Photoelectric material synthesis method based on physical model and machine learning model

Abstract

The invention discloses a photoelectric material synthesis method based on a physical model and a machine learning model, and the method comprises the following steps: determining the type of photoelectric data, and constructing a physical model; photoelectric data types are screened through a physical model, and a photoelectric database is constructed; and constructing a machine learning model, and guiding the synthesis of the photoelectric material through the machine learning model. According to the scheme, data screening is carried out by using the physical model, the difficulty of photoelectric database construction is reduced, and the fitting degree of the machine learning model when the data volume is small is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of optoelectronic material synthesis, and specifically relates to a method for synthesizing optoelectronic materials based on a physical model and a machine learning model. Background Art

[0002] Machine learning is a widely used big data processing technology. In recent years, it has been applied in the fields of materials, biomedicine, etc. Its main feature is to use a large amount of data to perform comprehensive data processing to achieve data prediction, rule summary and other functions. However, the data set of machine learning has the characteristics of large data volume, which is difficult to achieve in the field of materials. However, if the amount of data is too small, there will be distortion in the analysis process. Therefore, a method is needed to use less data in the field of materials to achieve distortion-free neural network synthesis of optoelectronic materials. Summary of the invention

[0003] The main purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and to provide a method for synthesizing optoelectronic materials based on physical models and machine learning models.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions: A method for synthesizing optoelectronic materials based on a physical model and a machine learning model, the method comprising the following steps: S1. Determine the data types of optoelectronic materials, including the types of optoelectronic material data, optoelectronic material synthesis formula parameters, optoelectronic material optical and electrical parameters, and optical and electrical parameters of optical devices constructed with optoelectronic materials; S2. Use COMSOL software and optoelectronic data to perform multi-physics simulation and build a physical model; S3, through the physical model screening to eliminate redundant data types, reduce the data size, and build an optoelectronic database; specifically, the physical model constructed in step S2 is used to eliminate the optoelectronic material synthesis formula parameters, optoelectronic material optical and electrical parameters with low correlation for the optoelectronic device performance parameters. S4. Build a machine learning model; S5. Predict optoelectronic material synthesis through machine learning models.

[0005] Furthermore, the types of photoelectric material data include the name of the material itself, the material concentration, and the material ratio; The optical and electrical parameters of the photoelectric material refer to the transmittance, absorption, bandgap width and conductivity of the material; The optical and electrical parameters of the optical device constructed with the optoelectronic material include the transmittance, response time, threshold voltage and life of the device.

[0006] Furthermore, the process of step S2 is as follows: S201, COMSOL software builds a 2D or 3D simulation model of the device according to the device structural parameters. The optoelectronic device structural parameters include boundary length, device layer width, and number of device layers; S202, assigning material parameters to the 2D or 3D simulation model, where the material parameters include material transmittance, absorption, bandgap width, and conductivity; S203, selecting the physical fields involved in the device according to the working principle of the device, including diffusion and drift motion of charged particles, particle recombination at the interface, and optical-electrical conversion; S204. Select the physical formula involved according to the selected physical field.

[0007] Among them, the physical formula includes the electric flux formula: Total flux N i = ∇(−D i ∇c i −z i μ i Fc i ∇ϕ i ) + uc i In the formula, 𝑐 𝑖 represents the concentration of the substance, 𝐷 𝑖 is the diffusion coefficient, 𝜙 𝑖 is the electric potential, F is the Faraday constant, 𝑧 𝑖 is the charge number, μ i is the electromigration number, u represents the fluid flow rate, and ∇ represents the gradient operation symbol; The physical formula also includes the Butler–Volmer formula: j loc =j0{exp (-nαFη / RT)-exp[n(1-α)Fη} Among them, 𝑗 l𝑜𝑐 and 𝑗0 are the local current density and exchange current density respectively, α is the transfer coefficient, n is the number of electrons transferred, R is the gas constant, T is the reaction temperature, and η is the overpotential.

[0008] Furthermore, in step S3, the physical model is used to exclude the optoelectronic material synthesis formula parameters, optoelectronic material optical and electrical parameters with low correlation with the optoelectronic device performance parameters.

[0009] Furthermore, the construction of the database in step S3 is actually to collect the synthesis formula parameters of the optoelectronic materials, and the optical and electrical parameters of the optoelectronic materials screened by the above physical model, and the process is as follows: S301. Normalize all selected data, that is, eliminate the dimension and convert it into pure numbers. The normalization method is: is the normalized data, is the original data, is the minimum value of the original data of this type. It is the maximum value of the original data of this type; S302, the selected data is composed of a group of data including material synthesis formula parameters, optical and electrical parameters of optoelectronic materials, and optoelectronic device parameters. The specific number of data in the group is specified in advance and is determined according to the scale of the database. All group data constitute a data set; S303. The selected data set is divided into K subsets (K>5), one of which is used as a test set, and the rest are used as training sets.

[0010] Furthermore, the machine learning models in step S4 include 7 machine learning models, namely linear regression (LR), Lasso regression (Lasso), kernel ridge regression (KRR), Gaussian process regression (GPR), support vector regression (SVR), k-nearest neighbor regression (KNR), and decision tree (DTR) regression. The machine learning models are directly imported from the python database. The process of step S4 is as follows: S401, import the test set selected in step S3 into the machine learning model, select the number of iterations N (N>10), and output the performance evaluation index; wherein the selected performance evaluation index includes the root mean square error RMSE and the determination coefficient R 2 , as follows; R 2 Represents the square of the correlation coefficient between the predicted value and the true value. The closer the value is to 1, the better the fitting effect of the machine learning model is. The calculation expression is: in, is the true value of the i-th sample, is the predicted value of the ith sample, is the mean of the true value of the sample, m is the total number of samples, Var is the variance of the true value of the sample, and the root mean square error RMSE is the mean square root of the square difference between the predicted value and the true value. The smaller the value, the better the fitting effect of the machine learning model. The calculation expression is: S402, repeat the cross validation 7 times, divide the data set into K subsets, use one of the subsets as the test set each time, and the rest as the training set, and use the test set to evaluate the model performance after each training. Repeat 7 times to obtain the performance evaluation indicators of 7 machine learning models, and then take the average value as the final evaluation result.

[0011] Furthermore, the process of step S5 is as follows: S501, selecting the machine learning model with the best fitting effect as the prediction model to predict the synthesis of optoelectronic materials; S502. Change the input data type of the prediction model, wherein the change of input data refers to any one or more of the synthesis formula parameters of optoelectronic materials, the optical and electrical parameters of optoelectronic materials, and the parameters of optoelectronic devices, and the output data is the remaining one or more of the optoelectronic device parameters, that is, any one or more of the optoelectronic device parameters are used as input, and the output of the machine learning model predicts the synthesis of optoelectronic materials, which is the optical and electrical parameters of the optical devices excluding the input, and the number of other optical and electrical parameters of the optical devices is the predicted output, and the optical and electrical parameters of the optical devices are obtained by the input optical and electrical parameters of the optical devices and the predicted output, thus completing the output of the synthesis of the optical devices, and by obtaining fixed optical and electrical parameters of the optical devices, the synthesis prediction of the optoelectronic materials is obtained.

[0012] Compared with the prior art, the present invention has the following advantages and beneficial effects: (1) The present invention discloses a method for synthesizing optoelectronic materials based on a physical model and a machine learning model. The technical solution designed is to first screen the data types using a physical model and then construct a database for machine learning. The physical model is used to screen the data types, which reduces the dimension of the database and thus reduces the amount of data required for machine learning.

[0013] (2) The present invention discloses a method for synthesizing optoelectronic materials based on a physical model and a machine learning model. The technical solution is designed to first screen the data types using a physical model and then construct a database for machine learning. The physical model is used to establish the correlation between data, thereby increasing the fitting degree of the machine learning model under a smaller database.

[0014] (3) The present invention discloses a method for synthesizing optoelectronic materials based on a physical model and a machine learning model, i.e., the steps of screening data types using a physical model, establishing a small database, and then performing machine learning can be applied to data prediction in other material fields and has good generalization. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 It is a flow chart of a method for synthesizing optoelectronic materials based on a physical model and a machine learning model disclosed in Example 1 of the present invention; Figure 2is a schematic diagram of a physical model of a liquid optoelectronic device disclosed in an embodiment of the present invention; Figure 3 It is a parameter diagram between the formula and performance of ammonium metatungstate-ferrous chloride in the embodiment of the present invention; Figure 4 It is a parameter diagram between the formula and performance of ammonium metatungstate-ferrous sulfate in the embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0018] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0019] Example 1 like Figure 1 As shown, this embodiment discloses a method for synthesizing optoelectronic materials based on a physical model and a machine learning model. Figure 1 The photoelectric material synthesis method comprises the following steps: S1. Determine the data types of optoelectronic materials, including the types of optoelectronic material data, optoelectronic material synthesis formula parameters, optoelectronic material optical and electrical parameters, and optical and electrical parameters of optical devices constructed from optoelectronic materials; where optoelectronic material data types include the name of the material itself, material concentration, and material ratio; the optical and electrical parameters of optoelectronic materials refer to the material's transmittance, absorption, bandgap, and conductivity; the optical and electrical parameters of optical devices constructed from optoelectronic materials include the device's transmittance, response time, threshold voltage, and lifespan. S2, using COMSOL software and optoelectronic data to perform multi-physics field simulation and build a physical model; in this embodiment, the process of step S2 is as follows: S201, COMSOL software builds a 2D or 3D simulation model of the device according to the device structural parameters. The optoelectronic device structural parameters include boundary length, device layer width, and number of device layers; S202, assigning material parameters to the 2D or 3D simulation model, where the material parameters include material transmittance, absorption, bandgap width, and conductivity; S203, selecting the physical fields involved in the device according to the working principle of the device, including diffusion and drift motion of charged particles, particle recombination at the interface, and optical-electrical conversion; S204. Select the physical formula involved according to the selected physical field.

[0020] In this embodiment, COMSOL software is used to construct a physical model of the liquid optoelectronic device. Figure 2 , a cathode potential is applied to boundary 1; an anode potential is applied to boundary 3; boundaries 4, 5, 6, and 7 are electrically insulated; and the lithium flux at boundaries 1, 3, 4, 5, 6, and 7 is 0.

[0021] S3. Eliminate redundant data types through physical model screening, reduce the data size, and build an optoelectronic database; specifically, use the physical model constructed in step S2 to eliminate optoelectronic material synthesis formula parameters, optoelectronic material optical and electrical parameters with low correlation for optoelectronic device performance parameters.

[0022] In this embodiment, simulation variables and parameters are input, as shown in Table 1, and electrical-optical multi-physical field coupling is performed after selecting the electrical physical field formula; the data group type selected is (ammonium metatungstate concentration, ferrocene concentration, modulation amplitude).

[0023] Table 1. Simulation variables and parameters S4, construct a machine learning model; the machine learning model includes 7 machine learning models, including linear regression (LR), Lasso regression (Lasso), kernel ridge regression (KRR), Gaussian process regression (GPR), support vector regression (SVR), k-nearest neighbor regression (KNR), and decision tree (DTR) regression. The machine learning model is directly imported from the python database; the process of step S4 in this embodiment is as follows: S401, import the test set selected in step S3 into the machine learning model, select the number of iterations N (N>10), and output the performance evaluation index; wherein the selected performance evaluation index includes the root mean square error RMSE and the determination coefficient R 2 , as follows; R 2 Represents the square of the correlation coefficient between the predicted value and the true value. The closer the value is to 1, the better the fitting effect of the machine learning model is. The calculation expression is: in, is the true value of the i-th sample, is the predicted value of the ith sample, is the mean of the true value of the sample, m is the total number of samples, Var is the variance of the true value of the sample, and the root mean square error RMSE is the mean square root of the square difference between the predicted value and the true value. The smaller the value, the better the fitting effect of the machine learning model. The calculation expression is: S402, repeat the cross validation 7 times, divide the data set into K subsets, use one of the subsets as the test set each time, and the rest as the training set, and use the test set to evaluate the model performance after each training. Repeat 7 times to obtain the performance evaluation indicators of 7 machine learning models, and then take the average value as the final evaluation result.

[0024] In this embodiment, the collected data were normalized and then a database was constructed. The database was divided into 10 groups. Seven machine learning models including linear regression (LR), Lasso regression (Lasso), kernel ridge regression (KRR), Gaussian process regression (GPR), support vector regression (SVR), k-nearest neighbor regression (KNR), and decision tree (DTR) regression were selected. After importing the data, the determination coefficients R of different models were obtained by cross-validation. 2 See Table 2 for values.

[0025] Table 2. Determination coefficient R 2 value S5. Predict optoelectronic material synthesis through machine learning models.

[0026] In this embodiment, the process of step S5 is as follows: S501, selecting the machine learning model with the best fitting effect as the prediction model to predict the synthesis of optoelectronic materials; S502. Change the type of input data of the prediction model, wherein the change of input data refers to any one or more of the synthesis formula parameters of optoelectronic materials, optical and electrical parameters of optoelectronic materials, and optoelectronic device parameters, and the output data is the remaining one or more of the optoelectronic device parameters.

[0027] In this embodiment, the SVR model is finally selected for data prediction with an accuracy of up to 99.35%.

[0028] Comparative Example 1 Referring to steps S1 to S5 of a method for synthesizing optoelectronic materials based on a physical model and a machine learning model disclosed in Example 1, this comparative example continues to disclose a method for synthesizing optoelectronic materials based on a physical model and a machine learning model.

[0029] Among them, the data group type is (ammonium metatungstate concentration, ferrocene concentration, device thickness, solution transmittance, modulation amplitude). After collecting and normalizing the data, a database was constructed, and the database was divided into 10 groups. A total of 7 machine learning models were selected, including linear regression (LR), Lasso regression (Lasso), kernel ridge regression (KRR), Gaussian process regression (GPR), support vector regression (SVR), k-nearest neighbor regression (KNR), and decision tree (DTR) regression. After importing the data, cross-validation was performed to obtain the determination coefficient R of different models. 2 The LR model was finally selected for data prediction with an accuracy of 39.25%.

[0030] Example 2 Referring to the various steps of a method for synthesizing optoelectronic materials based on a physical model and a machine learning model in Example 1, this embodiment specifically discloses a method for synthesizing two liquid optoelectronic devices, ammonium metatungstate-ferrous chloride and ammonium metatungstate-ferrous sulfate, respectively, and compares the parameter changes between the formula and performance of the two liquid optoelectronic devices, ammonium metatungstate-ferrous chloride and ammonium metatungstate-ferrous sulfate.

[0031] In this embodiment, each device is divided into four ammonium metatungstate concentration gradients and five ferrous chloride (or ferrous sulfate) concentration gradients, and a total of 20 devices are made in each group. Subsequently, the transmittance changes of each device at different current densities are measured to obtain their respective light modulation amplitudes, coloring time and fading time. After that, the measured data is used as a training set, and a total of 7 different machine learning models including linear regression, K-nearest neighbor regression, decision tree, logistic forest regression, kernel ridge regression, Gaussian process regression and support vector regression are used to construct photoelectric models of two liquid optoelectronic devices: ammonium metatungstate-ferrous chloride and ammonium metatungstate-ferrous sulfate, and predict the solution composition and concentration when the device performance is optimal, and finally form a mutual conversion model between formula, solution concentration, light modulation amplitude, coloring time and fading time. The parameter diagram between the final formula and performance is shown in Figure 3 and Figure 4 shown.

[0032] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0033] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for synthesizing optoelectronic materials based on a physical model and a machine learning model, characterized in that: The photoelectric material synthesis method comprises the following steps: S1. Determine the data types of optoelectronic materials, including the types of optoelectronic material data, optoelectronic material synthesis formula parameters, optoelectronic material optical and electrical parameters, and optical and electrical parameters of optical devices constructed with optoelectronic materials; S2. Use COMSOL software and optoelectronic data to perform multi-physics simulation and build a physical model; S3, through physical model screening to eliminate redundant data types, reduce the data scale, and build an optoelectronic database; S4. Build a machine learning model; S5. Predict optoelectronic material synthesis through machine learning models.

2. The method for synthesizing optoelectronic materials based on a physical model and a machine learning model according to claim 1, characterized in that: The types of photoelectric material data include the name of the material itself, the material concentration, and the material ratio; The optical and electrical parameters of the photoelectric material refer to the transmittance, absorption, bandgap width and conductivity of the material; The optical and electrical parameters of the optical device constructed with the optoelectronic material include the transmittance, response time, threshold voltage and life of the device.

3. The method for synthesizing optoelectronic materials based on a physical model and a machine learning model according to claim 1, characterized in that: The process of step S2 is as follows: S201, COMSOL software builds a 2D or 3D simulation model of the device according to the device structure parameters, and the device structure parameters include boundary length, device layer width, and number of device layers; S202, assigning material parameters to the 2D or 3D simulation model, where the material parameters include material transmittance, absorption, bandgap width, and conductivity; S203, selecting the physical fields involved in the device according to the working principle of the device, including diffusion and drift motion of charged particles, particle recombination at the interface, and optical-electrical conversion; S204. Select the physical formula involved according to the selected physical field.

4. The method for synthesizing optoelectronic materials based on a physical model and a machine learning model according to claim 3, characterized in that: The physical formula includes the electric flux formula: Total flux N i = ∇(−D i ∇c i −z i μ i Fc i ∇ϕ i ) + uc i In the formula, 𝑐 𝑖 represents the concentration of the substance, 𝐷 𝑖 is the diffusion coefficient, 𝜙 𝑖 is the electric potential, F is the Faraday constant, 𝑧 𝑖 is the charge number, μ i is the electromigration number, u represents the fluid flow rate, and ∇ represents the gradient operation symbol; The physical formula also includes the Butler–Volmer formula: j loc =j0{exp (-nαFη / RT )-exp[n(1-α)Fη} Among them, 𝑗 l𝑜𝑐 and 𝑗0 are the local current density and exchange current density respectively, α is the transfer coefficient, n is the number of electrons transferred, R is the gas constant, T is the reaction temperature, and η is the overpotential.

5. The method for synthesizing optoelectronic materials based on a physical model and a machine learning model according to claim 1, characterized in that: The process of step S3 is as follows: S301, normalizing all selected data; S302, the selected data is composed of a group of data including synthesis formula parameters of optoelectronic materials, optical and electrical parameters of optoelectronic materials, and parameters of optoelectronic devices, and all the groups of data constitute a data set; S303. The selected data set is divided into K subsets, K>5, one of which is used as a test set, and the rest are used as training sets.

6. The method for synthesizing optoelectronic materials based on a physical model and a machine learning model according to claim 5, characterized in that: The machine learning models in step S4 include linear regression, Lasso regression, kernel ridge regression, Gaussian process regression, support vector regression, k-nearest neighbor regression, and decision tree regression, a total of 7 machine learning models. The machine learning models are directly imported from the python database, and the process is as follows: S401, import the test set into the machine learning model, select the number of iterations N, N>10, and output the performance evaluation index; the selected performance evaluation index includes the root mean square error RMSE and the determination coefficient R 2 , as follows; R 2 Represents the square of the correlation coefficient between the predicted value and the true value. The closer the value is to 1, the better the fitting effect of the machine learning model is. The calculation expression is: in, is the true value of the i-th sample, is the predicted value of the ith sample, is the mean of the true value of the sample, m is the total number of samples, Var is the variance of the true value of the sample, and the root mean square error RMSE is the mean square root of the square difference between the predicted value and the true value. The smaller the value, the better the fitting effect of the machine learning model. The calculation expression is: S402. Repeat the cross-validation for 7 times. The data set is divided into K subsets. One of the subsets is used as the test set each time, and the rest are used as training sets. The test set is used to evaluate the performance of the machine learning model after each training. Repeat 7 times to obtain the performance evaluation indicators of 7 machine learning models, and then take the average value as the final evaluation result.

7. The method for synthesizing optoelectronic materials based on a physical model and a machine learning model according to claim 1, characterized in that: The process of step S5 is as follows S501, selecting the machine learning model with the best fitting effect as the prediction model to predict the synthesis of optoelectronic materials; S502. Change the type of input data of the prediction model, wherein the change of input data refers to any one or more of the synthesis formula parameters of optoelectronic materials, optical and electrical parameters of optoelectronic materials, and optoelectronic device parameters, and the output data is the remaining one or more of the optoelectronic device parameters.

Images