Method for preparing garnet type solid electrolyte under assistance of machine learning model

Abstract

The invention discloses a method for preparing garnet type solid electrolyte under the assistance of a machine learning model, and the method comprises the steps: inputting the data of a to-be-screened material into a structure stability parameter model, and screening out materials with stable structures to form a first material set; the to-be-screened material data set comprises a compound A-B-C-O containing a plurality of cations; determining a second material set meeting A, B and C element availability evaluation requirements, and selecting a structure stability test material from the second material set as a to-be-optimized material; a GBR model algorithm is used for the to-be-optimized material, and information of different doping structures is obtained; and according to the structure information, using an ionic conductivity model, and screening out a solid electrolyte material with high ionic conductivity as an output result through the ionic conductivity model. According to the method, the garnet type solid electrolyte with excellent ionic conductivity is obtained by screening an ionic conductivity model, including methods of optimizing structural parameter selection, element composition and the like.

Description

Technical Field

[0001] The present invention relates to the field of new energy material data mining, and in particular to a method for preparing a garnet-type solid electrolyte with the assistance of a machine learning model. Background Art

[0002] As we all know, with the rapid development of new energy industries such as the electric vehicle industry, the requirements for battery performance are getting higher and higher. All-solid-state batteries have the advantages of high safety, high temperature stability, and high energy density, which have attracted great attention from researchers. The performance of all-solid-state batteries depends largely on the performance of solid-state electrolytes. Scientific design and synthesis of high ionic conductivity solid electrolytes are crucial.

[0003] In recent years, researchers at home and abroad have carried out a series of work on the research of inorganic solid electrolyte materials based on material genome, and have made significant progress in high-throughput screening of inorganic solid electrolyte materials, machine learning prediction of new electrolyte materials, and data mining to discover the structure-activity relationship of electrolyte materials. This has greatly promoted the development of inorganic solid electrolytes. The exploration and development of inorganic solid electrolyte materials is a key task for all-solid-state batteries.

[0004] At present, many widely used solid electrolyte materials are discovered through the "cooking method", but the research and development cycle of solid electrolyte materials with excellent performance is long, and it is difficult to develop them using traditional experimental methods. Therefore, it is possible to combine machine learning with traditional experimental methods, use theoretical models to explore the structure-activity relationship of electrolyte materials, assist in designing materials and predict performance.

[0005] A large number of computational tasks are designed using machine learning models to assist in the preparation of solid electrolytes. The screening process usually includes steps such as iterative model optimization, structural parameter calculation, ionization energy analysis, and ionic conductivity calculation. The first step involves finding the best parameter value among a large number of candidate learners, maximum depth, learning rate, and other parameters. Since there are many parameters to be explored and they are highly complex, the iterative optimization cycle is long. Finally, multiple verification methods are needed to select the optimal model to predict ionic conductivity. Summary of the invention

[0006] The present invention discloses a method for preparing a solid electrolyte using a machine learning model, which is intended to solve the problem of lack of auxiliary design tools when machine learning methods are used to assist in finding garnet-type solid electrolyte materials with suitable material composition and stable structure. The method obtains a garnet-type solid electrolyte with excellent ionic conductivity by screening an ionic conductivity model, including optimizing structural parameter selection, element composition and other methods.

[0007] The present invention discloses a method for preparing a garnet-type solid electrolyte assisted by a machine learning model, which specifically comprises the following steps: S1. Inputting a data set of materials to be screened into an element classification model, and screening out materials of the category to which the elements belong to form a first material set: the data set of materials to be screened includes a compound ABCO containing multiple cations as a material to be optimized, and A, B, and C are all metal elements; S2. According to the predetermined A, B, C element classification parameters, determine a second material set from the first material set that meets the A, B, C element availability evaluation requirements; S3, performing doping modification on the material to be optimized according to the preselected doping material, obtaining derived structural information of different doping amounts and doping sites, obtaining structural stability parameters after doping, and forming a structural stability model as a third material set; S4. Input the third material set into the ionic conductivity model, and screen out the ionic conductivity corresponding to the derived structure as the optimized output result.

[0008] Preferably, the original compound of the material ABCO to be optimized in step S1 is Li-La-Zr-O; the characteristic element classification model classifies different metals into different sites of A, B, and C.

[0009] Preferably, the availability evaluation parameters A, B, and C in step S2 are determined by evaluating the price and abundance of the elements.

[0010] Preferably, in step S3, according to each derived structure information, a GBR model is used to calculate the structural stability parameters to determine the most stable structure at each doping amount.

[0011] Preferably, the structural stability parameter corresponding to the determined derived structural information is specifically: The stability parameters corresponding to each derived structure are determined according to the configuration, ion stacking position and ion radius, and the structural stability model is determined; among which, the stability parameters are structurally checked by the GBR model to determine the structural stability information.

[0012] Preferably, the structural stability model uses a GBR algorithm; the construction of the structural stability model includes: Using the inorganic crystal database, we can obtain all the crystals containing a specific cation M. n+ The basic data set is constructed by using compounds, where M n+ Including the first six periods of metal ions; based on the valence bond theory, the structural stability parameters of the garnet solid electrolyte are calculated, and the M with structural parameters between 0.8 and 1.3 is selected. n+ As a structural stability model data set.

[0013] Preferably, the ionic conductivity model in step S4 uses a GBR algorithm; the construction of the ionic conductivity model includes: Using the inorganic crystal database, we can obtain all the crystals containing a specific cation M. n+ The basic data set is constructed by using compounds, where M n+ Includes the first six periods of metal ions; The M was calculated based on the valence bond theory. n+ The ionization energy, affinity energy, and electronegativity parameters of the compound are used to train the ionic conductivity model.

[0014] Preferably, the screening optimization method is used for optimizing the screening of garnet-type solid electrolyte materials.

[0015] The present invention also discloses a method for preparing a garnet-type solid electrolyte assisted by a machine learning model. The garnet-type solid electrolyte material selected by the optimization is Li 6.4 Ln3Z 1.4 Ta 0.6 O 12 , where the stoichiometric ratio of Li, La, Zr, Ta, and O in the chemical formula is 32:15:7:3:60, and the space group is Ia- d.

[0016] Preferably, the garnet-type solid electrolyte material is a solid electrolyte material used in lithium-ion batteries or lithium metal batteries.

[0017] Therefore, the present invention adopts the above-mentioned machine learning model to assist in the preparation of a garnet-type solid electrolyte method, which has the following beneficial effects: The present invention uses a machine learning model to assist in the screening and optimization of solid electrolyte materials, and develops an accelerated screening process for garnet-type solid electrolyte materials. The machine learning model is used to accelerate the time-consuming steps in the screening process, which is used to accelerate the calculation of doping structure parameters and ionic conductivity, greatly reducing the material screening cycle, and more accurately achieving the screening and optimization of garnet-type solid electrolyte materials. The optimal doping scheme is explored through the selection optimization method, including the optimal structural stability parameters and the ratio of doping elements, so that the garnet-type solid electrolyte obtained by the screening optimization method of the present invention has better ionic conductivity. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flow chart of a method for preparing a garnet-type solid electrolyte assisted by a machine learning model according to the present invention; Figure 2 A fitting diagram of training data and test data between the ionic conductivity predicted by the model of the present invention and the actual ionic conductivity; Figure 3 is a graph of the Pearson coefficient of the material of the present invention; Figure 4 is a graph of the importance of the characteristics of the materials of the present invention; Figure 5 Ionic conductivity distribution diagram of garnet solid electrolyte material collected by the present invention; Figure 6 Schematic diagram of the LLZO crystal structure of the present invention; Figure 7 This is a schematic diagram of the LLZTO crystal structure after optimization in the present invention; Figure 8 The data fitting diagram of the solid electrolyte ion conductivity of the present invention is DETAILED DESCRIPTION

[0019] The present invention will be further described below through specific examples, but it should be understood that these examples are only used for more detailed description and should not be understood as limiting the present invention in any form, that is, they are not intended to limit the protection scope of the present invention.

[0020] The present invention proposes a method for accelerating the screening and optimization of garnet-type solid electrolyte materials using a machine learning model, which is used to accelerate the screening and optimization of garnet-type solid electrolytes. The main method steps are as follows: Figure 1 As shown, including; Step 1, inputting the material data set to be screened into the element classification model, and screening out the materials of the category to which the element belongs to form a first material set; Step 2, according to the predetermined A, B, C element classification parameters, determining a second material set that meets the A, B, C element availability evaluation requirements in the first material set; Step 3, doping and modifying the material to be optimized according to the pre-selected doping material to obtain derivative structures with different doping amounts and doping sites, and obtaining structural stability parameters after doping as a third material set; Establish a structural stability model and use the inorganic crystal database to obtain all the n+ The basic data set is constructed by using compounds, where M n+ Includes the first six periods of metal ions; Calculate the structural stability parameters of the compounds based on the structural stability parameter model, and select compounds with structural stability parameters between 0.8 and 1.3 as the model data set; The structural stability parameters of the compounds are used as labels for the model outputs, and the model data set composed of compounds with structural stability parameters selected by the present invention for model training contains 398 training samples in total.

[0021] Feature extraction of model data Material features used as model input: basic material properties based on composition and structure. Features can be divided into tetrahedral lattice ions, octahedral lattice ions, and dodecahedral lattice ions according to the type of ions of interest; Specifically, based on the basic properties of the material composition and structure, the features are divided into tetrahedral structure, octahedral structure and dodecahedral structure according to the different ion configurations of interest. The corresponding valence state, coordination number and radius features of different ions are added.

[0022] Based on the extracted features, the structural stability parameter model is trained.

[0023] Since the ionic radii of all compounds in the inorganic crystal database are known, the model can be trained based on the known ionic radii, with the training range being 0.8-1.3.

[0024] Step 4: Establish an ionic conductivity model: Using the inorganic crystal database, we can obtain all the crystals containing a specific cation M. n+ The basic data set is constructed by using compounds, where M n+ Includes the first six periods of metal ions; The M was calculated based on the valence bond theory. n+ The ionization energy, affinity energy, and electronegativity parameters of the compound are used to train the ionic conductivity model; The data defining the range of structural stability parameters were input into the ionic conductivity model.

[0025] The ionic conductivity model obtained through training has a score of 0.9 in the test set and an MSE of 2.17×10 -7 , the prediction results of this model can be used for ionic conductivity screening.

[0026] The relationship between the ionic conductivity predicted by the model and the actual ionic conductivity is shown in Figure 2 As shown in the figure, most of the data points are distributed around the reference line, indicating that the model has a good data fitting effect. The GBR algorithm is a regression analysis algorithm that enhances the performance of the basic algorithm and can give the most important features in the model. Figure 3 The ten most important features of the model are shown. Figure 4 The correlation coefficient graph between the ten most important features is given. Variables with positive correlation parameters are positively correlated, and vice versa.

[0027] The trained ionic conductivity model is used to screen polar materials in the material data set to be screened. The material data set to be screened includes ABCO, a compound containing multiple cations, and the original compound is Li-La-Zr-O, where A, B, and C are metal elements; During the implementation of the present invention, data of 398 garnet solid electrolyte materials were collected and applied to the training model. The distribution of ionic conductivity is shown in Figure 5 .

[0028] Through shap interpretability analysis, we find the influence of features on performance. According to the shap features, we adjust the proportion of elements in the model, obtain the weight values ​​of each feature, and select the material Li-La-Zr-Ta-O that meets the composite requirements of ion conductivity. In the compound ABCO, the original Zr of the C element is replaced by the Ta site. The selected stable structure is Li 6.4 Ln3Z 1.4 Ta 0.6 O 12 ; Among them, the stoichiometric ratio of Li, La, Zr, Ta, and O in the chemical formula is 32:15:7:3:60, and the space group is Ia- d.

[0029] The XRD structure of Li-La-Zr-Ta-O is shown in Figure 6 , The crystal structure of Li-La-Zr-Ta-O is shown in Figure 7 .

[0030] According to the crystal information corresponding to the derived structure, the ionic conductivity is calculated for different doping amounts and doping sites, and it is determined whether the fitting degree of the training set and the test set reaches the expected effect. The fitting degree of the training set and the test set in the present invention reaches 90%. If the fitting degree of the sample data of the training set and the test set does not reach 90%, the ionic conductivity model is reconstructed.

[0031] Performance prediction diagram Figure 8 .

[0032] The predicted sample is input into the ionic conductivity model after the fit test to predict the ionic conductivity of the solid electrolyte. In this example, after the ionic conductivity model is trained, it reaches a test set fit of more than 90%, and the real garnet-type solid electrolyte material is almost the same as the ionic conductivity material predicted by the ionic conductivity model, indicating that the ionic conductivity model of the present invention can well predict the ionic conductivity of the garnet-type solid electrolyte, and the process is fast and stable.

[0033] The specific implementation methods described above provide further information on the objectives, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific implementation method of the present invention and does not limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for preparing a garnet-type solid electrolyte assisted by a machine learning model, characterized in that: The specific steps include: S1. Inputting a data set of materials to be screened into an element classification model, and screening out materials of the category to which the elements belong to form a first material set: the data set of materials to be screened includes a compound ABCO containing multiple cations as a material to be optimized, and A, B, and C are all metal elements; S2. According to the predetermined A, B, C element classification parameters, determine a second material set from the first material set that meets the A, B, C element availability evaluation requirements; S3, performing doping modification on the material to be optimized according to the preselected doping material, obtaining derived structural information of different doping amounts and doping sites, obtaining structural stability parameters after doping, and forming a structural stability model as a third material set; S4. Input the third material set into the ionic conductivity model, and screen out the ionic conductivity corresponding to the derived structure as the optimized output result.

2. The method for preparing a garnet-type solid electrolyte assisted by a machine learning model according to claim 1, characterized in that: In step S1, the original compound of the material ABCO to be optimized is Li-La-Zr-O; the characteristic element classification model classifies different metals into different sites of A, B, and C.

3. The method for preparing garnet-type solid electrolyte assisted by a machine learning model according to claim 1, characterized in that: The availability evaluation parameters A, B, and C in step S2 are determined by evaluating the price and abundance of the elements.

4. The method for preparing garnet-type solid electrolyte assisted by a machine learning model according to claim 1, characterized in that: In step S3, according to each derived structure information, the GBR model is used to calculate the structural stability parameters to determine the most stable structure at each doping amount.

5. The method for preparing garnet-type solid electrolyte assisted by a machine learning model according to claim 4, characterized in that: The structural stability parameter corresponding to the derived structural information is specifically determined as follows: The stability parameters corresponding to each derived structure are determined according to the configuration, ion stacking position and ion radius, and the structural stability model is determined; among which, the stability parameters are structurally checked by the GBR model to determine the structural stability information.

6. The method for preparing garnet-type solid electrolyte assisted by a machine learning model according to claim 5, characterized in that: The structural stability model uses the GBR algorithm; The construction of the structural stability model includes: Using the inorganic crystal database, we can obtain all the crystals containing a specific cation M. n+ The basic data set is constructed by using compounds, where M n+ Including the first six periods of metal ions; based on the valence bond theory, the structural stability parameters of the garnet solid electrolyte are calculated, and the M with structural parameters between 0.8 and 1.3 is selected. n+ As a structural stability model data set.

7. The method for preparing a garnet-type solid electrolyte assisted by a machine learning model according to claim 1, characterized in that: The ionic conductivity model in step S4 uses the GBR algorithm; the construction of the ionic conductivity model includes: Using the inorganic crystal database, we can obtain all the crystals containing a specific cation M. n+ The basic data set is constructed by using compounds, where M n+ Includes the first six periods of metal ions; The M was calculated based on the valence bond theory. n+ The ionization energy, affinity energy, and electronegativity parameters of the compound are used to train the ionic conductivity model.

8. The method for preparing garnet-type solid electrolyte assisted by a machine learning model according to claim 1, characterized in that: The screening optimization method is used for optimizing the screening of garnet-type solid electrolyte materials.

9. A garnet-type solid electrolyte material, characterized in that: The garnet-type solid electrolyte material optimized and screened by a method for preparing garnet-type solid electrolytes assisted by a machine learning model is Li 6.4 Ln3Z 1.4 Ta 0.6 O 12 , where the stoichiometric ratio of Li, La, Zr, Ta, and O in the chemical formula is 32:15:7:3:60, and the space group is Ia- d.

10. The garnet-type solid electrolyte material according to claim 9, characterized in that: The garnet-type solid electrolyte material is a solid electrolyte material used in lithium-ion batteries or lithium metal batteries.

Images