A method and system for predicting the performance of doped molybdenum disulfide

Abstract

The present application relates to the technical field of molybdenum disulfide performance prediction, and particularly relates to a method and system for predicting the performance of doped molybdenum disulfide, comprising: constructing a doping model library, obtaining the performance value of corresponding molybdenum disulfide based on the molybdenum disulfide initial model, extracting the characteristic parameters of each molybdenum disulfide initial model, and constructing a training data set; training a preset prediction model using the training data set to obtain a trained prediction model; inputting the characteristic parameters of the molybdenum disulfide to be predicted into the trained prediction model to obtain the performance prediction value of the molybdenum disulfide to be predicted. The present application realizes rapid and accurate evaluation of the performance of doped molybdenum disulfide by constructing a structural model library under multiple doping conditions and constructing a training data set combined with multi-dimensional characteristic parameters for learning and prediction.

Description

Technical Field

[0001] This invention relates to the field of molybdenum disulfide performance prediction technology, specifically to a method and system for predicting the performance of doped molybdenum disulfide. Background Technology

[0002] Molybdenum disulfide, a typical transition metal chalcogenide, has attracted widespread attention in electronic devices, optoelectronic devices, and energy catalysis due to its tunable bandgap, excellent photoelectric properties, and good catalytic activity. Studies have shown that introducing exogenous dopants can effectively modulate the electronic structure and carrier behavior of molybdenum disulfide, thereby improving its electrical properties, optical response, and catalytic activity.

[0003] In existing technologies, the properties of molybdenum disulfide doping are mainly obtained through experimental testing methods. However, experimental methods are costly and time-consuming, which is not conducive to the design of high-throughput materials.

[0004] With the development of data-driven materials research, machine learning methods are increasingly being applied to predict material properties. However, existing methods typically rely on limited samples or single structural features, resulting in limited prediction accuracy and generalization ability. Furthermore, data on doped molybdenum disulfide is relatively scarce, making it difficult to establish a stable and reliable training data foundation. Summary of the Invention

[0005] (a) Purpose of the invention The purpose of this invention is to provide a method and system for predicting the performance of doped molybdenum disulfide. By constructing a structural model library with multiple doping conditions and combining it with multidimensional feature parameters to build a training dataset, the invention enables rapid and accurate evaluation of the performance of doped molybdenum disulfide.

[0006] (II) Technical Solution To address the above problems, this invention provides a method for predicting the properties of doped molybdenum disulfide, comprising: Construct a doping model library, which includes initial molybdenum disulfide models with different doping elements, different doping concentrations, and different doping positions; Based on the initial models of molybdenum disulfide described above, the corresponding performance values ​​of molybdenum disulfide are obtained; The feature parameters of each of the initial molybdenum disulfide models are extracted, and the feature parameters include: atomic property features and structural features of the doped atoms; The feature parameters are used as input variables, and the corresponding molybdenum disulfide performance values ​​are used as output labels to construct a training dataset. The preset prediction model is trained using the training dataset to obtain the trained prediction model; The characteristic parameters of molybdenum disulfide to be predicted are input into the trained prediction model to obtain the predicted performance value of molybdenum disulfide.

[0007] In another aspect of the present invention, preferably, the method for constructing the doping model library includes: Establish a single-layer or multi-layer crystal structure of molybdenum disulfide, and use the crystal structure as a substrate structure; Set a list of atomic numbers of dopant elements and a target doping concentration range, wherein the atomic number list is used to identify the types of dopant elements to be introduced; Based on the substrate structure, the list of atomic numbers of the doping elements, and the target doping concentration range, the doping model library is obtained.

[0008] In another aspect of the present invention, preferably, obtaining the doping model library based on the substrate structure, the atomic number list of doping elements, and the target doping concentration range includes: Based on the atomic number list of the doped elements and the target doping concentration range, and based on the ratio between the number of doped atoms and the total number of atoms in the supercell, the supercell expansion factor that meets the concentration requirements is determined. Based on the supercell expansion factor, construct the corresponding basic supercell structure; Based on the aforementioned basic supercell structure, non-equivalent topological sites are determined through crystal symmetry analysis; Based on the substrate structure, molybdenum or sulfur atoms are replaced at the non-equivalent topological sites to generate a substitution-type doped structure model. The substitution-doped structure model was verified to obtain the doping model library.

[0009] In another aspect of the present invention, preferably, the determination of non-equivalent topological sites based on the fundamental supercell structure through crystal symmetry analysis includes: Obtain the crystal symmetry group information of the basic supercell structure; Based on the crystal symmetry group information, potential doping atom sites in the basic supercell structure are determined, including molybdenum atom sites and sulfur atom sites; The symmetry operation of the crystal symmetry group is applied to the potential doping atomic sites to determine the symmetry equivalence relationship between each doping atomic site, thereby obtaining multiple equivalent site groups; A representative site is determined from each of the equivalent site groups and identified as the non-equivalent topological site.

[0010] In another aspect of the present invention, preferably, the verification of the substitution-doped structure model to obtain the doping model library includes: Structural relaxation calculations were performed on the substitution-doped structure model to obtain structural optimization results; The structure optimization results are judged based on the preset convergence criteria, and substitution-doped structure models that have not reached the convergence condition are eliminated. Based on the preset structural stability criteria, the stability of substitution-doped structural models that have reached the convergence condition is evaluated, and structurally unstable substitution-doped structural models are eliminated. The validated substitution-doped structure models are compiled to obtain the doping model library.

[0011] In another aspect of the present invention, preferably, obtaining the corresponding performance values ​​of molybdenum disulfide based on each of the initial molybdenum disulfide models includes: Electronic structure calculations were performed on each of the initial molybdenum disulfide models to obtain the electronic structure calculation results. Based on the electronic structure calculation results, electronic structure features are extracted; The performance values ​​of the initial molybdenum disulfide model are calculated using the electronic structure characteristics based on the preset physical performance calculation formula.

[0012] In another aspect of the present invention, preferably, the step of performing electronic structure calculations on each of the initial molybdenum disulfide models to obtain electronic structure calculation results includes: Based on density functional theory, the self-consistent electronic structure of each of the initial molybdenum disulfide models is calculated to obtain the corresponding band structure, density of states distribution and Fermi level. Identify the positions of the valence band top and conduction band bottom in the band structure and record the band gap width; Based on the density of states distribution, analyze the electronic state occupancy and density of states near the Fermi level; The band structure, band gap width, density of states distribution, electronic state occupancy near the Fermi level, and density of states values ​​are the results of electronic structure calculations.

[0013] In another aspect of the present invention, preferably, the atomic property characteristics of the doped atom include: atomic radius, electronegativity, and number of valence electrons; The structural features include: doping concentration and doping location.

[0014] In another aspect of the present invention, preferably, training a preset prediction model using the training dataset to obtain a trained prediction model includes: The training dataset is divided into a training subset and a validation subset; Construct a prediction model based on a deep neural network, and initialize the structural parameters and weight parameters of the prediction model; The feature parameters in the training subset are input into the prediction model to obtain the model output. The loss function is calculated based on the difference between the model output and the corresponding performance value, and the weight parameters of the prediction model are iteratively updated using the backpropagation algorithm. During training, the model prediction error is evaluated using the validation subset. Training stops when the preset convergence condition is met, and the trained prediction model is obtained.

[0015] In another aspect of the present invention, preferably, a system for predicting the properties of molybdenum disulfide after doping includes: First construction module: Constructing a doping model library, which includes initial molybdenum disulfide models with different doping elements, different doping concentrations, and different doping positions; The acquisition module obtains the corresponding performance values ​​of molybdenum disulfide based on the initial models described above. Extraction module: Extracts feature parameters of each initial molybdenum disulfide model, including atomic property features and structural features of doped atoms; The second construction module: uses the feature parameters as input variables and the corresponding molybdenum disulfide performance values ​​as output labels to construct a training dataset; Training module: Uses the training dataset to train the preset prediction model to obtain the trained prediction model; Prediction module: Input the feature parameters of molybdenum disulfide to be predicted into the trained prediction model to obtain the performance prediction value of molybdenum disulfide to be predicted.

[0016] (III) Beneficial Effects The above-described technical solution of the present invention has the following beneficial technical effects: This invention systematically constructs a structural model library containing different doping elements, doping concentrations, and doping positions. By combining multi-dimensional information such as atomic property characteristics and structural characteristics, it establishes a mapping relationship between doped structures and material properties, constructs a training dataset, and performs learning predictions to achieve rapid prediction of the properties of doped molybdenum disulfide. This invention can significantly reduce computation time and computational resource consumption while ensuring prediction accuracy, thereby improving the efficiency of material performance evaluation. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall process of one embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0019] Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0020] In the description of this invention, it should be noted that the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0021] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0022] Example 1 A method for predicting the properties of molybdenum disulfide after doping. Figure 1 An overall flowchart of one embodiment of the present invention is shown, as follows: Figure 1 As shown, it includes: A doping model library is constructed, comprising initial models of molybdenum disulfide with different doping elements, doping concentrations, and doping positions. Based on the monolayer or multilayer crystal structure of molybdenum disulfide, 2×2, 3×3, or 4×4 supercell structures can be used as the basic structure for the computational model. Doping methods include substitutional doping, which can be divided into molybdenum substitution and sulfur substitution. Doping elements can be main group or transition metal elements, such as boron, carbon, nitrogen, oxygen, phosphorus, silicon, iron, cobalt, or nickel. Doping concentration is controlled by changing the supercell size or the number of doped atoms; for example, single-atom substitution corresponds to a doping concentration of approximately 3.125% to 12.5%. Doping positions include surface sites, internal sites, and sites with different lattice symmetries.

[0023] Furthermore, in this embodiment, the method for constructing the doping model library includes: Establish a single-layer or multi-layer crystal structure of molybdenum disulfide, and use the crystal structure as the substrate structure; specifically, it can be a hexagonal 2H-MoS2 crystal structure with space group P63 / mmc. The initial lattice constant can be obtained from experimental values ​​or database data.

[0024] A list of atomic numbers for the doping elements and a target doping concentration range are defined. The atomic number list identifies the types of doping elements to be introduced. The atomic number list specifies the types of elements to be introduced, such as atomic numbers 5 (boron), 6 (carbon), 7 (nitrogen), 8 (oxygen), 14 (silicon), 15 (phosphorus), 26 (iron), 27 (cobalt), 28 (nickel), etc. The target doping concentration range can be set from 1% to 25%, or further from 3% to 15%, depending on requirements. The doping concentration can be determined using the following formula:

[0025] Where N_d represents the number of doped atoms, and N_total represents the total number of atoms in the supercell.

[0026] Based on the substrate structure, the atomic number list of dopant elements, and the target doping concentration range, the doping model library is obtained. This systematic method for constructing the doping model library can cover a variety of doping elements and concentrations, ensuring the comprehensiveness and representativeness of the model library. Specifically, it includes: Based on the atomic number list of the doping elements and the target doping concentration range, and considering the ratio between the number of doped atoms and the total number of atoms in the supercell, the supercell expansion factor that meets the concentration requirements is determined. Using the original MoS2 unit cell as the unit, a supercell structure is constructed by expanding n×n along the a and b directions. If a single original unit cell contains 1 molybdenum atom and 2 sulfur atoms, then the n×n supercell contains n atoms. 2 molybdenum atoms and 2n 2 One sulfur atom. The required value of n is determined by working backward from the target doping concentration and the number of doped atoms. For example, when the target concentration is 6.25% and single-atom substitution is used, a 4×4 supercell structure can be selected.

[0027] Based on the supercell expansion factor, the corresponding basic supercell structure is constructed; during the construction process, lattice symmetry and periodic boundary conditions are maintained to ensure that the supercell structure has no atomic overlap or unreasonable bond lengths.

[0028] Based on the aforementioned basic supercell structure, non-equivalent topological sites are determined through crystal symmetry analysis; through symmetry operation analysis, atoms that are equivalent under symmetry transformations in the crystal are classified, and symmetrically inequivalent molybdenum and sulfur sites are screened out.

[0029] Based on the substrate structure, molybdenum or sulfur atoms are replaced at the non-equivalent topological sites to generate a substitution-type doped structure model. According to the set dopant element type, Mo or S atoms at selected positions are replaced with the corresponding dopant atoms to generate a substitution-type doped structure model. For different doping concentrations, multiple non-equivalent sites can be combined for substitution to form a multi-atom synergistic doping structure. The periodic boundary conditions of the crystal remain unchanged during the substitution process. By using a method based on supercell ratio and crystal topology to generate doping models, precise control of doping concentration and diversity of doping sites are ensured, as well as diversity of training data, enhancing the applicability of the prediction model.

[0030] The substitution-doped structure model is validated to obtain the doping model library. The purpose of validation is to screen a set of doped structures with stable structures, controllable concentrations, and clearly defined sites, thus forming the doping model library.

[0031] Furthermore, in this embodiment, based on the fundamental supercell structure, non-equivalent topological sites are determined through crystal symmetry analysis, including: Obtain the crystal symmetry group information of the basic supercell structure; after completing the supercell construction and preliminary geometric optimization, identify the lattice parameters, atomic fraction coordinates and atomic type information of the supercell structure.

[0032] Based on the crystal symmetry group information, potential doping atomic sites in the basic supercell structure are determined. These potential doping atomic sites include molybdenum (Mo) atomic sites and sulfur (S) atomic sites. Fractional coordinates of all Mo and S atomic sites in the supercell are extracted to form sets of Mo and S sites, respectively. All of these sites are considered potential doping substitution sites. For both monolayer and multilayer structures, it is necessary to consider whether atoms in different layers remain equivalent under symmetry operations to avoid misjudgment.

[0033] The symmetry operations of the crystal symmetry group are applied to the potential doped atomic sites to determine the symmetry equivalence relationships between the doped atomic sites, thereby obtaining multiple sets of equivalent sites; the symmetry operations are expressed by the following formula:

[0034] Where r is the original coordinate of the atom, R is the rotation matrix, and t is the translation vector. The coordinates are compared one by one with the original coordinates of all potential doped atom sites to determine if they coincide within a given tolerance range. The tolerance can be set to 10. -3 ~10 -4 (Fractional coordinate units) are used to avoid numerical errors affecting the judgment. If two sites can be mapped to each other under at least one symmetry operation, they are determined to be symmetric equivalent sites. By traversing all sites and all symmetry operations, a complete equivalence relation matrix can be established. Based on the equivalence relation matrix, all mutually equivalent sites are divided into several equivalence site groups, and the sites within each group are completely equivalent in a symmetric sense.

[0035] A representative site is determined from each group of equivalent sites, and this site is designated as the non-equivalent topological site. The representative site can be selected according to the following rules: the site with the smallest fractional coordinate value or the site with the smallest number in the original atom numbering. This can compress the originally numerous potential doping sites into several symmetrically inequivalent representative sites, i.e., the non-equivalent topological sites. Using crystal symmetry analysis to determine the non-equivalent topological sites avoids redundant calculations of equivalent sites, saves computational resources, and ensures the representativeness and structural rationality of the doping model library.

[0036] Furthermore, in this embodiment, the verification of the substitution-doped structure model to obtain the doping model library includes: Structural relaxation calculations are performed on the substitution-doped structure models to obtain structural optimization results. For each generated substitution-doped structure model, a first-principles calculation method based on density functional theory is used to optimize the geometric structure. Through continuous iteration of self-consistent field calculations, the relaxed structural optimization results are finally obtained, including the optimized lattice parameters, atomic coordinates, and total system energy.

[0037] The structure optimization results are judged based on a preset convergence criterion, and substitution-doped structure models that have not met the convergence condition are eliminated; the convergence criterion includes: total energy convergence accuracy less than 1×10⁻⁶. -5 eV; the force on each atom is less than 0.01 eV / Å; the total stress of the system is less than 0.02 GPa; the iteration is completed within the preset maximum number of steps. If a structure still cannot meet any of the above convergence conditions within the maximum number of iterations, it is determined to be a non-convergent structure and is removed from the candidate model.

[0038] Based on a preset structural stability criterion, the stability of substitution-doped structural models that have reached convergence is evaluated, and structurally unstable substitution-doped structural models are eliminated. The stability evaluation includes checking for abnormal bond lengths or bond angles. For example, the distance between any two atoms must not be less than 80% of the sum of their corresponding covalent radii; the lattice distortion rate must not exceed a preset threshold (e.g., a relative change exceeding 10%); and there must be no cases of severe atomic overlap or significant detachment from the lattice plane.

[0039] The validated substitution-doped structure models are compiled to obtain the doping model library. By verifying the structural relaxation and stability of the substitution-doped structure models, unstable structures are eliminated, ensuring the physical rationality of each model in the doping model library and improving the reliability of the prediction results.

[0040] Based on the initial models of molybdenum disulfide described above, the corresponding performance values ​​of molybdenum disulfide are obtained, including: Electronic structure calculations were performed on each of the initial molybdenum disulfide models to obtain the electronic structure calculation results, including: Self-consistent electronic structure calculations were performed on the initial molybdenum disulfide models based on density functional theory (DFT) to obtain the corresponding band structure, density of states distribution, and Fermi level. The plane-wave pseudopotential method based on DFT can be used for calculation. Before starting the calculation, optimized lattice parameters and atomic fraction coordinates were input, and periodic boundary conditions were set. After the self-consistent calculation was completed, the band structure, density of states distribution, and Fermi level were obtained.

[0041] The positions of the valence band top and conduction band bottom in the band structure are identified, and the band gap width is recorded. The band gap width is calculated by the difference between the energy at the conduction band bottom and the energy at the valence band top. If the conduction band bottom and the valence band top are located at the same k-point, it is determined to be a direct band gap; if they are located at different k-points, it is determined to be an indirect band gap, and the band gap type information is recorded.

[0042] Based on the density of states distribution, the occupancy of electronic states and the density of states value near the Fermi level are analyzed, including whether there are localized states introduced by doping, whether impurity levels have entered the band gap, and whether the Fermi level has undergone a significant shift. For metallic systems, the density of states at the Fermi level is not zero; for semiconductor systems, the Fermi level is located in the band gap and the density of states is close to zero.

[0043] The band structure, band gap width, density of states distribution, electronic state occupancy near the Fermi level, and density of states values ​​are the results of electronic structure calculations. Based on the electronic structure calculation results, electronic structure features are extracted; the extracted electronic structure features may vary depending on the research objectives.

[0044] The performance values ​​of the initial molybdenum disulfide model are calculated using the pre-defined physical property calculation formula and the electronic structure characteristics. These performance values ​​may vary depending on the research objective.

[0045] For example, electrical properties, such as carrier mobility, are calculated using the following physical property calculation formula:

[0046] Where μ represents carrier mobility, e represents electron charge, τ represents relaxation time (which can be taken as an empirical value or estimated based on the theory of variable conditions), and m * For effective quality.

[0047] Electrical conductivity is calculated using the following physical property formula:

[0048] Where σ represents electrical conductivity, n represents carrier concentration, which can be approximately estimated by the Fermi level position and density of states integral, e represents electron charge, and μ represents carrier mobility.

[0049] Optical absorption capability is determined by the absorption edge position based on the band gap width Eg. The smaller Eg is, the wider the visible light response range.

[0050] Feature parameters are extracted from the initial models of molybdenum disulfide, including atomic and structural characteristics of the dopant atoms. Atomic characteristics include atomic radius (either covalent or effective atomic radius, in angstroms). This data can be obtained from a standard element database. Electronegativity is represented using the Paul electronegativity scale to characterize the strength of the dopant atom's ability to attract electrons. The number of valence electrons is defined as the number of outermost electrons or the number of valence electrons involved in bonding. For transition metals, this can be defined according to their common valence states or the number of d-orbital electrons. Structural characteristics include doping concentration and doping position, where the doping position distinguishes between substituted molybdenum and substituted sulfur sites. Numerical encoding can be used, for example, substituted molybdenum sites are denoted as 1, and substituted sulfur sites as 0. Combining the basic atomic properties (atomic radius, electronegativity, number of valence electrons) of the dopant atoms with structural characteristics such as doping concentration and position provides comprehensive feature input for the prediction model, improving its sensitivity to performance differences and prediction accuracy.

[0051] The feature parameters are used as input variables, and the corresponding molybdenum disulfide performance values ​​are used as output labels to construct a training dataset. The training dataset is used to train a preset prediction model to obtain a trained prediction model, including: The training dataset is divided into a training subset and a validation subset; the first 0.8N samples can be taken as the training subset and the remaining 0.2N samples as the validation subset.

[0052] A prediction model based on a deep neural network is constructed. This model includes an input layer, several hidden layers, and an output layer. The number of neurons in the input layer is the same as the dimension of the feature parameters. There can be 2 to 5 hidden layers, with the number of neurons in each layer selected based on the feature complexity. The ReLU activation function is used. The output layer is used to predict performance values, and its activation function is linear. The structural parameters and weight parameters of the prediction model are initialized. Weight parameters can be initialized using Xavier or He initialization methods to avoid gradient vanishing or exploding problems. Bias parameters can be initialized to zero or small random values.

[0053] The feature parameters in the training subset are input into the prediction model to obtain the model output. The loss function is calculated based on the difference between the model output and the corresponding performance value. The weight parameters of the prediction model are iteratively updated using the backpropagation algorithm. The loss function can be the mean squared error.

[0054] During training, the model prediction error is evaluated using the validation subset. Training stops when a preset convergence condition is met, such as when the error stops decreasing after several consecutive convergences or when a preset threshold is reached, and the trained prediction model is obtained.

[0055] The characteristic parameters of molybdenum disulfide to be predicted are input into the trained prediction model to obtain the predicted performance value of molybdenum disulfide.

[0056] This embodiment systematically constructs a structural model library containing different doping elements, doping concentrations, and doping locations. Combining multi-dimensional information such as atomic property characteristics and structural characteristics, it establishes a mapping relationship between doped structures and material properties, constructs a training dataset, and performs learning predictions to achieve rapid prediction of the properties of doped molybdenum disulfide. This avoids the high computational load and time-consuming methods of traditional experiments or density functional theory, and can significantly reduce computation time and computational resource consumption while ensuring prediction accuracy, thereby improving the efficiency of material performance evaluation.

[0057] Example 2 A system for predicting the properties of molybdenum disulfide after doping, comprising: First construction module: Constructing a doping model library, which includes initial molybdenum disulfide models with different doping elements, different doping concentrations, and different doping positions; The acquisition module obtains the corresponding performance values ​​of molybdenum disulfide based on the initial models described above. Extraction module: Extracts feature parameters of each initial molybdenum disulfide model, including atomic property features and structural features of doped atoms; The second construction module: uses the feature parameters as input variables and the corresponding molybdenum disulfide performance values ​​as output labels to construct a training dataset; Training module: Uses the training dataset to train the preset prediction model to obtain the trained prediction model; Prediction module: Input the feature parameters of molybdenum disulfide to be predicted into the trained prediction model to obtain the performance prediction value of molybdenum disulfide to be predicted.

[0058] It should be understood that the specific embodiments described above are merely illustrative or explanatory of the principles of the invention and do not constitute a limitation thereof. Therefore, any modifications, equivalent substitutions, improvements, etc., made without departing from the spirit and scope of the invention should be included within the protection scope of the invention. Furthermore, the appended claims are intended to cover all variations and modifications falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

[0059] The present invention has been described above with reference to embodiments thereof. However, these embodiments are merely illustrative and not intended to limit the scope of the invention. The scope of the invention is defined by the appended claims and their equivalents. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the invention, and all such substitutions and modifications should fall within the scope of the invention.

[0060] Although embodiments of the present invention have been described in detail, it should be understood that various changes, substitutions, and modifications can be made to the embodiments of the present invention without departing from the spirit and scope of the invention.

[0061] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for predicting the properties of molybdenum disulfide after doping, characterized in that, include: Construct a doping model library, which includes initial molybdenum disulfide models with different doping elements, different doping concentrations, and different doping positions; Based on the initial models of molybdenum disulfide described above, the corresponding performance values ​​of molybdenum disulfide are obtained; The feature parameters of each of the initial molybdenum disulfide models are extracted, and the feature parameters include: atomic property features and structural features of the doped atoms; The feature parameters are used as input variables, and the corresponding molybdenum disulfide performance values ​​are used as output labels to construct a training dataset. The preset prediction model is trained using the training dataset to obtain the trained prediction model; The characteristic parameters of molybdenum disulfide to be predicted are input into the trained prediction model to obtain the predicted performance value of molybdenum disulfide.

2. The method for predicting the properties of doped molybdenum disulfide according to claim 1, characterized in that, The method for constructing the doping model library includes: Establish a single-layer or multi-layer crystal structure of molybdenum disulfide, and use the crystal structure as a substrate structure; Set a list of atomic numbers of dopant elements and a target doping concentration range, wherein the atomic number list is used to identify the types of dopant elements to be introduced; Based on the substrate structure, the list of atomic numbers of the doping elements, and the target doping concentration range, the doping model library is obtained.

3. The method for predicting the properties of doped molybdenum disulfide according to claim 2, characterized in that, The doping model library, derived based on the substrate structure, the atomic number list of doped elements, and the target doping concentration range, includes: Based on the atomic number list of the doped elements and the target doping concentration range, and based on the ratio between the number of doped atoms and the total number of atoms in the supercell, the supercell expansion factor that meets the concentration requirements is determined. Based on the supercell expansion factor, construct the corresponding basic supercell structure; Based on the aforementioned basic supercell structure, non-equivalent topological sites are determined through crystal symmetry analysis; Based on the substrate structure, molybdenum or sulfur atoms are replaced at the non-equivalent topological sites to generate a substitution-type doped structure model. The substitution-doped structure model was verified to obtain the doping model library.

4. The method for predicting the properties of doped molybdenum disulfide according to claim 3, characterized in that, The determination of non-equivalent topological sites based on the aforementioned fundamental supercell structure through crystal symmetry analysis includes: Obtain the crystal symmetry group information of the basic supercell structure; Based on the crystal symmetry group information, potential doping atom sites in the basic supercell structure are determined, including molybdenum atom sites and sulfur atom sites; The symmetry operation of the crystal symmetry group is applied to the potential doping atomic sites to determine the symmetry equivalence relationship between each doping atomic site, thereby obtaining multiple equivalent site groups; A representative site is determined from each of the equivalent site groups and identified as the non-equivalent topological site.

5. The method for predicting the properties of doped molybdenum disulfide according to claim 3, characterized in that, The process of validating the substitution-doped structure model to obtain the doping model library includes: Structural relaxation calculations were performed on the substitution-doped structure model to obtain structural optimization results; The structure optimization results are judged based on the preset convergence criteria, and substitution-doped structure models that have not reached the convergence condition are eliminated. Based on the preset structural stability criteria, the stability of substitution-doped structural models that have reached the convergence condition is evaluated, and structurally unstable substitution-doped structural models are eliminated. The validated substitution-doped structure models are compiled to obtain the doping model library.

6. The method for predicting the properties of doped molybdenum disulfide according to claim 5, characterized in that, The process of obtaining the corresponding performance values ​​of molybdenum disulfide based on each of the initial models includes: Electronic structure calculations were performed on each of the initial molybdenum disulfide models to obtain the electronic structure calculation results. Based on the electronic structure calculation results, electronic structure features are extracted; The performance values ​​of the initial molybdenum disulfide model are calculated using the electronic structure characteristics based on the preset physical performance calculation formula.

7. The method for predicting the properties of doped molybdenum disulfide according to claim 6, characterized in that, The electronic structure calculation of each of the initial molybdenum disulfide models, to obtain the electronic structure calculation results, includes: Based on density functional theory, the self-consistent electronic structure of each of the initial molybdenum disulfide models is calculated to obtain the corresponding band structure, density of states distribution and Fermi level. Identify the positions of the valence band top and conduction band bottom in the band structure and record the band gap width; Based on the density of states distribution, analyze the electronic state occupancy and density of states near the Fermi level; The band structure, band gap width, density of states distribution, electronic state occupancy near the Fermi level, and density of states values ​​are the results of electronic structure calculations.

8. The method for predicting the properties of molybdenum disulfide after doping according to claim 1, characterized in that, The atomic properties of the doped atoms include: atomic radius, electronegativity, and number of valence electrons; The structural features include: doping concentration and doping location.

9. The method for predicting the properties of doped molybdenum disulfide according to claim 1, characterized in that, The step of training a preset prediction model using the training dataset to obtain a trained prediction model includes: The training dataset is divided into a training subset and a validation subset; Construct a prediction model based on a deep neural network, and initialize the structural parameters and weight parameters of the prediction model; The feature parameters in the training subset are input into the prediction model to obtain the model output. The loss function is calculated based on the difference between the model output and the corresponding performance value, and the weight parameters of the prediction model are iteratively updated using the backpropagation algorithm. During training, the model prediction error is evaluated using the validation subset. Training stops when the preset convergence condition is met, and the trained prediction model is obtained.

10. A system for predicting the properties of molybdenum disulfide after doping, characterized in that, include: First construction module: Constructing a doping model library, which includes initial molybdenum disulfide models with different doping elements, different doping concentrations, and different doping positions; The acquisition module obtains the corresponding performance values ​​of molybdenum disulfide based on the initial models described above. Extraction module: Extracts feature parameters of each initial molybdenum disulfide model, including atomic property features and structural features of doped atoms; The second construction module: uses the feature parameters as input variables and the corresponding molybdenum disulfide performance values ​​as output labels to construct a training dataset; Training module: Uses the training dataset to train the preset prediction model to obtain the trained prediction model; Prediction module: Input the feature parameters of molybdenum disulfide to be predicted into the trained prediction model to obtain the performance prediction value of molybdenum disulfide to be predicted.

Images