxometallate molecules and oxidation/reduction molecules adsorbed on the surface of the polyoxometallate or the semiconductor low-dimensional nano material, wherein in some examples, the composite structure schematic diagram is shown in figure 4, and the composite structure schematic diagram comprises a 6-substrate, a 5-gate dielectric layer, a 4-film electrode, a 1-carbon nanotube and a 2-polyoxometallate from bottom to top respectively. The thin film electrode is a metal thin film electrode, a graphene electrode or a composite thin film electrode of graphene and transition metal; the number of the film electrodes is multiple, and a film electrode array is formed, as shown in fig. 5, the polyoxometallate and carbon nano tube composite structure unit is built at two ends of the electrode, the electrode is used for collecting and outputting electric signals, and the lower graph of fig. 5 is a partial enlarged graph of the upper graph. The semiconductor low-dimensional nano material modified by polyoxometallate molecules is a semiconductor low-dimensional nano material adsorbed with polyoxometallate molecules, and the semiconductor low-dimensional nano material modified by the polyoxometallate molecules is erected between the film electrodes.

The composite structure of polyoxometallate and a semiconductor low-dimensional nano material is a structural state that polyoxometallate molecules are adsorbed on the surface of the semiconductor low-dimensional nano material and are in contact with the surface of the semiconductor low-dimensional nano material through physical means such as ultrasonic mixing and heating and chemical means such as modifying the surface of the polyoxometallate and the surface of the nano material. The invention constructs a topological structure network by taking the composite structure as a basic unit, and also comprises a composite network structure which is randomly constructed or constructed according to a certain topological structure.

In some embodiments, the topological structure network is a random network, a fully-connected neural network, a recurrent neural network, an echogenic neural network, a fluid state machine, a back propagation network, a self-organizing map, a Hopfield network, a Boltzmann machine, or the like.

In some embodiments, the thin film electrode is a graphene electrode, and the polyoxometalate molecule-modified semiconductor low-dimensional nanomaterial is connected with the graphene electrode by van der waals force; or the thin film electrode is a graphene/transition metal composite thin film electrode, and the two ends of the semiconductor low-dimensional nanomaterial modified by polyoxometallate molecules are connected with graphene in the graphene/transition metal composite thin film electrode through covalent bonds after being melted.

In some embodiments, the substrate material is Si, SiO₂、SiO₂and/Si, GaAs, GaN, SiC, BN, ceramic, sapphire, or the like.

In some embodiments, the thin film electrode array has a thickness of 50nm to 1 μm; the distance between the thin film electrodes is 0.1 to 500 μm, and the width of the thin film electrodes is 1 to 1000 nm.

In some embodiments, the polyoxometalate molecule-modified semiconductor low-dimensional nanomaterial, i.e., the polyoxometalate-semiconductor low-dimensional nanomaterial composite structure, is obtained by ultrasonically mixing a solution of the semiconductor low-dimensional nanomaterial with a solution of polyoxometalate.

In some embodiments, the polyoxometallate solution is dripped on the semiconductor low-dimensional nano material arranged between the film electrodes, and the semiconductor low-dimensional nano material modified by polyoxometallate molecules is obtained after standing.

In some embodiments, the concentration of the solution of the semiconductor low-dimensional nano material is 0.001-1000 mug/ml; the concentration of the polyoxometallate solution is 1 mu g/ml to 5000 mu g/ml; and standing for 2-30 min.

In some embodiments, the solvent of the solution of the semiconductor low-dimensional nanomaterial or the solution of the polyoxometalate is an organic solvent which is easy to dissolve the semiconductor low-dimensional nanomaterial or the polyoxometalate, including Dimethylformamide (DMF).

The invention also provides a ternary molecular neural network obtained based on the regulation and control method, wherein the ternary molecular neural network is built based on a composite structure of the oxidation/reduction molecules, the polyoxometallate and the semiconductor low-dimensional nano material.

In some embodiments, the structure of the ternary molecular neural network comprises, from bottom to top: the device comprises a substrate, a film electrode, a semiconductor low-dimensional nano material modified by polyoxometallate molecules and oxidation/reduction molecules adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure.

In some embodiments, the preparation of the ternary molecular neural network of the present invention comprises the following steps:

(2) mixing the solution of the semiconductor low-dimensional nano material with the solution of polyoxometallate and carrying out ultrasonic treatment to prepare a mixed solution of the semiconductor low-dimensional nano material modified by polyoxometallate molecules;

(4) and immersing the polyoxometallate-semiconductor low-dimensional nano material composite structure in the binary molecular neural network structure in the oxidation/reduction molecular environment, so that the oxidation/reduction molecules are adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure, and obtaining the ternary molecular neural network.

In some embodiments, the thin film electrode is a metal thin film electrode, and the step (1) includes the following sub-steps:

(1.1) preparing a turnover pattern of an electrode on the surface of the substrate by adopting a photoetching process; the substrate material is Si or SiO₂、SiO₂Any one of/Si, GaN, GaAs, ceramic or sapphire.

(1.2) depositing a metal film on the surface of the substrate by adopting a physical vapor deposition process; the thickness of the metal film is 20 nm-1.64 mu m;

(1.3) preparing a metal film electrode on the surface of the substrate by adopting a stripping process; the electrode spacing is 0.1 to 500 μm and the electrode width is 1 to 1000 nm.

In some embodiments, the concentration of the solution of the semiconductor low-dimensional nano material is 0.001-1000 mug/ml; the concentration of the polyoxometallate solution is 1 mu g/ml to 5000 mu g/ml; the mixed ultrasonic power of the solution of the semiconductor low-dimensional nano material and the solution of the polyoxometallate is 5W-300W, and the ultrasonic time is 5-30 h.

In some embodiments, the voltage of the AC signal used for dielectrophoresis is 10-18V_ppThe frequency is 1-10 MHz; the volume of the dielectrophoresis assembly semiconductor low-dimensional nano material solution is 1-15 mu l.

In some embodiments, the step (3) can be carried out by preparing a polyoxometallate-semiconductor low-dimensional nanomaterial composite neural network, and transferring the polyoxometallate-semiconductor low-dimensional nanomaterial composite neural network onto the thin film electrode array by means of transfer. Correspondingly, the preparation of the ternary molecular neural network comprises the following steps:

(3) performing membrane filtration on the mixed solution of the semiconductor low-dimensional nano materials modified by the polyoxometallate molecules obtained in the step (2), covering one surface, which contains the semiconductor low-dimensional nano materials/polyoxometallate, of the obtained filtration membrane on a multi-electrode array, and dissolving and removing the filtration membrane by using an organic chemical solvent such as acetone; obtaining a binary molecular neural network which is built by taking the polyoxometallate-semiconductor low-dimensional nano material composite structure as a basic unit and has a topological structure or random connection;

In other embodiments, the preparation of the ternary molecular neural network of the present invention comprises the steps of:

(2) assembling semiconductor low-dimensional nano materials between film electrodes by adopting a dielectrophoresis technology, and then dripping polyoxometallate solution between the film electrodes, wherein the polyoxometallate is dissolved in a DMF (dimethyl formamide) solution; standing for 1-2 hours;

(3) then dropping an organic solvent between the film electrodes to remove residual polyoxometallate solution, and obtaining a binary molecular neural network which is built by taking the polyoxometallate-semiconductor low-dimensional nano material composite structure as a basic unit and has a topological structure or random connection, wherein the organic solvent is preferably acetone solution;

In the above three methods, for step (4):

when the oxidation/reduction molecules are gas molecules, the polyoxometallate-semiconductor low-dimensional nano material composite structure or the binary molecular neural network is integrally placed in a closed container, vacuum pumping is performed, nitrogen or inert gas is introduced to normal pressure to eliminate the influence of water vapor in the environment, then the oxidation-reduction molecules with preset types and concentrations are introduced, and the ternary molecular network modified and regulated by the oxidation/reduction molecules is obtained after the system is stabilized.

When the oxidation/reduction molecules are liquid molecules, the polyoxometallate-semiconductor low-dimensional nano material composite structure or the binary molecular neural network is wholly immersed in the oxidation-reduction liquid molecular system, the oxidation/reduction molecules are adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure through standing, pressurizing, heating or electrifying and the like, and the ternary molecular network modified and regulated by the oxidation/reduction molecules is obtained after the system is stabilized.

In some embodiments, the molecular network modified and regulated by the oxidized/reduced molecules is encapsulated and stored and used at a low temperature to prevent desorption of the molecules.

In some embodiments, the oxidation/reduction molecules are gas molecules, and the operation is performed using an AES-4TH intelligent gas analysis system, wherein step (4) comprises the sub-steps of:

(4.1) vacuumizing, and vacuumizing the pressure in the container to be lower than 100 Pa;

(4.2) introducing a protective gas, i.e. introducing N₂Inert gases such as gas or argon and the like are carried out to normal pressure;

(4.3) introducing redox molecules of predetermined type and concentration.

And (4.4) measuring the resistance change between the electrodes to judge the stability of the system, obtaining the ternary molecular neural network modified and regulated by the oxidized/reduced molecules after the system is stable, and judging the regulated state according to the resistance change rate.

The invention improves the electron transfer and transmission mechanism in the composite structure by controlling the relative density of polyoxometallate and oxidation/reduction molecules, namely changing the type and density of the oxidation/reduction molecules adsorbed on the surface of the polyoxometallate-semiconductor nanowire/band/tube composite structure, so that the electric pulse behavior and the charge transmission mechanism generated by the polyoxometallate-semiconductor nanowire/band/tube composite structure can be regulated and controlled to a certain extent. The regulation and control means comprises processing the polyoxometallate and semiconductor nanowire/strip/tube composite structure by using the oxidation/reduction molecules, for example, the composite structure is contained in an oxidation/reduction molecule environment, the oxidation/reduction molecules are adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure by using a pressurizing mode, a heating mode, an electrifying mode and the like, and the adsorption quantity of the oxidation/reduction molecules is controlled by detecting the concentration change of the oxidation/reduction molecules in a sealed container.

The regulation and control result is that the electric pulse behavior of the polyoxometallate-semiconductor nanowire/band/tube composite structure is controlled, and the method specifically comprises the step of changing parameters of an electronic charge transfer function of a polyoxometallate-semiconductor nanowire/band/tube composite unit, such as a charge threshold, the number of electron transmission and the like, so that the amplitude, the frequency and the occurrence probability of the electric pulse are controlled.

The invention provides a method for regulating and controlling the generation of polyoxometallate-semiconductor nanowire/band/tube molecular network electric pulses by using oxidation/reduction molecules, the method uses oxidation/reduction molecules to adsorb on the surface of the polyoxometallate-semiconductor nanowire/strip/tube composite structure, the polyoxometallate-semiconductor nanowire/strip/tube composite structure is shown in figures 6 and 7, so that the conductance between the polyoxometallate and the semiconductor nanowire/strip/tube is regulated and controlled, wherein fig. 6 is a three-dimensional schematic diagram of a polyoxometallate and carbon nanotube composite structure of the invention, fig. 7(a) and fig. 7(b) are a front view and a top view of the three-dimensional schematic diagram of the polyoxometallate and graphene nanoribbon composite structure, respectively, and fig. 6 is a carbon nanotube 1; 2 is polyoxometallate; in fig. 7(a), reference numeral 3 denotes a graphene nanoribbon.

According to the cellular automata simulation model of charge transfer elements between phosphomolybdic acid molecules adsorbed on carbon nanotubes disclosed in the document "A molecular neural network device localization of single-walled carbon nanotubes complex with polyoxometalate, Nature Communications, volume 9, and aryl number:2693 (2018)", the model simulates the electron transfer and electric pulse behavior in a random network built by phosphomolybdic acid and carbon nanotubes, so that the model is also adopted for displaying the regulation and control result of the invention. Simplifying phosphomolybdic acid molecules in the model into a simulation unit, calculating the charge difference between a central unit and a peripheral unit of a cellular by using a cellular automaton in the first step to obtain two adjacent cells with the maximum electron number difference, wherein the electron number difference is marked as delta a_maxIf the number of electrons contained in the central unit cell is less than the threshold charge a_THA finite number of charges N_mRandomly transferring to the cell unit with the largest difference of adjacent electron numbers, wherein the random transfer probability function is P_cThe function is as follows:

wherein:

representing a down-rounding function,. epsilon.is the transfer coefficient,. gamma.is the sensitivity constant,. p and q are probability parameters,. a_i,jThe amount of charge that the cell having coordinates (i, j) in the simulation model has. As shown in fig. 8, where p is 1 and q is 0.95. P_cIndicating the charge transfer conductivity between the cell units and the capacitive nature of the phosphomolybdic acid molecules. Limited discharge charge N_m(Δa_max) Is an exponential relationship based on the markus theory. If the central unit cell contains electrons greater than the threshold charge a_THWhen the charge is discharged, the number of electrons contained in the charge is fully discharged to the adjacent cell with the largest difference value between the numbers of peripheral electrons and the second largest, and the charge discharge ratio is 9: 1. due to charge transfer in the network, adjacent cells typically have an excess charge (greater than or equal to a)_TH) In the next charge refresh cycle, the subsequent discharge causes a chain reaction (propagating discharge) in the network, thereby generating a charge cascade in the cellular automaton system, causing a sharp increase in the number of electrons passing through the feeding electrode, a current pulse across the electrode, and a charge threshold a_THThe size of the phosphomolybdic acid determines different charge transmission mechanisms among phosphomolybdic acid molecules, a specific cellular automata simulation model can be shown in FIG. 9, FIG. 10 is a topological structure schematic diagram of random network distribution, FIG. 11 is an electric pulse behavior of a random molecular neural network at two ends through electrodes, a small inset is an activation state of different units in a corresponding time network, the brighter the unit, the more electrons contained in the unit, the higher the density, the more the networkThe better the activation state of (c). According to the theoretical explanation in the literature, the charge threshold a in the model_THThe physical meaning of (a) has a direct correspondence with the size of a charge barrier generated by phosphomolybdic acid on the carbon nanotube, which can be expressed as the barrier height passed by electron tunneling between phosphomolybdic acid-carbon nanotube composite structural units. Meanwhile, the invention obtains the same verification through the simulation of the first principle. Based on the theoretical description of the literature and the simulation of the first principle and partial experimental verification, the simulation model can reasonably extend to the electron transmission between the polyoxometallate-semiconductor low-dimensional nano material composite structure units on the semiconductor low-dimensional nano material, and the charge threshold value a in the model_THThe corresponding relationship exists between the potential barriers.

When the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is adsorbed with the reductive molecule, electrons in the reductive molecule are transferred to the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, the hole concentration in the semiconductor low-dimensional nano material is reduced, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is reduced, the charge barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is increased, and the amplitude and the frequency of an electric pulse emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are reduced.

When the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is adsorbed by oxidizing molecules, electrons in the polyoxometallate-semiconductor low-dimensional nano material composite structure unit are transferred to the oxidizing/reducing molecules, so that the hole concentration in the semiconductor low-dimensional nano material is increased, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is increased, the charge barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure units is reduced, and the amplitude and the frequency of electric pulses emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are increased; the threshold variation can be regulated and controlled by the molecular type and number adsorbed on the surface of the unit polyoxometallate-semiconductor low-dimensional nano material composite structure. The parameters such as threshold value in an electron transport and transmission mechanism are corrected, so that the electric pulse generating behavior of the composite structure can be regulated and controlled to a certain extent.

The invention combines the first principle calculation, and explains the mechanism and experimental operation method for regulating the electric pulse behavior of the molecular neural network by using the oxidation/reduction molecules based on the molecular network simulation of the cellular automaton and the conductance experimental data after the oxidation/reduction molecules are adsorbed. When the oxidation type molecules are used for regulation, the polyoxometallate molecules and the semiconductor low-dimensional nano material can be deprived of partial electrons, so that the resistance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is increased, the threshold value is increased, and the amplitude and the frequency of electric pulses distributed in the molecular network are reduced; when the reducing molecules are used for regulation, the polyoxometallate molecules and the semiconductor low-dimensional nano material can obtain partial electrons, so that the resistance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is reduced, the threshold value is reduced, and the amplitude and the frequency of electric pulses distributed in the molecular network are increased.

The following are examples:

example 1

By using NH₃The regulation and control method for reducing the behaviors of the electrical pulse in the phosphomolybdic acid/carbon nanotube molecular neural network by modifying phosphomolybdic acid through gas takes a randomly-built molecular network unit structure as an example, and comprises the following steps from bottom to top: a silicon wafer substrate 6 with an oxide layer, a gate dielectric layer 5, a film electrode 4, a carbon nanotube 1 attached with polyoxometallate 2, and NH adsorbed on phosphomolybdic acid molecules₃The polyoxometallate is phosphomolybdic acid molecules, and the structure of the device is shown in figure 4;

NH₃modification of the phosphomolybdic acid molecule such that the phosphomolybdic acid molecule has 0.041 more electrons is shown in FIG. 12, as can be seen in FIG. 13, with 5ppm NH access₃After that, the resistance of the composite structure changed by 1.24% in the front and back, due to NH₃The modification(s) causes a change in the composite conductance.

The preparation steps of the regulation method and the corresponding ternary molecular neural network are as follows:

(1) taking a silicon wafer with an oxide layer as a substrate, and obtaining a reverse pattern of an electrode on the surface of the substrate by adopting a photoetching process;

(2) depositing a 90nm nickel film and a 510nm copper film on the surface of the substrate in sequence by adopting magnetron sputtering;

(3) putting the substrate in acetone for 3min, and removing the photoresist and the nickel/copper film on the photoresist; sequentially placing the nickel/copper electrode in ethanol and deionized water for ultrasonic cleaning for 15min to obtain a nickel/copper electrode on the surface of the substrate, wherein the electrode spacing is 1 micrometer, and the electrode width is 4 micrometers;

(4) mixing and ultrasonically treating a carbon nano tube/DMF solution with the concentration of 0.25 mu g/ml and a phosphomolybdic acid/DMF solution with the concentration of 2.5 mu g/ml, wherein the ultrasonic power is 300W, and the ultrasonic time is 30 h; the surface morphology of the carbon nanotube after phosphomolybdic acid modification is shown in fig. 14, fig. 14 is an AFM test image of the carbon nanotube adsorbing phosphomolybdic acid, the area with higher brightness is a phosphomolybdic acid molecular group, the darker area is the carbon nanotube, phosphomolybdic acid molecules are accumulated and adsorbed on the surface of the carbon nanotube, fig. 15 is a scanning electron microscope image of the carbon nanotube with a single mounted at two ends of the electrode, and fig. 17 is a scanning electron microscope image of a network of a phosphomolybdic acid and carbon nanotube composite structure randomly mounted at two ends of the electrode;

(5) connecting the nickel/copper electrode pair with a signal generator, wherein the sinusoidal signal voltage of the signal generator is 12Vpp, and the frequency is 1 MHz; and (3) dropping 1 mu l of phosphomolybdic acid/carbon nanotube mixed solution on the nickel/copper electrodes by using a liquid transfer machine, so that the phosphomolybdic acid modified carbon nanotubes are assembled between the nickel/copper electrodes.

(6) Phosphomolybdic acid-carbon nanotube composite structure in binary molecular neural network, which is carbon nanotube two-terminal network unit modified based on phosphomolybdic acid, is exposed to 5ppm of NH₃In the atmosphere, obtaining based on said NH₃A ternary molecular neural network built by a composite structure of phosphomolybdate and carbon nano tubes; NH of 5ppm concentration atmosphere₃The network resistance is changed by about 4.4%.

(7) The method analyzes that the carbon nano tube (8,0) modified by phosphomolybdic acid molecules adsorbs single NH by adopting the calculation of a first principle₃The charge transfer of the molecule is shown in FIG. 16. Adsorption of phosphomolybdic acid molecules on the surface of carbon nanotubes, oxidation of phosphomolybdic acidThe characteristic is that 3.47 electrons in the carbon nanotube are transferred to phosphomolybdic acid molecules, holes are generated in the carbon nanotube, and the structure of the ternary molecular neural network is shown in fig. 3 (a). Specifically, when the detected gas is NH₃When reducing gas is used, 0.26 electron in the gas molecule is transferred to the phosphomolybdic acid molecule, so that the electron transfer between the carbon nanotube and the phosphomolybdic acid molecule is weakened, the number of the electrons transferred from the carbon nanotube to the phosphomolybdic acid molecule is reduced by 1.8 electrons, the hole concentration in the carbon nanotube is reduced, the size between networks is increased, and the composite structure and NH are combined₃The position of the adsorption is shown in FIG. 3 (b).

(8) NH is analyzed by utilizing the molecular neural network pair based on the cellular automata to simulate₃After the adsorption on the surface of phosphomolybdic acid, the charge transmission between basic neurons conforms to the Marcus electron transfer theory, and when no redox molecules are added in the simulation of a cellular automaton model, the charge threshold value a is_THIs 6 when NH₃During adsorption, the charge threshold a is determined according to the change ratio of the corresponding charge barrier height_THAt 13, the simulation results of the pulse delivery are shown in FIG. 17, relative adsorption of NH₃Comparing the previous pulse distribution results with fig. 18, it is apparent that the amplitude of the electrical pulse and the distribution frequency are both reduced in the molecular neural network.

Example 2

By using NO₂A regulation and control method for increasing behaviors such as amplitude, frequency and the like of electric pulses in a phosphomolybdic acid/carbon nanotube molecular neural network by modifying phosphomolybdic acid, taking a molecular network unit structure as an example, comprises the following steps from bottom to top: a silicon wafer substrate 6 with an oxide layer, a gate dielectric layer 5, a film electrode 4, a carbon nanotube 1 attached with polyoxometallate 2, and NO adsorbed on phosphomolybdic acid molecules₂The polyoxometallate is phosphomolybdic acid molecules, and the structure of the device is shown in figure 4;

NO₂the modification of phosphomolybdic acid molecules to make phosphomolybdic acid molecules 0.062 electrons less is shown in FIG. 12, and it can be seen in FIG. 13 that 5ppm NO was introduced₂Then, the resistance of the composite structure changed by 3% in the front and rear, due to NO₂The modification(s) causes a change in the composite conductance.

(4) mixing and ultrasonically treating a carbon nano tube/DMF solution with the concentration of 0.25 mu g/ml and a phosphomolybdic acid/DMF solution with the concentration of 2.5 mu g/ml, wherein the ultrasonic power is 300W, and the ultrasonic time is 30 h;

(6) Respectively exposing phosphomolybdic acid-carbon nanotube composite structures in a carbon nanotube two-end random network modified based on phosphomolybdic acid, namely a binary molecular neural network to NO with the concentration of 5ppm₂In the atmosphere, obtaining based on said NO₂A ternary molecular neural network built by a composite structure of phosphomolybdate and carbon nano tubes. 5ppm concentration of NO₂Causing the network resistance to change by around 3.5%.

(7) The carbon nano tube (8,0) modified by phosphomolybdic acid molecules is analyzed to adsorb single NO by adopting the calculation of a first principle₂Charge transfer of the molecule. Phosphomolybdic acid molecules are adsorbed on the surface of the carbon nanotube, and the oxidation characteristic of the phosphomolybdic acid enables 3.47 electrons in the carbon nanotube to be transferred to the phosphomolybdic acid molecules, so that a hole is generated in the carbon nanotube. When oxidizing gas NO₂When adsorbed on phosphomolybdic acid molecules, 0.51 electrons in the phosphomolybdic acid molecules are transferred to gas molecules, so that the electron transfer between the carbon nanotubes and the phosphomolybdic acid molecules is enhanced, and the carbon nanotubes are transferred to the phosphomolybdic acid moleculesThe number of electrons of the molecule is increased by 0.9 electrons, so that the hole concentration of the carbon nano tube is increased, and the resistance of the network sensitive element is reduced.

(8) Utilizes molecular neural network pair based on cellular automata to simulate and analyze NO₂After modulation, the charge transfer between the basic neurons conforms to the linear separable electron transfer theory when NO is present₂During adsorption, the charge threshold a is determined according to the change ratio of the corresponding charge barrier height _TH4, the simulation results of pulse delivery are shown in FIG. 19, relative to adsorbed NO₂Comparing the previous pulse distribution results with fig. 18, it is apparent that the amplitude and the distribution frequency of the electric pulses in the molecular neural network are increased.

The invention aims to solve the technical problem of the prior art that the electric pulse behavior in the polyoxometallate-semiconductor nanowire/strip/tube composite network structure is not adjustable, and provides a method for controlling the charge transfer and transmission mechanism in the polyoxometallate-semiconductor nanowire/strip/tube composite network structure by using oxidation/reduction molecules to obtain the characteristics of the amplitude, the frequency, the occurrence probability and the like of the electric pulse, and a corresponding implementation method, so that the polyoxometallate-semiconductor nanowire/strip/tube composite structure neural network with the required specific electric pulse behavior is prepared. And a charge transmission mechanism of a single carbon nano tube/polyoxometallate is verified from theory to experiment in a mode of combining first-nature principle simulation with experiment, and the principle and the feasibility of realization of the regulation and control method are verified by utilizing the charge transmission mechanism and using a network simulation result and a part of experiment results based on a cellular automaton. The feasible electric pulse regulation and control method is provided for the network constructed based on the semiconductor nanowire/strip/tube-polyoxometallate, so that the learning mechanism and parameters of the network are controlled when the network is applied to artificial intelligence application, and meanwhile, the method has the characteristics of strong feasibility, high controllability and simple experiment; the invention also provides a ternary molecular neural network obtained based on the regulation and control method, wherein the ternary molecular neural network is built by a composite structure of the oxidation/reduction molecules, the polyoxometallate and the semiconductor low-dimensional nano material.

It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims

1. A method for regulating and controlling the electric pulse distribution behavior of a polyoxometallate-semiconductor low-dimensional nano material neural network by using oxidation/reduction molecules, the neural network is a molecular neural network built based on the polyoxometallate-semiconductor low-dimensional nano material composite structure, it is characterized in that the charge transfer between the polyoxometallate and the semiconductor low-dimensional nano material is changed by regulating and controlling the adsorption of oxidation/reduction molecules on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure, thereby changing the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure, thereby changing the charge potential barrier and the corresponding charge transmission mechanism between the polyoxometallate-semiconductor low-dimensional nano material composite structure units, thereby regulating and controlling the electric pulse issuing behavior of the polyoxometallate-semiconductor low-dimensional nano material neural network; the oxidation/reduction molecules are molecules capable of generating electron gain and loss with the polyoxometallate or the semiconductor low-dimensional nano material; the semiconductor low-dimensional nano material is a semiconductor one-dimensional nano material or a semiconductor two-dimensional nano material;

when the reducing molecules are adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, part of electrons of the reducing molecules are transferred into the polyoxometallate-semiconductor low-dimensional nano material composite structure unit, so that the hole concentration in the semiconductor low-dimensional nano material is reduced, the conductance of the polyoxometallate-semiconductor low-dimensional nano material composite structure is reduced, the charge potential barrier between the polyoxometallate-semiconductor low-dimensional nano material composite structure unit is increased, and the amplitude and the frequency of electric pulses emitted in the polyoxometallate-semiconductor low-dimensional nano material neural network are reduced;

2. The method of claim 1, wherein the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure is adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure by standing, pressurizing, heating or electrifying by immersing the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure in the oxidizing/reducing molecule environment.

3. The method of claim 1, wherein the polyoxometallate-semiconductor low-dimensional nanomaterial composite structure is placed in a closed container, after vacuum pumping, nitrogen or inert gas is introduced to normal pressure, and then the required type and concentration of the oxidation/reduction molecules are introduced, so that the oxidation/reduction molecules can be used as a control source for controlling the conductance, the charge barrier and the charge transport mechanism of the composite structure.

4. The method of claim 1, wherein the structure of the molecular neural network comprises, from bottom to top: the substrate, the thin film electrode and the semiconductor low-dimensional nano material modified by polyoxometallate molecules; the semiconductor low-dimensional nano material modified by polyoxometallate molecules is the polyoxometallate-semiconductor low-dimensional nano material composite structure;

5. The method of claim 1, wherein the semiconducting low dimensional nanomaterial is a semiconducting nanowire, semiconducting nanoribbon, or semiconducting nanotube;

the oxidation/reduction molecule is an oxidation/reduction gas molecule, an oxidation/reduction liquid molecule, or an oxidation/reduction solid molecule.

6. The ternary molecular neural network regulated and controlled by the method according to any one of claims 1 to 5, which is constructed on the basis of a composite structure of the oxidation/reduction molecule, the polyoxometallate and the semiconductor low-dimensional nano material.

7. The ternary molecular neural network of claim 6, wherein the structure of the ternary molecular neural network comprises, from bottom to top: the device comprises a substrate, a film electrode, a semiconductor low-dimensional nano material modified by polyoxometallate molecules and oxidation/reduction molecules adsorbed on the surface of the polyoxometallate-semiconductor low-dimensional nano material composite structure.

8. The ternary molecular neural network of claim 7, wherein said ternary molecular neural network is prepared by:

(3) by adopting a dielectrophoresis technology, dripping the mixed solution of the semiconductor low-dimensional nano material modified by polyoxometallate molecules between the film electrodes, and introducing alternating voltage to ensure that the polyoxometallate-semiconductor low-dimensional nano material composite structure is assembled between the film electrodes; then dropping an organic solvent in the middle of the electrode to remove the residual polyoxometallate solution, and obtaining a binary molecular neural network constructed on the basis of the polyoxometallate-semiconductor low-dimensional nano material composite structure; the organic solvent is an organic solvent which is easy to dissolve the semiconductor low-dimensional nano material and the polyoxometallate;

9. The ternary molecular neural network of claim 7, wherein said ternary molecular neural network is prepared by:

(3) performing membrane filtration on the mixed solution of the polyoxometallate molecule-modified semiconductor low-dimensional nano materials obtained in the step (2), covering one surface, containing the semiconductor low-dimensional nano materials/polyoxometallate, of the obtained filtration membrane on a multi-electrode array, and dissolving and removing the filtration membrane by using an organic solvent to obtain a binary molecular neural network constructed on the basis of the polyoxometallate-semiconductor low-dimensional nano material composite structure; the organic solvent is an organic solvent which is easy to dissolve the semiconductor low-dimensional nano material and the polyoxometallate;

10. The ternary molecular neural network of claim 7, wherein said molecular neural network is prepared by:

(3) dropping an organic solvent between the film electrodes to remove residual polyoxometallate solution, and obtaining a binary molecular neural network constructed on the basis of the polyoxometallate-semiconductor low-dimensional nano material composite structure; the organic solvent is an organic solvent which is easy to dissolve the semiconductor low-dimensional nano material and the polyoxometallate;