Prediction device and prediction method

Abstract

To realize a prediction device and a prediction method that predict the specific physical property value of an inorganic compound with high accuracy.SOLUTION: A prediction device 1 comprises an input acquisition unit (11) that acquires input information (31) concerning a first inorganic compound, and a prediction unit (13) that predicts, using a trained model (213), the specific physical property value (410) of the first inorganic compound from the input information (31). The trained model (213) has been trained for a relationship between reference information (32) concerning a second inorganic compound whose physical property value is known and the physical property value. The input information (31) and the reference information (32), respectively, include element information (310) that indicates the elements of inorganic compounds, composition information (311) that indicates the ratio of elements, and first structure information (312) that indicates the unit lattice and the transitional property and symmetry of crystal structure of inorganic compounds.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a prediction device and a prediction method for predicting the physical properties of a compound. [Background technology]

[0002] In material development, there is a demand for the ability to accurately predict the physical properties of materials before development. For example, Patent Document 1 discloses a technology for predicting the characteristic values ​​of a material by acquiring the composition formula and structural information of the material. Also, Non-Patent Document 1 discloses a technology for calculating the contribution of each element in an inorganic compound and predicting the physical properties of the inorganic compound. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Patent Publication 2020 / 031671 [Non-patent literature]

[0004] [Non-Patent Document 1] Anthony Yu-Tung Wang, npj Computational Materials (2021),77. Summary of the Invention [Problem to be solved by the invention]

[0005] One aspect of the present disclosure is to provide a prediction device and a prediction method that predict specific physical property values ​​of inorganic compounds with high accuracy. [Means for solving the problem]

[0006] A prediction device according to one embodiment of the present disclosure includes an input acquisition unit that acquires input information of a first inorganic compound whose specific physical property value is unknown, and a prediction unit that predicts the physical property value of the first inorganic compound from the input information using a trained model, wherein the trained model has undergone a training process to learn the relationship between reference information of a second inorganic compound whose physical property value is known and the physical property value, and the input information and the reference information each have elemental information indicating the elements of the inorganic compound, compositional information indicating the ratio of the elements, and first structural information indicating the unit cell and translational and symmetry properties of the crystal structure of the inorganic compound.

[0007] A prediction method according to one embodiment of the present disclosure includes an input acquisition step of acquiring input information of a first inorganic compound whose specific physical property value is unknown, and a prediction step of predicting the physical property value of the first inorganic compound from the input information using a trained model, wherein the trained model has been trained to determine the relationship between reference information of a second inorganic compound whose physical property value is known and the physical property value, and the input information and the reference information each have elemental information indicating the elements of the inorganic compound, compositional information indicating the ratio of the elements, and first structural information indicating the unit cell and translational and symmetry properties of the crystal structure of the inorganic compound.

[0008] The prediction device according to each aspect of the present disclosure may be realized by a computer. In this case, the control program for the prediction device that causes the computer to operate as each part (software element) of the prediction device to realize the prediction device, and the computer-readable recording medium on which the control program is recorded, also fall within the scope of the present disclosure. [Effects of the Invention]

[0009] According to one aspect of the present disclosure, it is possible to predict specific physical property values ​​of inorganic compounds with high accuracy. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a block diagram illustrating an example of a configuration of a prediction device according to an embodiment of the present disclosure. [Figure 2] FIG. 4 is a diagram illustrating an example of a data structure of input information. [Figure 3] FIG. 10 is a diagram illustrating an example of a data structure of output information. [Figure 4] FIG. 10 is a functional block diagram of a prediction unit that performs prediction processing based on a trained model. [Figure 5] FIG. 1 is a schematic diagram illustrating a part of the processing performed using a trained model. [Figure 6] FIG. 1 is a schematic diagram illustrating a part of the processing performed using a trained model. [Figure 7] FIG. 10 is a diagram illustrating an example of display information. [Figure 8] FIG. 10 is a diagram illustrating an example of display information. [Figure 9] FIG. 10 is a diagram illustrating an example of display information. [Figure 10] 10 is a flowchart illustrating an example of the flow of a prediction process executed by the prediction device. [Figure 11] FIG. 10 is a block diagram illustrating an example of a configuration of a prediction device according to a modified example of the present disclosure. [Figure 12] FIG. 10 is a diagram illustrating an example of display information. [Figure 13] 10 is a graph showing the correlation between the measured values ​​and predicted values ​​of magnetic moments related to training data of an example. [Figure 14] 10 is a graph showing the correlation between the measured value and the predicted value of the magnetic moment according to the verification data of the example. [Figure 15] 10 is a graph showing the correlation between the actual measured value and the predicted value of the magnetic moment related to the training data of the comparative example. [Figure 16] 10 is a graph showing the correlation between the measured value and the predicted value of the magnetic moment related to the verification data of the comparative example. DETAILED DESCRIPTION OF THE INVENTION

[0011] <Embodiment 1> [Outline of prediction device] In the past, developing new materials with target physical properties relied on the experience and intuition of researchers. However, the inventors of the present invention have established a method for predicting specific physical properties of unknown inorganic compounds by having a prediction device learn the relationship between the elemental information, composition information, and first structural information indicating the unit cell, translational property, and symmetry of the crystalline structure of the inorganic compound, of the inorganic compound, and the physical properties.

[0012] The prediction device of the present disclosure can be used, for example, in research facilities where research and development of new materials is conducted. The prediction device of the present disclosure uses, as input information, elemental information indicating the elements of an inorganic compound, compositional information indicating the ratio of the elements, and first structural information indicating the unit cell, symmetry, and translational nature of the crystalline structure of the inorganic compound. This allows the prediction device to accurately predict specific physical property values ​​of the inorganic compound.

[0013] The prediction device employs first structural information indicating the unit cell, translational property, and symmetry of the crystalline structure of the inorganic compound as input information, thereby enabling the prediction device to accurately predict the physical property values ​​even for inorganic compounds in which the crystalline structure dominates the physical property values.

[0014] Furthermore, by employing information indicating the unit cell, translational property, and symmetry of the crystalline structure of the inorganic compound as the first structural information, the user can easily input structural information without having to specify the detailed spatial arrangement of each atom of the inorganic compound when inputting the information. This also has the effect of making it easier for the user to prepare a data set of structural information when predicting the properties of not only known compounds but also unknown compounds.

[0015] [Configuration of prediction device 1] Next, the configuration of the prediction device 1 will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of the configuration of the prediction device 1.

[0016] The prediction device 1 is a computer that receives, as input information 31, elemental information indicating the elements of the inorganic compound, compositional information indicating the ratio of the elements, and first structural information indicating the unit cell, symmetry, and translational property of the crystal structure of the inorganic compound, and predicts specific physical property values ​​of the inorganic compound.

[0017] As shown in FIG. 1, the prediction device 1 includes a control unit 10 that controls each unit of the prediction device 1 in an integrated manner, a memory unit 20 that stores various data used by the control unit 10, and a communication unit 30 that communicates with external devices.

[0018] The storage unit 20 stores a trained model 213, a first inorganic compound database 210, and a second inorganic compound database 211.

[0019] The trained model 213 is a model that has undergone a learning process to learn the relationship between reference information 32 of the second inorganic compound, whose physical property values ​​are known, and specific physical property values. The reference information 32 has element information indicating the elements of the inorganic compound that is the second inorganic compound, composition information indicating the ratio of the elements, and first structure information indicating the unit lattice, translational property, and symmetry of the crystal structure of the inorganic compound that is the second inorganic compound. Details of the element information, composition information, first structure information, and physical property values ​​will be described later.

[0020] The first inorganic compound database 210 is a database in which element information, composition information, first structure information, etc. of inorganic compounds, which are first inorganic compounds with unknown physical properties, are stored in correspondence with each other. Here, if the physical properties differ for each unit lattice, "unknown" or "known" is determined for each unit lattice. For example, "unknown physical properties" corresponds to the fact that even if the chemical formula of an inorganic compound is known, the physical properties based on the unit lattice are unknown.

[0021] The second inorganic compound database 211 is a database that stores reference information 32 of second inorganic compounds whose physical property values ​​are known. Specifically, the reference information 32 is information in which element information, composition information, and first structure information of the second inorganic compound are associated with the physical property values. The second inorganic compound database 211 may further include second structure information and third structure information, which will be described later.

[0022] In this embodiment, the prediction device 1 includes a first inorganic compound database 210 and a second inorganic compound database 211, but is not limited to this. The first inorganic compound database 210 and the second inorganic compound database 211 may be stored on an external server, and the prediction device 1 may access each database stored on the server. In this case, the first inorganic compound database 210 and the second inorganic compound database 211 may be stored on the same server, or each database may be stored on a different server. In addition, the storage unit 20 may store a control program for performing various controls of the prediction device 1.

[0023] <Control unit 10> The control unit 10 includes an input acquisition unit 11, a prediction unit 13, a creation unit 14, and a learning unit 15. Each component included in the control unit 10 will be described below.

[0024] (Study Section 15) The learning unit 15 generates a trained model 213. The trained model 213 is generated using, as training data, reference information 32 of a second inorganic compound whose physical property values ​​are known, i.e., a data set of elemental information, compositional information, and first structural information, and the physical property values. The training data may be stored in, for example, the storage unit 20 described above. Specifically, the training data may be stored in the second inorganic compound database 211. In this case, the training data may be acquired from the second inorganic compound database 211 via the input acquisition unit 11 described later. Alternatively, the training data may be acquired from an external device via the input acquisition unit 11. Details of the elemental information, compositional information, first structural information, and physical property values ​​will be described later. Also, the learning unit 15 may change the hyperparameter set of the trained model 213 as appropriate.

[0025] (Input acquisition unit 11) The input acquisition unit 11 acquires input information 31 of a first inorganic compound whose specific physical property value is unknown. The input acquisition unit 11 may acquire the input information 31 from an external device other than the prediction device 1.

[0026] Furthermore, the input information 31 for one inorganic compound may be inputted per input, or the input information 31 for multiple inorganic compounds may be inputted per input. The input information 31 may be acquired from the first inorganic compound database 210.

[0027] 2 is a diagram showing an example of the data structure of input information 31. The input information 31 includes element information 310, composition information 311, and first structure information 312. The input information 31 may also include second structure information 313 and third structure information 314.

[0028] The element information 310 is information indicating the types of elements contained in an inorganic compound. For example, in Fig. 2, the element information 310 indicates that the inorganic compound BaTiO3 contains the elements "Ba," "Ti," and "O."

[0029] The composition information 311 is information indicating the ratio of elements contained in an inorganic compound. For example, in Fig. 2, the composition information 311 of the inorganic compound BaTiO3 is "Ba:Ti:O=1:1:3", which indicates that the ratio of the element Ba, the element Ti, and the element O is 1:1:3.

[0030] The first structural information 312 is information that indicates the unit cell, translational properties, and symmetry of the crystalline structure of the inorganic compound.

[0031] The unit cell of the crystal structure of an inorganic compound is the smallest unit contained in the crystal of the inorganic compound, and for example, a known unit cell can be used. Specifically, 14 types of Bravais lattices are adopted, taking into account seven types of crystal systems (cubic, tetragonal, orthorhombic, hexagonal, trigonal, monoclinic, and triclinic) and four types of lattice systems (simple (P), body-centered (I), face-centered (F), and base-centered (C)).

[0032] The translational and symmetry properties of a crystal structure are expressed in terms of a space group starting from the center of the crystal. The translational properties of a crystal structure specifically include translational, helical, and glide symmetry elements, while the symmetry specifically includes rotational, reflexive, inverted, and reflective symmetry elements without translation.

[0033] The first structural information 312 is determined to be one of 230 three-dimensional space groups by combining unit lattices, symmetries, and translational properties. Specifically, 230 three-dimensional spaces are determined by combining 14 types of Bravais lattices, 32 types of point groups, and translation operations. In other words, the first structural information 312 of an inorganic compound is assigned one of the 230 three-dimensional space groups. The space group symbol determined in international notation or a corresponding number may be input as the first structural information 312. In FIG. 2, as an example, the space group number "221" of the inorganic compound BaTiO3 is input as the first structural information 312. Hereinafter, the first structural information will also be referred to as the "space group."

[0034] The input information 31 may further include second structural information 313 indicating the length of each axis of the unit cell. The length of each axis of the unit cell may be expressed in, for example, Å (angstroms) or nm (nanometers).

[0035] In FIG. 2, the second structure information 313 is information "a=b=c=3" indicating that the lengths of the axes of the unit cell are the same and that the length of each axis is 3 Å.

[0036] In this way, by further using the second structural information 313 to predict the physical properties of the inorganic compound, the prediction device 1 can predict the physical properties with high accuracy.

[0037] The input information may further include third structural information 314 that indicates the angles between the edges of the unit cell.

[0038] In FIG. 2, the third structural information 314 is information that indicates that the angle α between axis a and axis b of the unit cell, the angle β between axis b and axis c, and the angle γ between axis c and axis a are all equal and each angle is 90°.

[0039] In this way, by further using the third structural information 314 to predict the physical properties of the inorganic compound, the prediction device 1 can predict the physical properties with higher accuracy.

[0040] (Prediction Section 13) The prediction unit 13 predicts a specific physical property value of an inorganic compound from the input information 31 using the trained model 213.

[0041] Specific physical properties of inorganic compounds include, for example, magnetic moment, formation energy of one molecule, formation energy per atom, band gap, density, stability from thermodynamic convex hull, and volume of unit structure.

[0042] The trained model 213 is a model that has undergone a training process to learn the relationship between the reference information 32 (element information, composition information, and first structure information) of the second inorganic compound, whose physical property values ​​are known, and the physical property values. Details of the trained model 213 will be described later.

[0043] The prediction unit 13 may predict, from the input information 31, a first contribution indicating the degree of influence of each element contained in the first inorganic compound on the physical property value of the first inorganic compound.

[0044] The elements contained in inorganic compounds affect the physical properties of the inorganic compound. Furthermore, the multiple elements contained in inorganic compounds do not all affect the physical properties in the same way; the degree of influence on the physical properties varies depending on the type of element. The first contribution is thus information that indicates the degree of influence each element has on the physical properties of the inorganic compound, i.e., the magnitude of its contribution to the physical properties.

[0045] The first contribution may be a value calculated for each element contained in the first inorganic compound, or may be the total value of the contributions of all elements contained in the first inorganic compound. The first contribution may be expressed, for example, as a numerical value. It can be said that the larger the numerical value of the first contribution, the greater the influence that the element exhibiting that first contribution has on the physical property value of the first inorganic compound.

[0046] The prediction unit 13 being able to predict the first contribution allows the user to know which element contained in the first inorganic compound has a large effect on the physical property value of the first inorganic compound.

[0047] The prediction unit 13 may predict a second contribution indicating the degree of influence on the physical property value of the first inorganic compound that is influenced by the structure indicated by the structural information from the input information 31.

[0048] Structural information of inorganic compounds affects the physical properties of the inorganic compounds. For example, even if the types of elements contained in inorganic compounds are the same, different structural information can result in different physical properties. The second contribution is thus information that indicates the degree of influence that structural information has on the physical properties of inorganic compounds, i.e., the magnitude of its contribution to the physical properties.

[0049] The second contribution is expressed, for example, by a numerical value. The larger the numerical value of the second contribution, the greater the influence of the structural information indicating the second contribution on the physical property value of the inorganic compound. The second contribution and the above-mentioned first contribution can be compared, for example. Therefore, for example, if the structure of an inorganic compound has a greater influence on a physical property value than the influence of a certain element on the physical property value, the second contribution may be greater than the first contribution. In this case, for example, when a user attempts to control (e.g., increase) the physical property value of this inorganic compound, the user can obtain a guideline that indicates that they should place importance on the design of the inorganic compound's structure.

[0050] 3 is a diagram showing an example of output information 41 output by prediction unit 13 when input information 31 about inorganic compound BaTiO3 shown in FIG. 2 is input. Output information 41 includes first contribution 411, second contribution 412, and physical property value 410. Furthermore, first contribution 411 indicates the contribution of each element of inorganic compound BaTiO3. First contribution 411, second contribution 412, and physical property value may each be expressed as a numerical value.

[0051] The prediction unit 13 may predict the physical property value based on a first contribution indicating the degree of influence of each element contained in the first inorganic compound on the physical property value of the first inorganic compound, and a second contribution indicating the degree of influence of the structure indicated by the structural information on the physical property value of the first inorganic compound, as described below. Specifically, the prediction unit 13 may calculate the physical property value of the first inorganic compound by calculating the average of the first contribution and the second contribution of each element. This allows the prediction device 1 to predict the physical property value of the first inorganic compound with high accuracy.

[0052] The prediction unit 13 may calculate a third contribution from the first contribution and the second contribution of at least one element included in a predetermined period among the elements included in the first inorganic compound. The third contribution indicates the degree of influence of at least one element included in the predetermined period on the physical property values ​​of the first inorganic compound. When the first inorganic compound has a predetermined structure, the third contribution is an index that indicates the degree to which at least one element included in the first inorganic compound contributes to the physical property values ​​of the inorganic compound when the first inorganic compound has a predetermined structure.

[0053] The first contribution is the degree of influence on the physical property values ​​of an inorganic compound calculated for each element, and does not take into account structural information of the inorganic compound. However, the degree of influence of an element on the physical property values ​​of an inorganic compound can vary depending on the structure in which the element of the inorganic compound is arranged. Therefore, by calculating the third contribution, the user can know the degree of influence that an element of the inorganic compound has on the physical property values ​​of the inorganic compound when the element of the inorganic compound is arranged in a specified structure.

[0054] The predetermined structure is a structure indicated by the structural information included in the input information 31. In this case, the structural information may be only the first structural information 312, or may include the first structural information 312, the second structural information 313, and the third structural information 314.

[0055] The predetermined period is not particularly limited, but may be, for example, an element of the fourth period or the fifth period.

[0056] [Trained model] As described above, the function of the prediction unit 13 to predict physical property values ​​is realized by the prediction unit 13 inputting the input information 31 into the trained model 213.

[0057] The trained model 213 will be described below. FIG. 4 shows an example of a functional block diagram of the prediction unit 13 that performs prediction processing based on the trained model 213. Here, the information obtained after performing predetermined processing on the element information 310 and the composition information 311 will also be simply referred to as "element information" and "composition information." Furthermore, the information obtained after performing predetermined processing on the first structure information 312, the second structure information 313, and the third structure information 314 will also be simply referred to as "structure information."

[0058] The trained model 213 includes a first conversion unit 2130, a first NN unit 2131, a second conversion unit 2132, an addition unit 2133, a vector conversion unit 2134, a scaling unit 2135, a structural information integration unit 2136, a second NN unit 2137, an integration unit 2138, a relationship learning unit 2139, a third NN unit 2140, and a calculation unit 2141.

[0059] The first conversion unit 2130 converts the input element information 310 into a unique array. Specifically, the first conversion unit 2130 converts the atomic numbers of the elements that make up the inorganic compound into an embedded representation of the element information. As an example, the first conversion unit 2130 converts information such as the atomic number "1" of a hydrogen atom into a unique array such as [0.0571, -0.2255, ]. As an example, the first conversion unit 2130 uses Mat2vec embedding processing on the element information 310.

[0060] The first NN unit 2131 converts the number of dimensions of the element information having a unique sequence converted by the first conversion unit 2130 into a predetermined number of dimensions through a neural network. The neural network used in the first NN unit 2131 may be, for example, a neural network that does not have an intermediate layer between the input layer and the output layer, and the input layer and the output layer may be weighted by a predetermined weight parameter W1. In other words, the weight W1 is a parameter that is learned during model learning.

[0061] The second conversion unit 2132 converts the input composition information 311 into an array of a specific length. As an example, positional encoding processing is used.

[0062] The addition unit 2133 adds the element information and the composition information together to generate information in which the composition information is added to the element information. As an example, the second conversion unit 2132 converts element information of Mn (manganese) into information of "Mn2" by adding the composition ratio of an inorganic compound.

[0063] The vector conversion unit 2134 vectorizes the input first structure information 312 and further converts it into a fixed-length vector. Specifically, it converts the one-hot vectorized information of the space group number, which is the first structure information 312, into a fixed-length vector. The space group number is a number assigned to a specific space group, and the space group number itself has no meaning. However, the number indicating this space group number is converted into a similar arrangement for information indicating similar space groups, and into a completely different arrangement for information indicating dissimilar space groups.

[0064] The scaling unit 2135 performs a process of standardizing the input second structure information 313 and third structure information 314. While the first structure information 312 indicating the space group number described above simply indicates numerical values, the lengths of the axes of the unit cell, which is the second structure information 313, and the angles between the edges of the unit cell, which is the third structure information 314, are each a one-dimensional array. As an example, the angles between the edges of the unit cell in the third structure information 314 are represented as a one-dimensional array such as [90, 90, 90]. Therefore, unlike the first structure information 312, the second structure information 313 and the third structure information 314 are subjected to a standardization process by the scaling unit 2135.

[0065] The structural information integration unit 2136 integrates the first structural information after the vector conversion process with the standardized second structural information and the standardized third structural information.

[0066] The second NN unit 2137 converts the number of dimensions of the structural information integrated in the structural information integration unit 2136 into a predetermined number of dimensions through a neural network. The neural network used in the second NN unit 2137 may be, for example, a neural network that does not have an intermediate layer between the input layer and the output layer, and the input layer and the output layer may be weighted by a predetermined weight parameter W2. In other words, the weight W2 is a parameter that is learned during model learning.

[0067] FIG. 5 is a diagram illustrating the processing performed by the second NN unit 2137 based on input of the first structural information 312, the second structural information 313, and the third structural information 314. FIG. 5 illustrates, as an example, the case where information on the inorganic compound "MnO2" is input. For example, the space group number "166" input as the first structural information 312 is converted by the vector conversion unit 2134 into a predetermined vector (first structural information m1) and then into a fixed-length vector (first structural information m2). Furthermore, the lengths of the axes of the unit cell input as the second structural information 313 ([2.93, 2.93, 2.93]) and the angles between the edges of the unit cell input as the third structural information 314 ([77.0, 77.0, 60.0]) are standardized by the scaling unit 2135. The standardized second structural information and the third structural information may be combined to generate information indicating a predetermined sequence (structural information m3). The first structural information m2 after vector conversion and the structural information m3 after standardization are integrated in a structural information integration unit 2136. In addition, the integrated structural information m4 becomes information of a predetermined number of dimensions (structural information m5) in a second NN unit 2137.

[0068] Returning to FIG. 4, the integrated structural information in the second NN unit 2137 is further integrated in an integrating unit 2138 into information in which element information and composition information are added together.

[0069] 6 is a diagram schematically illustrating the process of integrating element information m6 converted in the first NN unit 2131, composition information m7 converted in the second conversion unit 2132, and structural information m5 integrated in the structural information integration unit 2136. For example, the converted element information m6 and the converted composition information m7 are added in the addition unit 2133 to become element information m8. The integration unit 2138 further integrates the element information m8 with structural information m5, which is an integration of the first to third structural information, to become integrated information m9. The integrated information m9 is input to the relationship learning unit 2139, which will be described next.

[0070] Returning to FIG. 4, the relationship learning unit 2139 learns the mutual relationships between each element and the structural information by taking the inner product between each element and the structural information for the integrated information m9, which is an integration of element information, composition information, and structural information. The processing performed by the relationship learning unit 2139 clarifies which elements affect the physical property values ​​of inorganic compounds. The relationship learning unit 2139 may use a known attention mechanism. For example, multi-head attention is used.

[0071] The third NN unit 2140 converts the information processed by the relationship learning unit 2139 into information with a predetermined number of dimensions through a neural network.

[0072] The calculation unit 2141 calculates the contribution of each element by applying a sigmoid function to the information output by the third NN unit 2140. The calculation unit 2141 also calculates the average of the contributions of all elements to calculate the physical property value and outputs the output information 41.

[0073] (Creation Section 14) The creation unit 14 creates display information 42 based on the output information 41 predicted and output by the prediction unit 13. Here, the display information 42 is not particularly limited, and may be, for example, an image, a graph, a numerical value, a list, or the like based on the result predicted by the prediction unit 13. The display information 42 may also be a GUI (Graphical User Interface screen) to be presented to the user. The creation unit 14 may also acquire data necessary to create a predetermined GUI from, for example, the second inorganic compound database 211. Specific examples of the display information 42 created by the creation unit 14 will be described below.

[0074] The creation unit 14 may create, as the display information 42, a heat map showing the distribution of the third contribution degree, with the first structural information 312 and the number of electrons affecting the physical property value of the first inorganic compound as two axes. The creation unit 14 may use structural information that has as input information the second structural information 313 and the third structural information 314 in addition to the first structural information 312. The number of electrons affecting the physical property value of the first inorganic compound is not particularly limited, and is, for example, the number of outermost electrons of at least one element included in a predetermined period in the first inorganic compound.

[0075] FIG. 7 is an example of a heat map created by the creation unit 14. In the heat map 141 of FIG. 7, the horizontal axis indicates the space group (space group number) and the vertical axis indicates the number of outermost electrons. There are 230 types of space groups, and the number of outermost electrons ranges from 0 to 12. The third contribution corresponding to each space group and each number of outermost electrons is indicated by a color gradient. The third contribution ranges from 0 to 200, and in the heat map 141, areas with relatively lighter colors indicate higher third contributions. It can be seen from the heat map 141 that the third contributions of elements of inorganic compounds classified into space group numbers 160 and 166 are relatively high. In this way, the creation unit 14 creates a heat map showing the distribution of the third contributions corresponding to the space group and the number of outermost electrons, allowing the user to visually determine the magnitude of the third contribution via, for example, the display device 2. Specifically, the user can visually identify the number of outermost electrons required for an element to have a high third contribution, and the space group to which the element belongs to have a high third contribution.

[0076] Furthermore, the creation unit 14 may further create, as the display information 42, a diagram showing the correlation between the number of electrons that affect the physical property values ​​of the inorganic compound and the third contribution in the first structure information 312, for an area with a high third contribution in the heat map created by the creation unit 14. Specifically, the diagram showing the correlation between the number of electrons and the third contribution may be a graph.

[0077] Fig. 8 is an example of a graph showing the correlation between the number of electrons and the third contribution. Graph 142 in Fig. 8 shows, as an example, the correlation between the number of electrons and the third contribution for data belonging to space group number 166, which is a region with a high third contribution from heat map 141 in Fig. 7. Graph 142 shows the third contribution on the vertical axis and the correlation with the number of electrons in space group number 166 on the horizontal axis. Graph 142 may also show the number of data extracted from second inorganic compound database 211 in addition to the correlation between the third contribution and the number of electrons.

[0078] The number of data is the number of partial data extracted from the second inorganic compound database 211. More specifically, the number of data is the number of compounds that have a structure with space group number 166 and have an element with 0 to 10 outermost electrons among the inorganic compounds included in the second inorganic compound database 211. That is, according to the graph 142, the user can know how many compounds that have an element with a high third contribution are included in the second inorganic compound database 211.

[0079] Furthermore, the creation unit 14 may add an approximation curve based on the plot of the third contribution to the correlation diagram between the number of electrons and the third contribution. In FIG. 8, the curve indicated by the dotted line is the approximation curve. This provides display information that allows the user to easily understand, for example, the number of electrons that maximizes the third contribution from the approximation curve. Furthermore, the creation unit 14 may extract, from the second inorganic compound database 211, the number of electrons corresponding to the point showing the maximum value on the approximation curve, i.e., the compound having the number of electrons that is estimated to maximize the third contribution.

[0080] In addition, the creation unit 14 may further create, as the display information 42, a list of inorganic compounds having first structure information 312 indicating the third contribution and the number of electrons that affect the physical property values ​​of the inorganic compound.

[0081] 9 is an example of a list of inorganic compounds having first structural information (space group) 312 indicating the third contribution and the number of electrons (number of outermost electrons) that affect the physical properties of the inorganic compound. List 143 in FIG. 9 is a list displaying inorganic compounds extracted from second inorganic compound database 211, which contains data on multiple inorganic compounds, and containing an element with space group number 160 and the number of outermost electrons being 5. ID is an identification number assigned to each extracted inorganic compound.

[0082] As an example, the list 143 includes inorganic compounds with the same ID and the same composition (No. 3 and No. 4). These indicate that the inorganic compounds have the same composition but different target elements. That is, No. 3 indicates the third contribution of Fe (iron), and No. 4 indicates the third contribution of Co (cobalt). This allows the user to visually identify inorganic compounds with high third contributions 413 and the elements contained in the inorganic compounds.

[0083] The display information 42 created by the creation unit 14 may be output to an external display device 2 and displayed on the display device 2, or, if the prediction device 1 is equipped with a display unit (not shown), may be output to the display unit of the prediction device 1. This allows the user to visually identify the result predicted by the prediction device 1 by looking at the display information 42 created by the creation unit 14.

[0084] [Flow of prediction process in prediction device 1] Next, the flow of the prediction process in the prediction device 1 will be described with reference to Fig. 10. Fig. 10 is a flowchart showing an example of the flow of the prediction process.

[0085] First, the input acquisition unit 11 acquires input information 31 of a first inorganic compound whose specific physical property value is unknown (Step S1: input step). The input information 31 includes element information indicating the elements of the inorganic compound, composition information indicating the ratio of the elements, and first structure information 312 indicating the unit cell, translational property, and symmetry of the crystalline structure of the inorganic compound.

[0086] The prediction unit 13 predicts a specific physical property value of the first inorganic compound from the input information 31 using the trained model. Specifically, the prediction unit 13 predicts a first contribution indicating the degree of influence of each element contained in the inorganic compound on the physical property value of the first inorganic compound from the converted input information (step S2). Then, the prediction unit 13 predicts a second contribution indicating the degree of influence on the physical property value of the first inorganic compound influenced by the structure indicated by the structural information (step S3). The prediction unit 13 predicts the physical property value of the first inorganic compound based on the first contribution and the second contribution (step S4). The trained model has undergone a learning process to determine the relationship between the element information, composition information, and first structural information 312 of the second inorganic compound, whose physical property values ​​are known, and the physical property values.

[0087] In addition, the prediction unit 13 calculates a third contribution indicating the degree of influence of at least one element on the physical property values ​​of the first inorganic compound from the first contribution and second contribution of at least one element included in a predetermined period among the elements included in the first inorganic compound (step S5: calculation step).

[0088] The creation unit 14 creates an image or a list as the display information 42 related to the third degree of contribution calculated by the prediction unit 13 (step S6: creation step).

[0089] The creation unit 14 outputs the created display information 42 to, for example, the display device 2, which is an external device of the prediction device 1 (step S7: output step), and the process ends.

[0090] Thus, the prediction process of the prediction device 1 includes an input acquisition step of acquiring input information 31, and a prediction step of predicting a specific physical property value of the first inorganic compound from the input information using a trained model.

[0091] By employing this prediction method, the prediction device 1 can accurately predict specific physical property values ​​of the first inorganic compound. Furthermore, by inputting first structure information 312 indicating the unit cell, translational property, and symmetry of the crystal structure of the first inorganic compound as input information 31, the prediction device 1 can accurately predict physical property values ​​even for first inorganic compounds whose crystal structure dominates the physical property values.

[0092] Furthermore, by employing information indicating the unit cell, translational property, and symmetry of the crystal structure of the first inorganic compound as the first structural information 312, the user can easily input structural information without having to set the spatial arrangement of each atom of the first inorganic compound in detail when inputting information. This also has the effect of making it easier for the user to prepare a data set of structural information when predicting the physical properties of not only known compounds but also unknown compounds.

[0093] In the prediction process of this embodiment, the prediction unit 13 predicts the physical property values, and then further calculates the third contribution, and the creation unit 14 creates display information, but this is not limited to this form, and the prediction unit 13 may predict only the physical property values ​​and output the results to the display device 2.

[0094] Furthermore, the prediction result predicted by the prediction unit 13 may be transmitted to an external device via the communication unit 30 without being sent to the creation unit 14. In this case, the above-described display information 42 may be created in the external device.

[0095] <Modification> Modifications of the present disclosure will be described below. For convenience of explanation, the same reference numerals will be used to designate components having the same functions as those described in the above embodiment, and the description thereof will not be repeated.

[0096] In the first embodiment, an example was described in which the input acquisition unit 11 acquires element information 310 indicating the elements of the inorganic compound, composition information 311 indicating the ratio of the elements, and first structure information 312 from the first inorganic compound database 210. In this modified example, an example will be described in which an optimal condition prediction unit 17 is provided to predict an optimal first inorganic compound. The optimal condition prediction unit 17 can predict at least one optimal condition of the first inorganic compound. In this modified example, the first inorganic compound can be expressed, for example, by a formula including one or more variables, and the optimal condition prediction unit 17 predicts the optimal condition of the variables. Note that the predicted optimal condition may be structure information. The input acquisition unit 11 acquires conditions predicted as optimal conditions as input information 31 from the optimal condition prediction unit 17. In this modified example, a specific example will be described in which the variables are element information 310 and composition information 311.

[0097] [Configuration of prediction device 1A] FIG. 11 is a block diagram showing an example of the configuration of a prediction device 1A according to this modification.

[0098] As shown in FIG. 11, the prediction device 1A includes a control unit 10A that controls all the components of the prediction device 1A, a memory unit 20A that stores various data used by the control unit 10A, and a communication unit 30 that communicates with external devices.

[0099] The storage unit 20A stores a trained model 213, a first inorganic compound database 210, a second inorganic compound database 211, a statistical model 214, and a search space database 220.

[0100] The statistical model 214 is a model that indicates the distribution of probabilities regarding what values ​​will be output when the prediction unit 13 predicts physical property values ​​using the trained model 213. The statistical model 214 may be updated by the prediction unit 13.

[0101] The search space database 220 stores a plurality of combinations of the type of additive metal and the amount of additive metal used by the search space determination unit 16 (described later) to determine the search space.

[0102] In this modification, the prediction device 1A includes the first inorganic compound database 210, the second inorganic compound database 211, the statistical model 214, and the search space database 220. However, the present invention is not limited to this. For example, the first inorganic compound database 210, the second inorganic compound database 211, the statistical model 214, and the search space database 220 may be stored on an external server, and the prediction device 1A may access each database stored on the server. In this case, the first inorganic compound database 210, the second inorganic compound database 211, the statistical model 214, and the search space database 220 may be stored on the same server, or each database may be stored on a different server. Furthermore, the storage unit 20A may store a control program for performing various controls of the prediction device 1A.

[0103] <Control unit 10A> The control unit 10A includes a search space determination unit 16, an optimum condition prediction unit 17, an input acquisition unit 11, a prediction unit 13A, a creation unit 14, and a learning unit 15. Each component included in the control unit 10A will be described below.

[0104] (Search space determination unit 16) The search space determination unit 16 determines a search space used by the optimal condition prediction unit 17 to predict an optimal first inorganic compound. More specifically, the search space determination unit 16 determines a search space used by the optimal condition prediction unit 17 to predict optimal conditions for variables in a formula representing the first inorganic compound. As a specific example, the search space determination unit 16 determines a search space used by the optimal condition prediction unit 17 to predict optimal conditions for element information 310 and composition information 311. That is, the search space determination unit 16 identifies candidate element types and ranges of composition ratios to define a search space used by the optimal condition prediction unit 17 to predict the element information 310 and composition information 311. As an example, if the element information 310 includes "FeCo," which is known as a promising magnetic material among magnetic materials, the first inorganic compound can be expressed by the formula M_xFe_(1-x / 2)Co_(1-x / 2), where the type of metal (M) added to FeCO and the amount of metal added (x) are variables. The search space determination unit 16 determines candidates for the type of metal (M) to be added to FeCO and the range of the amount of metal (x) added. The search space determination unit 16 may extract candidates for the type of added metal and the range of the amount of metal added from the search space database 220. As an example, the search space determination unit 16 determines the search space as follows: added metal M: [Sc-Au transition metal element], added amount x = [0-1] (M_xFe_(1-x / 2)Co_(1-x / 2)). The added amount x is a factor that determines the composition ratio of the composition formula of the inorganic compound.

[0105] The first structure information 312 may be input in advance to the search space determination unit 16. That is, the search space determination unit 16 may determine the search space in a state in which the first structure information 312 has been taken into consideration in advance. Alternatively, the first structure information 312 may be input to the input acquisition unit 11 from the first inorganic compound database 210 after the optimum condition prediction unit 17, which will be described later, predicts the optimum conditions for the element information 310 and the composition information 311.

[0106] (Optimal condition prediction unit 17) The optimal condition prediction unit 17 predicts the optimal first inorganic compound. More specifically, the optimal condition prediction unit 17 predicts the optimal conditions of variables in a formula representing the first inorganic compound from the search space defined by the search space determination unit 16. As a specific example, the optimal condition prediction unit 17 predicts the optimal conditions of element information 310 and composition information 311. The search space determined by the search space determination unit 16 is input to the optimal condition prediction unit 17. The optimal condition prediction unit 17 predicts the optimal conditions of the element information 310 and composition information 311 from within the search space based on the statistical model 214 stored in the memory unit 20A. Specifically, the optimal condition prediction unit 17 predicts the optimal conditions of the type M of the added metal and the amount x of the added metal described above.

[0107] The optimum condition prediction unit 17 predicts optimum conditions for the element information 310 and composition information 311 based on part of the input element information 310 and part of the composition information 311 using Bayesian optimization.

[0108] Bayesian optimization is an algorithm that uses a statistical model to represent the relationship between an input and a corresponding output, and determines the optimal input conditions to be considered next based on an acquisition function, which is an evaluation index derived from the statistical model. In this modification, the statistical model 214 indicates the probability distribution of what value will be output when the prediction unit 13 predicts a physical property value using the trained model 213, as described above.

[0109] In this way, the optimum condition prediction unit 17 predicts the optimum conditions for the type M of added metal and the amount x of added metal using Bayesian optimization. The element information 310 and the composition information 311 are uniquely determined.

[0110] (Input acquisition unit 11) The input acquisition unit 11 acquires, as input information 31, elemental information 310 and compositional information 311 based on the conditions predicted by the optimal condition prediction unit 17. As described above, the first structural information 312 may be acquired from the optimal condition prediction unit 17 together with the elemental information 310 and compositional information 311, or the first structural information 312 may be acquired separately from the first inorganic compound database 210.

[0111] (Prediction section 13A) Similar to the prediction unit 13 of the first embodiment, the prediction unit 13A predicts a specific physical property value of the inorganic compound from the input information 31 using the trained model 213. The prediction unit 13A may also update the statistical model 214 based on the predicted physical property value. Thereafter, using the updated statistical model 214, the optimal condition prediction unit 17 may further determine the type M of added metal and the amount x of added metal that are predicted as the optimal condition.

[0112] The series of processes performed by the optimal condition prediction unit 17, the input acquisition unit 11, and the prediction unit 13A may be repeated a predetermined number of times. The prediction unit 13A may determine the optimal conditions by comparing a plurality of physical property values ​​predicted by the repetition. Each time the optimization process by the optimal condition prediction unit 17, the input process by the input acquisition unit 11, and the prediction process by the prediction unit 13A are repeated, the statistical model 214 is updated by the prediction unit 13A. The predetermined number of times may be determined in advance.

[0113] (Creation Section 14) The creation unit 14 creates display information based on the output information predicted and output by the prediction unit 13 A. In this modification, the creation unit 14 may create display information every time the prediction unit 13 A outputs one physical property value, or may create display information after a series of processes from the optimization process to the prediction process have been performed a predetermined number of times.

[0114] FIG. 12 shows a graph 144 created by the creation unit 14, which shows the correlation between the type of metal (M) to be added and the amount of metal (x) added, along with the magnetic value. The magnetic value may be represented, for example, by different colors or shades of the plots. In the graph 144, for example, the darker the shade of the plot, the lower the magnetic value, and the lighter the shade of the plot, the higher the magnetic value. In this way, the creation unit 14 shows the correlation between the type of metal (M) and the amount of metal (x) added, along with the magnetic value, allowing the user to visually recognize the prediction result of how much of which metal needs to be added to obtain a highly magnetic material. Alternatively, the creation unit 14 may create display information 42 based on output information 41 output based on the optimal conditions determined by the prediction unit 13A.

[0115] In this way, by including the optimal condition prediction unit 17 in the prediction device 1A, the optimized elemental information 310 and compositional information 311 are input to the input acquisition unit 11. Furthermore, the prediction unit 13A predicts physical property values ​​based on the optimized elemental information 310, the optimized compositional information 311, and the first structure information 312. This reduces the number of times that the prediction process can be repeated until the prediction unit 13A outputs optimal physical property values ​​compared to when non-optimized elemental information 310 and compositional information 311 are input. In other words, the time required for prediction can be shortened.

[0116] In this modification, an example has been described in which the prediction device 1A is provided with the optimal condition prediction unit 17, but an optimal condition prediction device that predicts the optimal conditions for the elemental information 310 and the composition information 311 may be external to the prediction device 1 of embodiment 1. In this case, the prediction device 1 may acquire the elemental information 310 and the composition information 311 from the optimal condition prediction device. Alternatively, the storage unit of the optimal condition prediction device may be provided with a statistical model and a search space database, and the statistical model may be updated by acquiring the results predicted by the prediction device 1.

[0117] [Software implementation example] The functions of the prediction device 1 (hereinafter referred to as the "device") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device (particularly each part included in the control unit 10).

[0118] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program, thereby realizing the functions described in each of the above embodiments.

[0119] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0120] In addition, some or all of the functions of each of the control blocks can be realized by logic circuits. For example, integrated circuits in which logic circuits that function as each of the control blocks are formed are also included in the scope of the present disclosure. In addition, the functions of each of the control blocks can also be realized by, for example, a quantum computer.

[0121] 〔summary〕 A prediction device 1 according to aspect 1 of the present disclosure includes an input acquisition unit 11 that acquires input information 31 of a first inorganic compound whose specific physical property value is unknown, and a prediction unit 13 that predicts a specific physical property value 410 of the first inorganic compound from the input information 31 using a trained model 213. The trained model 213 has undergone a training process to learn the relationship between reference information 32 of a second inorganic compound whose physical property value is known and the physical property value. The input information 31 and the reference information 32 each have elemental information indicating the elements of the inorganic compound, compositional information indicating the ratio of the elements, and first structural information 312 indicating the unit cell, translational property, and symmetry of the crystal structure of the inorganic compound.

[0122] In the prediction device 1 according to aspect 2 of the present disclosure, in aspect 1, the prediction unit 13 may be capable of predicting a first contribution 411 indicating the degree of influence of each element contained in the first inorganic compound on the physical property value of the first inorganic compound from the input information 31.

[0123] In the prediction device 1 according to aspect 3 of the present disclosure, in aspect 2, the prediction unit 13 may be capable of predicting a second contribution 412 indicating the degree of influence on the physical property value of the first inorganic compound affected by the structure indicated by the first structural information 312 from the input information 31.

[0124] In the prediction device 1 according to aspect 4 of the present disclosure, in the above-described aspect 3, the prediction unit 13 may calculate the physical property value 410 of the first inorganic compound based on the first contribution 411 and the second contribution 412.

[0125] In the prediction device 1 according to aspect 5 of the present disclosure, in any one of aspects 1 to 4, the input information 31 may include second structure information 313 indicating the length of each axis of the unit cell.

[0126] In the prediction device 1 according to a sixth aspect of the present disclosure, in any one of the first to fifth aspects, the input information 31 may include third structural information 314 indicating angles between edges of the unit cell.

[0127] In the prediction device 1 according to aspect 7 of the present disclosure, in aspect 3 or 4, the prediction unit 13 may calculate a third contribution 413 indicating the degree of influence of at least one element included in a predetermined period on the physical property value of the first inorganic compound from the first contribution 411 and second contribution 412 of at least one element included in a predetermined period among the elements included in the first inorganic compound.

[0128] The prediction device 1A according to an eighth aspect of the present disclosure may further include an optimal condition prediction unit 17 that predicts an optimal first inorganic compound in any one of the first to seventh aspects.

[0129] The prediction device 1 according to aspect 9 of the present disclosure may further include a creation unit 14 that creates a heat map 141 showing the distribution of the third contribution, with the first structural information 312 and the number of electrons affecting the physical property value 410 of the first inorganic compound as the two axes, in the seventh aspect.

[0130] In the prediction device 1 according to aspect 10 of the present disclosure, in aspect 9, the creation unit 14 may further create a graph showing the correlation between the number of electrons in the first structural information 312 and the third contribution 413 for regions in the heat map 141 where the third contribution 413 is high.

[0131] In the prediction device 1 according to aspect 11 of the present disclosure, in aspect 9 or 10, the creation unit 14 may further create a list 143 of inorganic compounds having first structural information 312 indicating a predetermined third contribution 413 and the number of electrons.

[0132] A prediction method according to aspect 12 of the present disclosure includes an input acquisition step (S1) of acquiring input information 31 of a first inorganic compound whose specific physical property value is unknown, and a prediction step (S4) of predicting a specific physical property value 410 of the first inorganic compound from the input information 31 using a trained model 213, wherein the trained model has been trained to learn the relationship between reference information 32 of a second inorganic compound whose specific physical property value is known and the physical property value, and the input information 31 and the reference information 32 each have elemental information indicating the elements of the inorganic compound, compositional information indicating the ratio of the elements, and first structural information 312 indicating the unit cell and translational and symmetry properties of the crystal structure of the inorganic compound. [Example]

[0133] An example of the present disclosure is described below. As an example, a prediction device 1 was used, and input information was (1) elemental information, (2) composition information, and (3) first structural information indicating a unit cell, translational properties, and symmetry. The physical property to be predicted is magnetic moment. Furthermore, the dataset used for training, verifying, and testing the trained model 213 was a set of data including (1) elemental information, (2) composition information, (3) information indicating a unit cell, translational properties, and symmetry, and (4) magnetic moment. Figures 13 and 14 are graphs of the example.

[0134] In addition, as a comparative example, only (1) element information and (2) composition information were used as input information, and data sets for training, verifying, and testing the trained model were data sets containing (1) element information, (2) composition information, and (4) magnetic moment. Figures 15 and 16 are graphs of the example.

[0135] The example differs from the comparative example in that the input information used is first structural information indicating a unit cell and translational and symmetry properties.

[0136] In both the example and the comparative example, the prediction device 1 used 35,000 data sets, which were randomly divided into training data, validation data, and test data, with 70% used for training, 15% for evaluation, and 15% for testing. The training data was used to instruct the trained model 213 to learn, the validation data was used to verify the accuracy of the trained model 213 in the learning process, and the test data was used to verify the accuracy of the final trained model 213 constructed. The trained model 213 was created using the training data. The data shown as "Training data" in FIGS. 13 and 15 is the training data, and the data shown as "Test data" in FIGS. 14 and 16 is the test data. In all of FIGS. 13 to 16, predicted values ​​for actual measured values ​​of magnetic moment are plotted in the range of 1 to 250 μm. B / nm 3This is a graph plotted in the range of

[0137] 13 to 16 show the true regression line obtained when the predicted value of the magnetic moment and the calculated value of the magnetic moment agree with each other. 2 indicates the correlation with the true regression line. 2 The closer to 1, the closer it is to the true regression line, i.e., the higher the prediction accuracy of the prediction device. The value indicated by mse indicates mean squared error (mse), which is obtained by calculating the square of the difference between the predicted value and the actual measured value for each data, and then dividing the sum by the number of data. The value indicated by mae indicates mean absolute error (mae), which is obtained by calculating the absolute value of the difference between the predicted value and the actual measured value for each data, and then dividing the sum by the number of data. For both mse and mae, the smaller the value, the smaller the error between the predicted value and the actual measured value. In other words, the smaller the values ​​of mse and mae, the higher the prediction accuracy of the prediction device.

[0138] <Result> (1) Example In the training data of the example, the coefficient of determination R 2 The coefficient of determination (R) was 0.85, the MSE was 146.78, and the MAE was 7.54 (Fig. 13). 2 was 0.75, MSE was 258.93, and MAE was 9.26 (Figure 14).

[0139] (2) Comparative Example In the training data of the comparative example, the coefficient of determination R 2 The coefficient of determination (R) was 0.61, the MSE was 383.49, and the MAE was 10.99 (Fig. 15). 2 was 0.53, MSE was 454.49, and MAE was 12.37 (Figure 16).

[0140] Coefficient of determination R 2From the values, it was found that the predicted values ​​predicted by the prediction device 1 of the embodiment show a good correlation with the actual measured values. In addition, for test data that is not included in the training data, the physical properties can be predicted with the same accuracy as the training data, demonstrating that the prediction device 1 has good performance.

[0141] The invention according to the present disclosure has been described above based on the drawings and examples. However, the invention according to the present disclosure is not limited to the above-described embodiments. In other words, the invention according to the present disclosure can be modified in various ways within the scope of the present disclosure, and embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the invention according to the present disclosure. In other words, it should be noted that a person skilled in the art can easily make various modifications or corrections based on the present disclosure. It should also be noted that these modifications or corrections are included in the scope of the present disclosure. [Explanation of symbols]

[0142] 1 Prediction device 11 Input acquisition unit 13 Prediction Department 14 Creation Department 16 Search space determination unit 17 Optimum condition prediction section 31 Input information 32 References 141 Heatmap 143 List 213 trained models 310 Element information 311 Composition information 312 First structure information 313 Second structure information 314 Third structure information 410 Physical properties 411 First Contribution 412 Second Contribution

Claims

1. an input acquisition unit that acquires input information of a first inorganic compound whose specific physical property value is unknown; a prediction unit that predicts the physical property value of the first inorganic compound from the input information using a trained model; The trained model has undergone a training process to learn a relationship between reference information of a second inorganic compound whose physical property values ​​are known and the physical property values, A prediction device, wherein the input information and the reference information each have element information indicating the elements of the inorganic compound, composition information indicating the ratio of the elements, and first structure information indicating the unit lattice and translational and symmetry properties of the crystal structure of the inorganic compound.

2. The prediction device according to claim 1 , wherein the prediction unit is capable of predicting a first contribution indicating a degree of influence of each element contained in the first inorganic compound on a physical property value of the first inorganic compound from the input information.

3. The prediction device according to claim 2 , wherein the prediction unit is capable of predicting a second contribution indicating a degree of influence on the physical property value of the first inorganic compound that is influenced by the structure indicated by the first structural information from the input information.

4. The prediction device according to claim 3 , wherein the prediction unit calculates the physical property value of the first inorganic compound based on the first contribution and the second contribution.

5. The prediction device according to claim 1 , wherein the input information includes second structural information indicating the lengths of the axes of the unit cell.

6. The prediction device according to claim 1 , wherein the input information includes third structural information indicating angles between edges of the unit cell.

7. The prediction device described in claim 3, wherein the prediction unit calculates a third contribution indicating the degree of influence of at least one element included in a predetermined period on the physical property value of the first inorganic compound from the first contribution and the second contribution of at least one element included in the predetermined period among the elements included in the first inorganic compound.

8. The prediction device according to claim 1 , further comprising an optimal condition prediction unit that predicts an optimal first inorganic compound.

9. The prediction device according to claim 7, further comprising a creation unit that creates a heat map showing the distribution of the third contribution, with the first structural information and the number of electrons that affect the physical property value of the first inorganic compound as two axes.

10. The prediction device according to claim 9 , wherein the creation unit further creates a diagram showing a correlation between the number of electrons in the first structural information and the third contribution for a region in the heat map where the third contribution is high.

11. The prediction device according to claim 9 , wherein the creation unit further creates a list of inorganic compounds having the first structural information indicating the predetermined third contribution and the number of electrons.

12. an input acquisition step of acquiring input information of a first inorganic compound whose specific physical property value is unknown; a prediction step of predicting the physical property value of the first inorganic compound from the input information using a trained model, The trained model has undergone a training process to learn a relationship between reference information of a second inorganic compound whose physical property values ​​are known and the physical property values, A prediction method in which the input information and the reference information each have elemental information indicating the elements of the inorganic compound, compositional information indicating the ratio of the elements, and first structural information indicating the unit cell and translational and symmetry properties of the crystal structure of the inorganic compound.

Images