Crystal analysis method, crystal analysis device, and crystal analysis program that can be suitably applied to quantum computing technology
By treating anionic atoms as part of cationic atoms in the graph creation process, the method addresses the challenge of high bit requirements in ionic crystal analysis, enabling accurate similarity calculations through reduced node representation and efficient computational methods.
Patent Information
- Application Number
- JP2020213847
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2020-12-23
- Publication Date
- 2024-11-13
- Estimated Expiration
- 2040-12-23
AI Technical Summary
Existing methods for analyzing ionic crystals face challenges in accurately determining similarity due to the large number of atoms in the unit cell, which increases the number of bits required for calculation, making it difficult to obtain precise results.
A crystal analysis method that reduces the number of nodes in the lattice by treating anionic atoms as part of cationic atoms in the graph creation process, allowing for the creation of a graph that accurately represents the repeating structure of ionic crystals, and uses this graph to analyze similarity by solving the maximum independent set problem.
This approach effectively reduces the number of bits required for analysis, enabling accurate similarity calculations between ionic crystals by minimizing the number of nodes in the lattice and improving computational efficiency.
Smart Images

Figure 0007581845000025 
Figure 0007581845000026 
Figure 0007581845000027
Abstract
Description
[Technical field]
[0001] The present invention relates to a crystal analysis method, a crystal analysis apparatus, and a crystal analysis program that can be suitably applied to quantum computing technology. [Background technology]
[0002] In recent years, computer-based analysis of materials has been used to design new materials, predict the properties of existing materials, and find new uses for existing materials.
[0003] More specifically, for example, when calculating the similarity between molecules or between crystals having a repeating structure, a "conflict graph" that represents the combination of atoms in the structure for which the similarity is to be calculated can be used. As a method of using the conflict graph, a technology has been proposed in which a search for a substructure common to each structure is performed by solving the maximum independent set problem of the conflict graph using an annealing machine or the like to specify the similarity (see, for example, Non-Patent Document 1).
[0004] Here, when calculating the similarity between crystals having a repeating structure, the number of atoms contained in each unit cell for comparing the structures may be large (the number of bits required to calculate the similarity may be large). In particular, when the crystal is an ionic crystal, the unit cell contains many anions (e.g., oxygen atoms), and the number of atoms contained in the unit cell is large. In addition, when calculating similarity using an annealing machine, since the annealing machine has a limit to the number of bits that can be used in the calculation, if the unit cell contains a large number of atoms (the number of bits required for the calculation is large), it can be difficult to obtain an accurate similarity. [Prior art documents] [Non-patent literature]
[0005] [Non-Patent Document 1] Maritza Hernandez, Arman Zaribafiyan, Maliheh Aramon, Mohammad Naghibi “A Novel Graph-based Approach for Determining Molecular Similarity”. arXiv:1601.06693(https: / / arxiv.org / pdf / 1601.06693.pdf) Summary of the Invention [Problem to be solved by the invention]
[0006] An object of the present invention is to provide a crystal analysis method, a crystal analysis device, and a crystal analysis program that can reduce (suppress) the number of bits required when analyzing ionic crystals. [Means for solving the problem]
[0007] One embodiment of the means for solving the above problems is as follows. That is, in one embodiment, the crystallography method comprises: a graph creation step of creating a graph in which a repeating unit cell including an intra-lattice node indicating data of an atom in the repeating unit cell and an intra-lattice edge indicating data of a chemical bond between two atoms in the repeating unit cell is expanded to the repeating unit cell and an adjacent repeating unit cell adjacent to the repeating unit cell; an analysis step of analyzing the crystal using the graph; Including, In the graph creation step, when the crystal is an ionic crystal, when the number of intra-lattice nodes which are data on anionic atoms bonded to cationic atoms in one repeating unit lattice in the ionic crystal is n, the number of intra-lattice nodes which are data on anionic atoms is set to n-1 or less. Effect of the Invention
[0008] In one aspect, the present invention provides a crystal analysis method, a crystal analysis device, and a crystal analysis program that can reduce (suppress) the number of bits required when analyzing an ionic crystal. [Brief description of the drawings]
[0009] [Figure 1] FIG. 1 is a diagram showing an example of how acetic acid and methyl acetate are graphed. [Diagram 2] FIG. 2 is a diagram showing an example of a combination in which the same elements in molecules A and B are combined to form nodes in a conflict graph. [Diagram 3] FIG. 3 is a diagram showing an example of rules for creating edges in a conflict graph. [Figure 4] FIG. 4 is a diagram showing an example of a conflict graph between molecule A and molecule B. [Diagram 5] FIG. 5 is a diagram illustrating an example of a maximal independent set in a graph. [Figure 6] FIG. 6 is a diagram showing an example of a flow in determining the maximum common substructure between molecules A and B by determining a maximum independent set in a conflict graph (by solving a maximum independent set problem). [Figure 7] FIG. 7 is a diagram illustrating an example of a technique for searching for a maximum independent set in a graph having six nodes. [Figure 8] FIG. 8 is an explanatory diagram illustrating an example of a technique for searching for a maximum independent set in a graph having six nodes. [Figure 9] FIG. 9 is a diagram showing an example of a maximum independent set in a conflict graph. [Figure 10] FIG. 10 is a schematic diagram showing an example of a periodic structure in a crystal (crystal material) having a repeating structure. [Figure 11] FIG. 11 is a diagram showing an example of a unit lattice in a method of expanding and handling a unit lattice by using an expansion node. [Figure 12]FIG. 12 is a diagram showing an example of a unit cell in a method of treating a repeating structure of a crystal as a loop structure. [Figure 13A] FIG. 13A is a diagram showing an example of a graph of a unit cell of a crystal. [Figure 13B] FIG. 13B is a diagram showing another example of a graph of a unit cell of a crystal. [Figure 14A] FIG. 14A is a diagram for explaining an example of a technique for integrally multiplying unit cell A and unit cell B. FIG. [Figure 14B] FIG. 14B is a diagram for explaining an example of a technique for integrally multiplying unit cell A and unit cell B. In FIG. [Figure 14C] FIG. 14C is a diagram for explaining an example of a technique for integrally multiplying unit cell A and unit cell B. In FIG. [Figure 15A] FIG. 15A is a diagram showing an example of a unit cell in an ionic crystal. [Figure 15B] FIG. 15B is a diagram showing another example of a unit cell in an ionic crystal. [Figure 16] FIG. 16 is a diagram showing a comparison between an example of the prior art and an embodiment of the present invention. [Figure 17] FIG. 17 is a diagram showing an example of the relationship between the combination of nodes in the first graph and the second graph and c'ij which represents the weight related to the bond order between cationic atoms. [Figure 18] FIG. 18 is a diagram showing an example of the hardware configuration of the crystal analysis apparatus disclosed in the present application. [Figure 19] FIG. 19 is a diagram showing another example of the hardware configuration of the crystal analysis apparatus disclosed in the present application. [Figure 20] FIG. 20 is a diagram showing an example of the functional configuration of the crystal analysis apparatus disclosed in the present application. [Figure 21] FIG. 21 is a diagram showing an example of a flowchart for analyzing ionic crystals using an example of the technology disclosed in the present application. [Figure 22] FIG. 22 is a diagram showing a part of an example of a flowchart for creating a graph by a method of treating a repeating crystal structure as a loop structure, using an example of the technology disclosed in the present application. [Diagram 23] FIG. 23 is a diagram showing a part of an example of a flowchart for creating a graph by a method of treating a repeating crystal structure as a loop structure, using an example of the technology disclosed in the present application. [Figure 24] FIG. 24 is a diagram showing a part of an example of a flowchart for creating a graph by a method of treating a repeating crystal structure as a loop structure, using an example of the technology disclosed in the present application. [Diagram 25] FIG. 25 is a diagram showing a part of an example of a flowchart for creating a graph by a method of treating a repeating crystal structure as a loop structure, using an example of the technology disclosed in the present application. [Figure 26] FIG. 26 is a diagram showing a part of an example of a flowchart for creating a graph by a method of treating a repeating crystal structure as a loop structure, using an example of the technology disclosed in the present application. [Figure 27] FIG. 27 is a diagram showing a part of an example of a flowchart for creating a graph by a method for expanding and handling a unit lattice using an expansion node, using an example of the technology disclosed in the present application. [Figure 28] FIG. 28 is a diagram showing a part of an example of a flowchart for creating a graph by a method for expanding and handling a unit lattice using an expansion node, using an example of the technology disclosed in the present application. [Figure 29] FIG. 29 is a diagram showing a part of an example of a flowchart for creating a graph by a method for expanding and handling a unit lattice using an expansion node, using an example of the technology disclosed in the present application. [Diagram 30] FIG. 30 is a diagram illustrating an example of a functional configuration of an annealing machine that performs the annealing method. [Diagram 31] FIG. 31 is a diagram showing an example of an operational flow of the transition control unit 114. [Diagram 32] FIG. 32 is a diagram showing an example in which the results of Example 1 and Comparative Example 1 are summarized. [Diagram 33] FIG. 33 is a diagram showing an example of a summary of the results of the second embodiment. [Diagram 34]FIG. 34 is a diagram showing an example of the relationship between the type of ionic crystal and the calculated similarity in Example 3 and Comparative Example 2. [Diagram 35] FIG. 35 is a diagram showing an example of the relationship between the type of ionic crystal and the number of bits required for calculating the similarity in Example 3 and Comparative Example 2. [Diagram 36] FIG. 36 is a diagram showing an example of the relationship between the type of ionic crystal and the calculated similarity in Example 4. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0010] (Crystal analysis method) The technology disclosed in this application is based on the inventor's knowledge that, in the conventional technology, when analyzing an ionic crystal, if the unit lattice in the ionic crystal contains a large number of atoms, the number of bits required for calculation increases, making accurate analysis difficult. Therefore, before describing the details of the technology disclosed in this application, the problems of the conventional technology will be described.
[0011] As described above, when performing an analysis to calculate the similarity between molecules or between crystals having a repeating structure, for example, a technique using a "conflict graph" that represents the combination of atoms in the structure for which the similarity is to be calculated can be used. In this technique, for example, the maximum independent set is searched for in the conflict graph (the maximum independent set problem is solved) to identify a substructure that is common to each structure, thereby obtaining the similarity. Here, when calculating the structural similarity between compounds by solving the maximum independent set problem of a conflict graph, the compounds are graphed for handling. Here, graphing a compound means, for example, expressing the structure of the compound using information on the types of atoms (elements) in the compound and information on the bonding state of each atom.
[0012] In the following, a method for creating a conflict graph will be described, firstly by taking as an example a case where a conflict graph between acetic acid (CH3COOH) and methyl acetate (CH3COOCH3) is created as an example for finding the similarity between molecules. First, when acetic acid (hereinafter sometimes referred to as "molecule A") and methyl acetate (hereinafter sometimes referred to as "molecule B") are graphed, the result is as shown in Figure 1. In Figure 1, the atoms that form acetic acid are indicated as A1, A2, A3, and A5, and the atoms that form methyl acetate are indicated as B1 to B5. Also, in Figure 1, A1, A2, B1, B2, and B4 represent carbon, A3, A5, B3, and B5 represent oxygen, and single bonds are indicated with thin solid lines and double bonds with thick solid lines.
[0013] Next, vertices (atoms) in the graphed molecules A and B are combined to create vertices (nodes) of a conflict graph. At this time, it is preferable to combine the same elements in molecules A and B to create nodes of the conflict graph, for example, as shown in Fig. 2. In the example shown in Fig. 2, the combination of A1, A2, B1, B2, and B4 representing carbon, and the combination of A3, A5, B3, and B5 representing oxygen are set as nodes of the conflict graph.
[0014] Next, edges (branches, sides) in the conflict graph are created. At this time, two nodes are compared, and if the nodes are composed of atoms in different situations (for example, atomic numbers, the presence or absence of bonds, bond orders, etc.), an edge is created between the two nodes. On the other hand, if the two nodes are compared, and the nodes are composed of atoms in the same situation, no edge is created between the two nodes.
[0015] Now, with reference to FIG. 3, the rules for creating edges in a conflict graph will be described. First, in the example shown in Figure 3, we will explain whether to create an edge between node [A1B1] and node [A2B2]. As can be seen from the graphed structure of molecule A in Figure 3, carbon A1 of molecule A included in node [A1B1] and carbon A2 of molecule A included in node [A2B2] are bonded to each other (single bond). Similarly, carbon B1 of molecule B included in node [A1B1] and carbon B2 of molecule B included in node [A2B2] are bonded to each other (single bond). In other words, the bond situation between carbon A1 and carbon A2 and the bond situation between carbon B1 and carbon B2 are identical to each other. In this way, in the example of Figure 3, the situation of carbon A1 and carbon A2 in molecule A is the same as the situation of carbon B1 and carbon B2 in molecule B, and node [A1B1] and node [A2B2] are nodes composed of atoms in the same situation. For this reason, in the example shown in Figure 3, no edge is created between node [A1B1] and node [A2B2].
[0016] Next, in the example shown in FIG. 3, whether or not to create an edge between node [A1B4] and node [A2B2] will be described. As can be seen from the graphed structure of molecule A in FIG. 3, carbon A1 of molecule A included in node [A1B4] and carbon A2 of molecule A included in node [A2B2] are bonded to each other (single bond). On the other hand, as can be seen from the graphed structure of molecule B, carbon B4 of molecule B included in node [A1B4] and carbon B2 of molecule B included in node [A2B2] are not directly bonded, with oxygen B3 sandwiched between them. In other words, the bond situation between carbon A1 and carbon A2 is different from the bond situation between carbon B4 and carbon B2. In other words, in the example of Figure 3, the situation of carbon A1 and carbon A2 in molecule A is different from the situation of carbon B4 and carbon B2 in molecule B, and node [A1B4] and node [A2B2] are nodes composed of atoms in different situations. Therefore, in the example shown in Figure 3, an edge is created between node [A1B4] and node [A2B2].
[0017] In this way, a conflict graph can be created based on the rule that if the nodes are composed of atoms in different states, then an edge is created between the nodes, and if the nodes are composed of atoms in the same state, then an edge is not created between the nodes. Fig. 4 is a diagram showing an example of a conflict graph for molecules A and B. As shown in Fig. 4, for example, in nodes [A2B2] and [A5B5], the bonding situation between carbon A2 and oxygen A5 in molecule A is the same as the bonding situation between carbon B2 and carbon B5 in molecule B. For this reason, since nodes [A2B2] and [A5B5] are nodes composed of atoms in the same situation, no edge is created between nodes [A2B2] and [A5B5].
[0018] Next, an example of a method for solving the maximum independent set problem of the created conflict graph will be described. A maximum independent set (MIS) in a conflict graph refers to a set that contains the largest number of nodes that have no edges between them among the sets of nodes that make up the conflict graph. In other words, a maximum independent set in a conflict graph refers to a set that has the largest size (number of nodes) among the sets formed by nodes that have no edges between each other.
[0019] Fig. 5 is a diagram showing an example of a maximum independent set in a graph. In Fig. 5, nodes included in the set are marked with a code "1", nodes not included in the set are marked with a code "0", and where edges exist between nodes, the nodes are connected with solid lines, and where no edges exist, the nodes are connected with dotted lines. Note that, for the sake of simplicity, a graph with six nodes will be used as an example for explanation, as shown in Fig. 5. In the example shown in Fig. 5, among the sets consisting of nodes with no edges between them, there are three sets with the maximum number of nodes, and the number of nodes in these sets is 3. That is, in the example shown in Fig. 5, the maximum independent sets in the graph are the three sets surrounded by dashed lines.
[0020] As described above, the conflict graph is created based on the rule that if the nodes are composed of atoms in different situations, an edge is created between the nodes, and if the nodes are composed of atoms in the same situation, an edge is not created between the nodes. Therefore, in the conflict graph, finding a maximum independent set, which is a set with the maximum number of nodes among sets composed of nodes with no edges between them, is synonymous with finding the largest common substructure between two molecules. In other words, finding a maximum independent set in the conflict graph makes it possible to identify the largest common substructure in two molecules.
[0021] Figure 6 shows an example of the flow for finding the maximum common substructure between molecules A (acetic acid) and B (methyl acetate) by finding the maximum independent set of a conflict graph (solving the maximum independent set problem). As shown in Figure 6, molecules A and B are graphed, and the same elements are combined to form nodes, and edges are formed according to the status of the atoms that make up the nodes to create a conflict graph. Then, by finding the maximum independent set in the created conflict graph, the maximum common substructure between molecules A and B can be found.
[0022] Here, an example of a specific method for determining (searching for) a maximum independent set in a conflict graph will be described. The search for the maximum independent set of the conflict graph can be performed, for example, by using a Hamiltonian whose minimization means searching for the maximum independent set. More specifically, the search can be performed, for example, by using the Hamiltonian (H) shown in the following formula.
[0023]
number
[0024] The relationship between the Hamiltonian expressed by the above formula and the search for the maximum independent set will be explained in more detail. The above formula is a Hamiltonian that represents the Ising model formula in the Quadratic Unconstrained Binary Optimization (QUBO) format. In the above formula, x i When x is 1, it means that the i-th node is included in the set that is a candidate for the maximal independent set, and when it is 0, it means that the i-th node is not included in the set that is a candidate for the maximal independent set. Similarly, in the above formula, x j When it is 1, it means that the jth node is included in a set that is a candidate for the maximal independent set, and when it is 0, it means that the jth node is not included in a set that is a candidate for the maximal independent set. Therefore, for the above formula, under the constraint that there are no edges between nodes whose state is 1 (bit is set to 1), it is possible to search for a maximum independent set by searching for combinations in which as many nodes as possible have a state of 1.
[0025] Here, each term in the above formula will be explained. The term on the right hand side of the above formula (the term with coefficient -α) is x i The more i's that are 1 (the more nodes are included in the set that is a candidate for the maximum independent set), the smaller the value of this term becomes. Note that a small value in one item on the right side of the above formula means that it becomes a large negative number. In other words, in the above formula, when the bits of many nodes become 1 due to the action of one item on the right side, the value of the Hamiltonian (H) becomes small.
[0026] The two terms on the right side of the above equation (terms with coefficient β) are the terms that are used when an edge exists between nodes with a bit of 1 (w ij is a non-zero positive number). In other words, the two terms on the right side of the above equation are 0 if there is no edge between nodes with a bit set to 1, and are positive numbers otherwise. In other words, in the above equation, due to the action of the two terms on the right side, if there is an edge between nodes with a bit set to 1, the value of the Hamiltonian (H) becomes larger.
[0027] As mentioned above, the value of the above equation becomes smaller when the bits of many nodes are 1, and becomes larger when there are edges between nodes whose bits are 1. Therefore, minimizing the above equation means searching for the maximum independent set.
[0028] Here, the relationship between the Hamiltonian expressed by the above formula and the search for a maximal independent set will be described using an example with reference to the drawings. Consider a graph with six nodes, where bits are set for each node as in the example shown in Figure 7. In the example in Figure 7, just like in Figure 5, where edges exist between nodes, the nodes are connected by solid lines, and where no edges exist, the nodes are connected by dotted lines.
[0029] For the example in Figure 7, b in the above formula i is 1, and w when an edge exists between the i-th node and the j-th node ijIf we set it to 1, the above formula becomes:
[0030]
number
[0031] In this way, in the example of Figure 7, if there are no edges between nodes with bits set to 1 (if there are no contradictions as independent sets), the two items on the right-hand side will be 0, and the value of the first item will be the value of the Hamiltonian.
[0032] Next, consider the case where bits are set for each node as shown in the example in Figure 8. As in the example in Figure 7, i is 1, and w when an edge exists between the i-th node and the j-th node ij If we set it to 1, the above formula becomes:
[0033]
number
[0034] Thus, in the example of Fig. 8, there are some edges between nodes where the bit is 1, so the two terms on the right side are not 0, and the value of the Hamiltonian is the sum of the two terms on the right side. Here, in the examples shown in Figs. 7 and 8, for example, if α>5β, then -3α<-4α+5β, and therefore the value of the Hamiltonian in the example of Fig. 7 is smaller than the value of the Hamiltonian in the example of Fig. 8. The example of Fig. 7 is a set of nodes that is consistent as a maximal independent set, and it can be seen that a maximal independent set can be searched for by searching for a combination of nodes that reduces the value of the Hamiltonian in the above formula (1).
[0035] Next, an example of a method for calculating the similarity between the structures of molecules based on the searched maximal independent set will be described. The similarity between the structures of molecules can be calculated, for example, using the following formula:
[0036]
number
[0037] In addition, in the above similarity formula, max{A,B} means to select the larger value of A or B, and min{A,B} means to select the smaller value of A or B.
[0038] Here, as in the examples shown in Figs. 1 to 8, a method for calculating the similarity will be described using acetic acid (molecule A) and methyl acetate (molecule B) as an example. In the conflict graph shown in Figure 9, the maximum independent set is composed of four nodes: node [A1B1], node [A2B2], node [A3B3], and node [A5B5]. That is, in the example of Figure 9, |V A | is 4, and |V c A | is 4, and |V B | is 5, and |V c Bis 4. In addition, in this example, if δ is set to 0.5 and the average of the first molecule and the second molecule is taken (treated equally), the above similarity formula becomes as follows.
[0039]
number
[0040] Thus, in the example of FIG. 9, the structural similarity between the molecules can be calculated to be 0.9 based on the above similarity formula.
[0041] Although the method for calculating the similarity between molecules has been described in detail above, the same method can be appropriately used for calculating the similarity between crystals having a repeating structure. When calculating the similarity between crystals, since crystals have a repeating structure (periodic structure, crystal structure) in which unit cells are repeated three-dimensionally, it is necessary to appropriately represent this repeating structure in order to calculate the similarity with high accuracy.
[0042] FIG. 10 is a schematic diagram showing an example of a periodic structure in a crystal (crystal material) having a repeating structure. In the example shown in FIG. 10, a crystal 10X has a structure in which unit cells 10 containing nine distinguishable types of atoms (atoms A to I) are repeated. For convenience of explanation, FIG. 10 shows a structure in which atoms and unit cells are arranged two-dimensionally, but an actual crystal has a structure in which atoms and unit cells are arranged three-dimensionally. In FIG. 10, dotted lines on the top, bottom, left, and right indicate that unit cells 10 are repeated. 10, a crystal 10X is an aggregate of many unit cells 10. For this reason, when analyzing a crystal 10X (such as calculating similarity), it is necessary to take into account not only the atoms contained in one unit cell 10 but also the atoms contained in the unit cells 10 surrounding it.
[0043] One example of a method for calculating the similarity of crystals taking into account the repeating structure in the crystal is to expand a repeating unit cell in the crystal to include atoms (extended nodes) of other unit cells adjacent to the one unit cell. This method will be described with reference to Fig. 11. As shown in Fig. 11, in the method of expanding and handling a unit lattice using an expansion node, for example, nodes representing atoms A to I are respectively called node A to node I. Then, in this method, for example, each of these nodes (nodes in the lattice) is connected by an edge so as to correspond to the chemical bonds between atoms in a crystal.
[0044] Furthermore, in the method of expanding and handling a unit lattice using an extension node, for example, nodes corresponding to atoms adjacent to and bonded to an atom in a unit lattice including atoms A to I in another unit lattice adjacent to the unit lattice including the atoms A to I are prepared as extension nodes. In the example shown in Fig. 11, for example, an extension node G1 corresponding to atom G in a unit lattice adjacent to the upper side of the unit lattice including node A and an extension node C4 corresponding to atom C in a unit lattice adjacent to the left side of the unit lattice including node A are prepared for node A. Similarly, as shown in Fig. 11, extension nodes corresponding to atoms in adjacent unit lattices are prepared for nodes A to I. Next, as shown in FIG. 11, for nodes A to I, edges (extended edges) are created between the nodes A to I and the prepared extended nodes so as to correspond to the chemical bonds between atoms in the crystal.
[0045] In this way, in the method of expanding and handling unit lattices using expansion nodes, a unit lattice is expanded by using expansion nodes corresponding to atoms in other unit lattices adjacent to the unit lattice, taking into account the bonding state of atoms in the crystal. In this way, the method of expanding and handling unit lattices using expansion nodes can appropriately represent the repeating structure of unit lattices in a crystal, so that analysis (such as calculation of similarity) of crystals can be performed with high accuracy.
[0046] Another example of a method that takes into account the repeating structure of a crystal is a method in which, for a unit cell in a crystal, atoms in one unit cell that correspond to atoms in another unit cell adjacent to the unit cell are identified and expressed as a loop structure. This method will be described with reference to Fig. 12. As shown in Fig. 12, in the method of treating the repeating structure of a crystal as a loop structure, for example, nodes representing atoms A to I are designated as node A to node I, respectively, similar to the example of Fig. 11. Then, in this method, each of these nodes (nodes in a lattice) is connected by an edge so as to correspond to the chemical bonds between atoms in the crystal, similar to the example of Fig. 11.
[0047] Furthermore, as shown in FIG. 12, in a method of treating a repeating structure of a crystal as a loop structure, a loop edge is created so as to correspond to a chemical bond between an atom in another unit cell adjacent to one unit cell including nodes A to I. 12, the loop edge extending above node A represents the same loop edge as the loop edge extending below node G. More specifically, the loop edge connecting node A and node G corresponds to a virtual chemical bond between node A in one unit lattice and node G in another unit lattice adjacent to the upper side of the one unit lattice. Similarly, the other loop edges shown in FIG. 12 each represent a loop edge corresponding to a chemical bond between atoms contained in adjacent unit cells.
[0048] In this way, in a method that treats the repeating structure of a crystal as a loop structure, the chemical bonds between atoms in adjacent unit cells are represented by loop edges, making it possible to express information about the repeating structure of the unit cell (information about the arrangement of the unit cells) without increasing the number of nodes representing the unit cells. That is, in the method of treating the repeating structure of a crystal as a loop structure, for a unit cell in a crystal, atoms in one unit cell that correspond to atoms in another unit cell adjacent to the one unit cell are specified and expressed as a loop structure. In this way, the method of treating the repeating structure of a crystal as a loop structure can appropriately express the repeating structure of the unit cell in the crystal, so that analysis (such as calculation of similarity) targeting the crystal can be performed with high accuracy.
[0049] Here, when calculating the similarity between crystals having a repeating structure, it is preferable to calculate the similarity by assuming multiple unit lattices (by multiplying the unit lattice by an integer) as necessary so that the number of atoms (number of nodes) contained in the structures for which the similarity is to be calculated is close. The relationship between the number of atoms (number of nodes) included in the structure for which the similarity is to be calculated and the value of the calculated similarity will be described below.
[0050] First, a case where the number of atoms (the number of nodes) included in the structures for which similarity is to be calculated is significantly different will be described. As shown in Figures 13A and 13B, a unit cell 200A has nine distinguishable types of atoms (atoms A to I), and a unit cell 200B has 24 distinguishable types of atoms (atoms A to X). When all atoms in the unit cell 200A in Fig. 13A form a common substructure with atoms in the unit cell 200B in Fig. 13B (the unit cell 200A in Fig. 13A is included in the unit cell 200B in Fig. 13B), the similarity between these structures is usually considered to be very large. However, when the similarity is calculated based on the above-mentioned similarity formula, the similarity between these structures is calculated to be "0.675", which is not a large value. From this, it can be seen that the number of nodes (number of atoms) contained in unit cell 200A in Figure 13A and unit cell 200B in Figure 13B are significantly different, which causes the similarity to be calculated as low, resulting in low analysis accuracy.
[0051] The similarity S between unit cell A and unit cell B can be calculated by the following formula by simplifying the above similarity formula. S = {N(A ∩ B) / N(A) + N(A ∩ B) / N(B)} × 0.5 Here, in the above formula, "N(A ∩ B) ≤ N(A)" and "N(A ∩ B) ≤ N(B)".
[0052] In the above formula, when "N(A) ≠ N(B)", the similarity S does not become "1". Specifically, when N(A) = λN(B) (0 < λ ≤ 1), the maximum value that the similarity S can take is "(1 + λ) / 2". Also, since a crystal can be considered as a periodic structure in which the unit cell is repeated semi-infinitely, the fact that the similarity S does not take the value of "1" due to the difference in the number of atoms (node numbers) per unit cell cannot be regarded as an essential discussion of the similarity S of crystals.
[0053] Here, as a method of representing the possible values of the similarity S as 0 to 1, for example, the following two methods can be cited.
[0054] · Normalization of similarity S When normalizing the similarity S, for example, the possible values of the similarity S (0 ≤ S ≤ (1 + λ) / 2) are made to be 0 to 1, and the similarity S' is normalized by the following formula. S' = {N(A ∩ B) / N(A) + N(A ∩ B) / N(B)} / (1 + λ)
[0055] · Unit cells A and B as integer multiples By appropriately multiplying the unit cells A and B by integer multiples and treating them as multiple lattices, the difference between N(A) and N(B) is made as small as possible, and the possible values of the similarity S are made to be 0 to 1.
[0056] When normalizing the similarity S, when N(A) < N(B), even if S' = 1, there exists a substructure B\A (the difference set obtained by subtracting the elements of unit cell A from unit cell B) that is not included in "A ∩ B". That is, regardless of the structure of the substructure B\A and the number of atoms constituting each unit cell, there may be a case where the similarity S' = 1. Thus, the method of normalizing the similarity S is not suitable as a method for evaluating similarity.
[0057] On the other hand, in the method of integer multiples of unit lattice A and unit lattice B, for example, A" is a structure of multiple lattices of unit lattice A, and B" is a structure of multiple lattice of unit lattice B, so that N(A") = N(B"), and the similarity S''' is expressed by the following formula. S'''={N(A”∩B”) / N(A”)+N(A”∩B”) / N(B”)} / (1+λ) However, in order to make N(A") = N(B"), it may be necessary to increase the number of target unit lattices (for example, to make it the least common multiple of unit lattices A and B). In such a case, it is preferable to multiply at least one of unit lattices A and B by an integer so that N(A") ≒ N(B"), that is, the number of nodes in the structures for which similarity is calculated, is close to each other. As a specific guideline, when N(A") = λN(B"), it can be set to "9 / 10≦λ≦10 / 9". By defining it in this way, it is possible to reduce the number of lattices required for the calculation of similarity while making the number of nodes included in the target structures closer, and to improve the accuracy of the calculated similarity.
[0058] Here, a method for integrally multiplying unit cell A and unit cell B will be described with reference to FIGS. 14A to 14C. In the examples of Figures 14A to 14C, graph A is a graph of unit lattice A using nodes and edges, and graph B is a graph of unit lattice B using nodes and edges. In the method of integer multiplying unit lattice A and unit lattice B, for example, in order to make the number of nodes in graph A and the number of nodes in graph B closer to each other, at least one of graph A and graph B is duplicated by integer multiplication. At this time, for example, it is preferable to perform this so that the absolute value of the difference between the total number of nodes in graph A after duplication and the total number of nodes in graph B after duplication is minimized.
[0059] In this example, in the method of integer multiplying unit lattice A and unit lattice B, for example, graph A is duplicated four times and graph B is duplicated nine times. In the graph obtained by integrating the duplicated graphs, for example, graphs A may be integrated with each other and graphs B may be integrated with each other as shown in Fig. 14B. In this case, integration means providing an edge between the graphs (between graphs A and graphs B). In addition, in the graph obtained by integrating the duplicated graphs, for example, graphs A may not be integrated with each other and graphs B may not be integrated with each other as shown in Fig. 14C.
[0060] In this way, in the method of integer-multiplying unit cell A and unit cell B, for example, at least one of the graphs of the unit cells of the two crystals is duplicated. By duplicating the graph and bringing the number of atoms (number of nodes) of the two unit cells being compared closer together, it is possible to avoid a decrease in similarity caused by the difference in the number of atoms (number of nodes). As a result, it is possible to improve the accuracy of the analysis (e.g., similarity).
[0061] Here, when calculating the similarity between crystals, as described above, the number of atoms contained in each unit lattice for comparing structures may be large (the number of bits required to calculate the similarity may be large). In particular, when the crystal is an ionic crystal, the unit lattice contains many anions (e.g., oxygen atoms), and the number of atoms contained in the unit lattice is large. FIG. 15A shows an example of a unit lattice in an ionic crystal. Ionic crystal A shown in FIG. 15A has 8 lithium (Li), 4 manganese (Mn), 16 phosphorus (P), and 48 oxygen (O) atoms in the unit lattice, for a total of 76 atoms. Similarly, FIG. 15B shows another example of a unit lattice in an ionic crystal. Ionic crystal B shown in FIG. 15B has 16 lithium (Li), 8 manganese (Mn), 16 phosphorus (P), and 56 oxygen (O) atoms in the unit lattice, for a total of 96 atoms.
[0062] In the example shown in FIG. 15A and FIG. 15B, when calculating the similarity between ionic crystal A and ionic crystal B, for example, the above-mentioned conflict graph node is created for each type of atom. At this time, the number of nodes of the conflict graph (number of bits required for calculation) is "8×16=128" for lithium, "4×8=32" for manganese, "16×16=256" for phosphorus, and "48×56=2688" for oxygen. Therefore, in order to calculate the similarity between ionic crystal A and ionic crystal B, for example, the number of nodes of the conflict graph (number of bits required for calculation) is "128+32+256+2688=3104".
[0063] In this way, when the number of atoms (nodes) included in a unit lattice is large, it is difficult to calculate the similarity assuming multiple unit lattices due to the limitation of the number of bits that can be handled by the annealing machine used to calculate the similarity. For example, the number of bits available for an annealing machine is at most about 8,000 bits, and when 3,000 bits or more are required even for comparing unit lattices as in the above example, it is difficult to handle the unit lattice as an integer multiple due to a shortage of available bits. As described above, in the prior art, when analyzing ionic crystals, if the unit cell of the ionic crystal contains a large number of atoms, the number of bits required for calculation becomes large, making accurate analysis difficult.
[0064] Therefore, the present inventors have conducted extensive research into methods for reducing (suppressing) the number of bits required for analyzing ionic crystals, and have obtained the following findings. That is, the present inventors have found that the number of bits required for analyzing ionic crystals can be reduced (suppressed) by using the following crystal analysis method or the like. The crystallography method as one example of the technology disclosed in the present application is a method for analyzing a crystal, a graph creation step of creating a graph in which a repeating unit cell including an intra-lattice node indicating data of an atom in the repeating unit cell and an intra-lattice edge indicating data of a chemical bond between two atoms in the repeating unit cell is expanded to the repeating unit cell and an adjacent repeating unit cell adjacent to the repeating unit cell; an analysis step of analyzing the crystal using the graph; Including, In the graph creation step, when the crystal is an ionic crystal, when the number of intra-lattice nodes which are data on anionic atoms bonded to cationic atoms in one repeating unit lattice in the ionic crystal is n, the number of intra-lattice nodes which are data on anionic atoms is set to n-1 or less.
[0065] Here, in one example of the technique disclosed herein, in the graph creation step, for example, a repeating unit lattice is prepared that includes an intra-lattice node that indicates data of atoms in the repeating unit lattice and an intra-lattice edge that indicates data of chemical bonds between two atoms in the repeating unit lattice. That is, in one example of the technique disclosed herein, for example, in order to represent atoms of a unit lattice in a crystal and chemical bonds between the atoms, a repeating unit lattice (graph of a unit lattice) that includes the intra-lattice node and the intra-lattice edge is prepared. In one example of the technology disclosed herein, in the graph creation step, for example, a graph is created in which one repeating unit lattice is expanded to adjacent repeating unit lattices adjacent to the one repeating unit lattice. That is, in one example of the technology disclosed herein, for example, in order to express the repeating structure (periodic structure, crystal structure) of an ionic crystal, a graph is created in which one repeating unit lattice is expanded in consideration of the relationship with atoms of the surrounding repeating unit lattices. In this way, in one example of the technology disclosed herein, the repeating structure in the crystal can be appropriately expressed, and the crystal can be analyzed with high accuracy.
[0066] Furthermore, in one example of the technique disclosed herein, in the graph creation step, for example, when the crystal is an ionic crystal, and the number of intra-lattice nodes that are data on anionic atoms bonded to cationic atoms in one repeating unit lattice in the ionic crystal is n, the number of intra-lattice nodes that are data on anionic atoms is set to n-1 or less. In other words, in the graph creation step, for example, at least one of the anionic atoms contained in one repeating unit lattice is not regarded as a node (deleted), so that the cationic atom and the anionic atom are regarded as one node together. In this way, in one example of the technology disclosed herein, the number of nodes in a lattice in one repeating unit lattice can be suppressed (reduced), and therefore the number of bits required for calculation can be reduced (reduced). In other words, in one example of the technology disclosed herein, by creating a graph in which at least some of the nodes of anionic atoms in an ionic crystal are deleted, the number of nodes included in the graph (the number of bits required for calculation) can be reduced.
[0067] Next, in an example of the technique disclosed herein, in the analysis step, for example, the graph created in the graph creation step is used to analyze the crystal. That is, in an example of the technique disclosed herein, in the analysis step, for example, the ionic crystal is analyzed based on a graph in which at least some of the nodes of anionic atoms in the ionic crystal are deleted. By doing this, one example of the technology disclosed herein uses a graph in which the number of nodes in the lattice is suppressed, making it possible to perform highly accurate analysis, such as, for example, multiplying the unit lattice by an integer to make the number of nodes uniform and then calculating the similarity.
[0068] Here, the number of nodes in the lattice, which is data of anionic atoms in the graph created in the graph creation step, is not particularly limited as long as it is n-1 or less, and can be appropriately selected according to the purpose, as described above, but it is preferable to set it to "0 (zero)". In other words, in one example of the technology disclosed in the present application, it is preferable not to regard all anionic atoms contained in one repeating unit lattice as nodes (delete all anions). The form in which the number of intra-lattice nodes, which are data of anionic atoms, in the graph created in the graph creating step is set to "0" will be described with reference to FIG.
[0069] FIG. 16 shows an example in which one repeating unit cell in an ionic crystal contains one each of lithium (Li), manganese (Mn), and phosphorus (P), and 14 oxygen atoms (O). As shown in the example of the conventional technology in FIG. 16, in an ionic crystal, each cationic atom (lithium, manganese, and phosphorus) is bonded with a number of oxygen atoms according to the valence by a single bond. In this way, in an ionic crystal, cationic atoms and anionic atoms are usually bonded to each other. In the example of the conventional technology, a graph is created by regarding these atoms as nodes and interatomic bonds as single bond edges. In the example shown in Fig. 16, for example, there are two bonds between phosphorus (P) and manganese (Mn) located at the bottom left of the unit lattice via oxygen (O), and phosphorus (P) and manganese (Mn) located at the upper side of the unit lattice are bonded to each other via one oxygen (O) of a single bond. Furthermore, in the example shown in Fig. 16, for example, manganese (Mn) and lithium (Li) are bonded to each other via two oxygen (O) of a single bond.
[0070] As shown in one embodiment of the present application in Fig. 16, in one example of the technology disclosed herein, a cationic atom (+) and an anionic atom (-) surrounding the cationic atom are collectively treated as one node. In other words, as shown in one embodiment of the present application in Fig. 16, in one example of the technology disclosed herein, for example, anionic atoms are not considered as nodes, and only cationic atoms are considered as nodes, thereby making it possible to reduce the number of nodes included in a unit lattice. In addition, in one example of the technology disclosed herein, the bond order between cationic atoms can be the number of chemical bonds between the cationic atoms via anions that are considered not to exist (the number of anions deleted between the cations). In other words, in one example of the technology disclosed herein, the bond order between cationic atoms can be determined by the number of bonds between the cationic atoms via an anionic atom. For example, in the example shown in Figure 16, the bond order between phosphorus (P) and manganese (Mn) located at the lower left of the unit lattice can be "2", and the bond order between phosphorus (P) and manganese (Mn) located at the upper side of the unit lattice can be "1". Furthermore, in the example shown in Figure 16, the bond order between manganese (Mn) and lithium (Li) can be "2".
[0071] In one example of the technology disclosed herein, as shown in one embodiment of the present invention in Fig. 16, for example, the number of nodes included in a unit lattice can be reduced by considering only cationic atoms as nodes, without considering anionic atoms as nodes. In the example shown in Fig. 16, the number of nodes included in the unit lattice is 17 in one example of the conventional technology, but in one embodiment of the present invention, the number of nodes included in the unit lattice can be reduced to 4.
[0072] In this manner, in one example of the technology disclosed herein, by deleting at least a portion of the anionic atoms contained in the unit lattice and treating the cationic atoms and anionic atoms together as one node, the number of nodes contained in the unit lattice can be reduced. Therefore, in one example of the technology disclosed herein, it is possible to reduce (suppress) the number of nodes included in a graph used in analyzing an ionic crystal, that is, the number of bits required for calculation.
[0073] Each step in the crystal analysis method disclosed herein will be described in detail below. The crystal analysis method disclosed herein includes, for example, a graph creation step and an analysis step, and may further include other steps as necessary.
[0074] <Graph creation process> In the graph creation step of the example of the technique disclosed herein, as described above, a repeating unit lattice is prepared, which includes an intra-lattice node indicating data of atoms in the repeating unit lattice and an intra-lattice edge indicating data of chemical bonds between two atoms in the repeating unit lattice. Then, in the graph creation step, a graph is created in which the repeating unit lattice is expanded to an adjacent repeating unit lattice adjacent to the repeating unit lattice.
[0075] <<Graph>> In one example of the technology disclosed herein, a graph has nodes that are, for example, data on atoms, and edges that are data on chemical bonds between two atoms. Here, in one example of the technology disclosed herein, a graph can be, for example, an abstract data type composed of a group of nodes (vertices) and a group of edges (branches) that represent the connections between the nodes. A graph is represented, for example, as G=(V,E), where V is a set of nodes and E is a set of edges. V is a finite set, and E is a set of sets consisting of two elements selected from V. A node may represent an atom within a unit cell of a crystal. Hereinafter, a node representing an atom within a unit cell of a crystal may be referred to as an "intra-cell node." Additionally, an edge may represent a chemical bond between atoms within a unit lattice. Hereinafter, an edge representing a chemical bond between atoms within a unit lattice may be referred to as an "intra-lattice edge."
[0076] The graph may have nodes (hereinafter sometimes referred to as "extended nodes") that indicate atoms in adjacent unit cells adjacent to the repeating unit cell, the atoms having chemical bonds with atoms in the repeating unit cell. In this case, the graph may have an edge (hereinafter, sometimes referred to as an "extended edge") indicating a chemical bond between an atom corresponding to a node in the lattice and an atom corresponding to an extended node. Furthermore, in this case, the graph may have an edge (hereinafter, sometimes referred to as an "extended inter-node edge") indicating a chemical bond between two atoms corresponding to the extended node.
[0077] That is, in one example of the technology disclosed in the present application, the graph created in the graph creation step is: an intra-lattice node representing data for an atom in one repeating unit cell; and an intra-lattice edge representing data for a chemical bond between two atoms in the unit cell. It is preferable that the data includes extended nodes, which are data on atoms in adjacent repeating unit lattices that are chemically bonded to atoms in a repeating unit lattice and are adjacent to the repeating unit lattice, and extended edges, which are data on chemical bonds between atoms corresponding to intra-lattice nodes in a repeating lattice unit and atoms corresponding to extended nodes. In this way, in one example of the technology disclosed in the present application, the method of expanding and handling a unit lattice by using an expansion node, which has been described with reference to FIG. 11, can be suitably used.
[0078] The graph may also have an edge indicating a virtual chemical bond between atom X in the unit lattice and atom Y in the unit lattice, which atom Y' has a chemical bond with atom X in the unit lattice and corresponds to atom Y' in an adjacent unit lattice adjacent to the unit lattice. Here, atom Y in the unit lattice corresponding to atom Y' in the adjacent unit lattice means an atom in the unit lattice that overlaps with atom Y' of the adjacent unit lattice when the adjacent unit lattice and the unit lattice are overlapped so that the constituent atoms overlap. In the following description, the edge indicating the virtual chemical bond between atom Y in the unit lattice and atom X in the unit lattice may be referred to as a "loop edge."
[0079] That is, in one example of the technology disclosed in the present application, the graph created in the graph creation step is: an intra-lattice node representing data for an atom in one repeating unit cell; and an intra-lattice edge representing data for a chemical bond between two atoms in the unit cell. and a loop edge indicating data of a hypothetical chemical bond between atom Y in one repeating unit lattice and atom X, the atom Y' being chemically bonded to atom X in one repeating unit lattice and corresponding to atom Y' in an adjacent repeating unit lattice adjacent to the one repeating unit lattice. In this way, in one example of the technique disclosed herein, the method of treating the repeating structure of the crystal as a loop structure, as described with reference to FIG. 12, can be suitably used.
[0080] The nodes within the lattice and the extended nodes each have data such as the type of atom, the valence of the atom, and the charge of the atom. The intra-lattice edges, loop edges, and extension edges have data such as bond type, bond angle, bond distance, bond order, etc. Bond angle and bond distance can be represented using, for example, atomic or chemical bond coordinate data.
[0081] Furthermore, the intra-lattice edges, loop edges, and extension edges are preferably created, for example, by Voronoi division between the nodes. The Voronoi division is a method of dividing the nearest neighboring region of each node by drawing a perpendicular bisector on a line connecting adjacent nodes.
[0082] Here, the Voronoi diagram and the Voronoi division will be explained. A Voronoi diagram is a diagram in which multiple points (nuclear points) placed at any position in a certain metric space are divided into regions based on which other points in the same metric space are closer to each other. This division is also called Voronoi tessellation. In a Voronoi diagram, the division pattern is determined only by the positions of the kernel points. When a unit lattice of a crystal is represented diagrammatically, including chemical bonds, the chemical bonds of the unit lattice of the crystal are usually not uniquely determined. However, by creating edges (internal lattice edges, loop edges, extended edges) using Voronoi tessellation, the edges (internal lattice edges, loop edges, extended edges) can be uniquely determined. As a result, edges can be uniquely determined for multiple crystals, improving the accuracy of crystal analysis.
[0083] In addition, the intra-lattice edges, loop edges, and extended edges can also be created based on, for example, the standard ionic radius of the atoms. For example, a rule may be applied that when the distance between any two nodes is d and the ionic radii of each node are r1 and r2, if "d≦a(r1+r2)+b", an edge between the nodes is created. Here, a is a constant equal to or greater than 0, and b is a constant. Also, 1.0≦a≦1.2 and 0 Å≦b≦0.4 Å are preferable. The method of creating the intra-lattice edges, loop edges, and extended edges can be freely selected from, for example, the above-mentioned Voronoi division method and the method using the standard ion radius. When both are selected, for example, among the intra-lattice edges, loop edges, and extended edges that can be determined by Voronoi division, those that are considered to be intra-lattice edges, loop edges, and extended edges by the method using the standard ion radius of atoms may be created as intra-lattice edges, loop edges, and extended edges.
[0084] Furthermore, in one example of the technology disclosed herein, as described above, in the graph creation process, when the crystal is an ionic crystal, when the number of intra-lattice nodes which are data on anionic atoms bonded to cationic atoms in one repeating unit lattice in the ionic crystal is n, the number of intra-lattice nodes which are data on anionic atoms is set to n-1 or less. Here, the ionic crystal means, for example, a crystal formed by ionic bonding between a cation (positive ion, cationic atom) and an anion (negative ion, anionic atom). In one example of the technology disclosed herein, the ionic crystal to be analyzed is not particularly limited and can be appropriately selected according to the purpose.
[0085] In addition, in one example of the technology disclosed herein, as described above, it is preferable not to regard all anionic atoms contained in the repeating unit lattice as nodes (delete all anions). In other words, in one example of the technology disclosed herein, it is preferable to regard the nodes of anionic atoms as "0" and all cationic atoms as being directly bonded to each other. More specifically, in one example of the technique disclosed herein, in the graph creation step, it is preferable to set the number of intra-lattice nodes, which are data on anionic atoms, to 0. By doing so, the number of nodes contained in the repeating unit lattice can be further reduced, and the number of bits required for analyzing an ionic crystal can be further suppressed.
[0086] In the graph creation process, when the number of intra-lattice nodes, which are data on anionic atoms, is not set to 0 (when a portion of all anionic atoms contained in the repeating unit lattice is left), the anionic atoms regarded as intra-lattice nodes can be appropriately selected depending on the purpose. In one example of the technology disclosed herein, for example, whether or not to regard an anionic atom as a node in a lattice (whether or not to delete it) may be determined based on the bonding relationship of each anionic atom in a repeating unit lattice. In this case, for example, an anionic atom bonded to two or more cationic atoms may be deleted, and an anionic atom bonded to only one cationic atom may be left.
[0087] In addition, in one example of the technology disclosed herein, the anionic atoms that are not considered to be nodes in the lattice (anionic atoms to be deleted) are not particularly limited and can be appropriately selected depending on the purpose, and examples thereof include oxygen atoms, chlorine atoms, and fluorine atoms. Anionic atoms that are not considered to be nodes within the lattice can be selected appropriately depending on the ionic crystal to be analyzed. For example, by selecting anionic atoms with a large number of atoms in the ionic crystal, the number of bits required for calculation can be further reduced.
[0088] In one example of the technology disclosed herein, when an anionic atom is deleted, the bond order between cationic atoms can be, for example, the number of chemical bonds between the cationic atoms via an anion that is considered not to exist between the cationic atoms. In other words, in one example of the technology disclosed herein, the bond order between cationic atoms can be determined by the number of bonds between the cationic atoms via an anionic atom, as described above. In other words, in one example of the technology disclosed herein, in the graph creation process, when the number of nodes in a lattice, which is data on anionic atoms, is set to 0 and cationic atoms in a repeating unit lattice are considered to be directly chemically bonded to each other, it is preferable that the bond order of the chemical bonds between the cationic atoms is set to the number of chemical bonds between the cationic atoms via anionic atoms that are considered to be non-existent. By doing so, in one example of the technology disclosed herein, as shown in FIG. 16, the bonding state of each atom in the ionic crystal can be more appropriately taken into account, enabling more accurate analysis.
[0089] <Analysis process> In the analysis step in the example of the technique disclosed herein, as described above, the graph created in the graph creation step is used to analyze the crystal. The analysis of the crystals performed in the analysis step is not particularly limited and can be appropriately selected depending on the purpose. For example, the analysis can include analysis of the similarity of the structures of multiple crystals and prediction of the properties of the crystals.
[0090] The method of analyzing the similarity of the structures (crystal structures) of multiple crystals is not particularly limited, and can be appropriately selected according to the purpose, as long as it is a method of analyzing the similarity between the crystal structure of a crystal and the crystal structure of another crystal using the graph created in the graph creation step. For example, the method of analyzing the similarity of the structures of multiple crystals includes a method of calculating a score of similarity, and judging that the larger the score, the higher the similarity.
[0091] The method for calculating the similarity score is not particularly limited and can be appropriately selected depending on the purpose. For example, the following methods (1) and (2) can be mentioned. Method (1) Fingerprinting Method (2) A method to analyze the similarity of structures by searching for common substructures, expressing the maximum independent set problem of a conflict graph as an Ising model formula and solving it with an annealing machine or the like.
[0092] In the fingerprint method as method (1), for example, whether or not a partial structure in a reference structure is included in a structure to be compared is expressed as 0 or 1 to evaluate the similarity. As for the technique (2) using a conflict graph, the technique described with reference to Fig. 1 to Fig. 9 can be suitably used. Details of the similarity analysis using a conflict graph in one example of the technology disclosed in the present application will be described later.
[0093] It should be noted that the method of calculating the similarity score is not limited to these methods, and the similarity score may be calculated, for example, using cosine similarity, correlation coefficient, correlation function, edit distance (also called Levenshtein distance), etc., which are related to similarity calculation.
[0094] The method for predicting the crystal characteristics is not particularly limited as long as it is a method that predicts the crystal characteristics using the graph created in the graph creation process, and can be selected appropriately depending on the purpose, such as predicting characteristics using a learning model.
[0095] An example of predicting characteristics using a learning model is a method of creating a learning model by machine learning and predicting the characteristics of a crystal using the learning model. Machine learning can be exemplified by supervised learning and unsupervised learning. There is no particular limitation on the learning method used to create a learning model by machine learning, and it can be appropriately selected according to the purpose. In the case of supervised learning, for example, a method can be used to perform learning using a graph of a crystal and a data set of teacher data having the characteristics of the crystal as labels. As training data in machine learning, for example, data having the graph created in the graph creation process and data on the properties of the crystal that corresponds to the graph can be used. The predicted properties are not particularly limited and may be appropriately selected depending on the purpose, and may be chemical properties or physical properties. Specific examples of the properties include electrical conductivity, ionic conductivity, discharge potential, dielectric constant, thermal conductivity, and specific heat.
[0096] In addition, when predicting characteristics, the graph (graph-structured data) created in the graph creation process may be processed and used. For example, by converting the graph into a tensor, the graph-structured data can be learned with high accuracy using machine learning (e.g., deep learning) technology. Here, the term "tensor" refers to data expressed as a multidimensional array, which generalizes the concepts of matrices and vectors.
[0097] In one example of the technique disclosed herein, as described above, it is preferable to calculate the similarity in the ionic crystal in the analysis step. In other words, in one example of the technology disclosed herein, in the graph creation process, it is preferable to create a first graph for the first crystal and a second graph for the second crystal, and in the analysis process, calculate the similarity between the first crystal and the second crystal based on the first graph and the second graph.
[0098] Furthermore, in one example of the technology disclosed herein, as described above, when calculating the similarity, it is preferable to calculate the similarity by assuming multiple unit lattices (by multiplying the unit lattice by an integer) as necessary so that the number of atoms (number of nodes) contained in the structures for which the similarity is to be calculated is close. In other words, in one example of the technology disclosed herein, it is preferable that the calculation of the similarity in the analysis process is performed after at least one of a repeating unit lattice for the first graph and a repeating unit lattice for the second graph is integer-multiplied so that the numbers of nodes within the lattice in the first graph and the second graph are approximated. By doing so, in one example of the technology disclosed herein, by bringing the numbers of atoms (number of nodes) of the two repeating unit cells whose structures are being compared closer together, it is possible to avoid a decrease in similarity caused by the difference in the number of atoms, and to further improve the accuracy of the similarity calculation results.
[0099] In addition, in one example of the technique disclosed herein, as described above, it is preferable to calculate the similarity in the analysis step by solving a maximum independent set problem of a conflict graph. As a method for calculating the similarity between the first crystal and the second crystal by solving the maximum independent set problem of a conflict graph, the method described with reference to Figs. 1 to 9 can be suitably used. The conflict graph used to calculate the similarity between a first crystal and a second crystal has nodes that are, for example, combinations of each node atom that constitutes the repeating unit lattice of the graphed first crystal and each node atom that constitutes the repeating unit lattice of the graphed second crystal. Here, in one example of the technology disclosed herein, "two nodes are compared to be identical to each other" means that when two nodes are compared to each other, these nodes are composed of node atoms in the same state (bonding state). Similarly, in one example of the technology disclosed herein, "two nodes are compared to be not identical to each other" means that when two nodes are compared to each other, these nodes are composed of node atoms in different states (bonding states).
[0100] More specifically, in one example of the technology disclosed herein, for example, it is preferable that the similarity in the analysis step is calculated by searching for a maximum independent set of a conflict graph between a first graph and a second graph using the following formula (1): Moreover, the following formula (1) is used when the number of nodes in the lattice, which is data of anionic atoms, is set to 0 in the graph creation step.
number
[0101] Here, c' in the above formula (1) ij We will explain in detail about this. First, as described above, in one example of the technology disclosed herein, the bond order between cationic atoms when an anionic atom is deleted can be determined, for example, by the number of bonds between cationic atoms via an anionic atom. For this reason, the bond order between cationic atoms is determined by the combination of cationic atoms, and can be incorporated as a weight in the penalty term (term with coefficient β) of the Ising model formula (formula (1)) for solving the maximum independent set problem of the conflict graph. In other words, in one example of the technology disclosed in the present application, the weight (c′ ij ) the bond order between cationic atoms can be expressed as , whereby the similarity can be calculated taking into account the bond order between cationic atoms.
[0102] For example, c' represents the weight of the bond order between cationic atoms. ij can be determined, for example, according to the following rules: If there is no bond between the combination of atoms in the i-th node and the combination of atoms in the j-th node, then "c′ ij =0". If a bond exists between the pair of atoms in the i-th node and the pair of atoms in the j-th node, then "c′ ij =λ1". If there is a bond between the combination of atoms in the i-th node and the combination of atoms in the j-th node, then "c′ ij = λ2(1-min{c1 / c2,c2 / c1}). However, λ1 and λ2 are positive constants, and it is usually preferable to set λ1 = λ2 = 1. In the following explanation, the case of λ1 = λ2 = 1 will be explained as an example.
[0103] Here, min{A,B} means to select the smaller value of A and B. Furthermore, c1 means the bond order between the atom of the i-th node and the atom of the j-th node in the first graph, and c2 means the bond order between the atom of the i-th node and the atom of the j-th node in the second graph.
[0104] Figure 17 shows the combination of nodes in the first and second graphs and the weights of the bond orders between cationic atoms, c′ ij An example of the relationship is shown below. As shown in Figure 17, c′ represents the weight on the bond order. ij According to the rule for, for example, if there is no connection between node A1 and node A2 in the first graph and between node B1 and node B2 in the second graph, then c′ ijIn other words, in this case, these nodes are composed of atoms in the same situation (bonding situation) (there is no contradiction in the bonding situation), so c′ ij will result in "0". Similarly, if there is a bond with the same bond order between node A1 and node A2 in the first graph and between node B1 and node B2 in the second graph (in the example of FIG. 17, both have a bond order of 1), then c′ ij will result in "0".
[0105] Furthermore, as shown in FIG. ij According to the rule for, for example, if there is a connection between node A1 and node A2 in the first graph and there is no connection between node B1 and node B2 in the second graph, then c′ ij In other words, in this case, these nodes are composed of atoms in completely different situations (bonding situations) from each other (the bonding situations are completely contradictory), so the penalty term with coefficient β is set to a large value, and c′ ij will result in "1".
[0106] Also, as shown in FIG. ij According to the rule for, for example, when there are bonds with different bond orders between node A1 and node A2 in the first graph and between node B1 and node B2 in the second graph, c′ ij is a number between 0 and 1 according to the bond order. In other words, in this case, these nodes are not contradictory in terms of bonding, but since the bond orders are different, they are composed of atoms in somewhat different situations (bonding situations). Therefore, c' is set so that the value of the penalty term with coefficient β corresponds to the bond order. ij means "a number between 0 and 1." In this manner, in one example of the technology disclosed herein, c′, which represents the weight related to the bond order, is used. ijcan be a multi-valued weight (weight according to the difference in bond order) according to the bond order, instead of a weight (binary value of 0 or 1) that considers only whether or not there is a bond between nodes. By doing so, in one example of the technology disclosed in the present application, it is possible to calculate the similarity with higher accuracy by considering the bond order between cationic atoms. In the above explanation, the case where λ1=λ2=1 has been described as an example, but the technology disclosed in the present application is not limited to this, and λ1 and λ2 can be positive constants other than 1 as necessary.
[0107] In one example of the technology disclosed herein, when searching for a maximum independent set using the above formula (1), it is not essential to create a conflict graph, and it is sufficient to at least minimize the above formula (1). In other words, in one example of the technology disclosed herein, the search for a maximum independent set in a conflict graph is replaced with and solved as a combinatorial optimization problem in a Hamiltonian, where minimization means searching for a maximum independent set. Here, minimization of the Hamiltonian expressed by the Ising model formula in the QUBO format as in the above formula (1) can be performed in a short time by performing simulated annealing using an annealing machine or the like. Therefore, in one aspect, the technology disclosed in the present application uses the above formula (1) to search for the maximum independent set by a simulated annealing method using an annealing machine or the like, so that the similarity of the crystals can be calculated in a shorter time. In other words, in one aspect, the technology disclosed in the present application uses the simulated annealing method to minimize the Hamiltonian (H) in the above formula (1) to search for the maximum independent set, so that the similarity of the crystals can be calculated in a shorter time. Details of the simulated annealing method will be described later.
[0108] Here, the annealing machine is not particularly limited as long as it is a computer that employs an annealing method for performing a ground state search for an energy function represented by an Ising model, and can be appropriately selected according to the purpose. Examples of the annealing machine include a quantum annealing machine, a semiconductor annealing machine using semiconductor technology, and a machine that performs simulated annealing executed by software using a CPU or a GPU (Graphics Processing Unit). In addition, for example, a Digital Annealer (registered trademark) may be used as the annealing machine.
[0109] In the above example, the formula was described for the case where the number of nodes in the lattice, which are data of anionic atoms, is set to 0. However, if the number of nodes in the lattice, which are data of anionic atoms, is not set to 0 (i.e., if some of the anionic atoms are left), for example, the following formula can be used.
number
[0110] In the above formula, "c'' i" has been introduced, and this "c'' i By introducing ", we can add "combinations of the same anionic atoms" to the nodes of the graph. Here, "c'' i " can be a constant related to the selection of the combination of the i-th node in the conflict graph between the first graph and the second graph. More specifically, the constant related to the combination between cationic atoms is "c'' j " and the constant related to the combination between anionic atoms is "c'' k ", "c' j ≧c′′ k ≧0". This is because, in one example of the technology disclosed herein, the number of anionic atoms is reduced, and therefore it is not essential that the similarity calculation results are affected by the selection of anionic atoms to be deleted.
[0111] Furthermore, in an example of the technology disclosed herein, similarly to the method described with reference to Figures 1 to 9, the similarity related to the characteristics of the searched maximal independent set can be calculated using the following formula (2). In other words, in an example of the technology disclosed herein, it is preferable that the calculation of the similarity in the analysis step is performed for the searched maximal independent set using the following formula (2).
[0112]
number
[0113] <Other processes> The other steps are not particularly limited and can be appropriately selected depending on the purpose.
[0114] (Crystal analyzer) The crystal analysis apparatus disclosed in this application is In the crystal, a graph creating unit that creates a graph in which a repeating unit lattice including an intra-lattice node indicating data of an atom in the repeating unit lattice and an intra-lattice edge indicating data of a chemical bond between two atoms in the unit lattice is expanded to the repeating unit lattice and an adjacent repeating unit lattice; an analysis unit that performs crystal analysis using a graph; having When the crystal is an ionic crystal, the graph creation unit sets the number of intra-lattice nodes that are data on anionic atoms bonded to cationic atoms in a repeating unit lattice in the ionic crystal to n-1 or less when the number of intra-lattice nodes that are data on anionic atoms bonded to cationic atoms in a repeating unit lattice in the ionic crystal is n.
[0115] The crystal analysis apparatus disclosed herein may be, for example, an apparatus for executing the crystal analysis method disclosed herein. In addition, a preferred embodiment of the crystal analysis apparatus disclosed herein may be, for example, the same as a preferred embodiment of the crystal analysis method disclosed herein. Here, the crystal analysis device disclosed herein has a graph creation unit and an analysis unit, and may further have other units (means) as necessary. Note that the graph creation unit and the analysis unit can be realized, for example, by combining a processor such as a CPU (Central Processing Unit) with a memory such as a RAM (Random Access Memory), or as part of an annealing machine.
[0116] (Crystal analysis program) The crystal analysis program disclosed in this application is In the crystal, a graph creation step of creating a graph in which a repeating unit cell including an intra-lattice node indicating data of an atom in the repeating unit cell and an intra-lattice edge indicating data of a chemical bond between two atoms in the unit cell is expanded to the repeating unit cell and an adjacent repeating unit cell adjacent to the repeating unit cell; an analysis step of analyzing the crystal using the graph; Including, In the graph creation step, when the crystal is an ionic crystal, when the number of nodes in the lattice which are data on anionic atoms bonded to cationic atoms in one repeating unit lattice in the ionic crystal is n, the number of nodes in the lattice which are data on anionic atoms is n-1 or less. The processing is carried out by a computer.
[0117] The crystal analysis program disclosed herein may be, for example, a program for causing a computer to execute the crystal analysis method disclosed herein. In addition, a preferred embodiment of the crystal analysis program disclosed herein may be, for example, the same as a preferred embodiment of the crystal analysis method disclosed herein.
[0118] The crystal analysis program disclosed herein can be created using various known programming languages depending on the configuration of the computer system used and the type and version of the operating system.
[0119] The crystal analysis program disclosed herein may be recorded on a recording medium such as an internal hard disk or an external hard disk, or may be recorded on a recording medium such as a CD-ROM, a DVD-ROM, an MO disk, or a USB memory. Furthermore, when the crystal analysis program disclosed in the present application is recorded on the above-mentioned recording medium, it can be used directly or by installing it on a hard disk via a recording medium reader possessed by a computer system, as necessary. The crystal analysis program disclosed in the present application may also be recorded in an external storage area (such as another computer) accessible from the computer system via an information and communication network. In this case, the optimization program disclosed in the present application recorded in the external storage area can be used directly or by installing it on a hard disk via an information and communication network from the external storage area, as necessary. The crystal analysis program disclosed herein may be divided into any number of processing units and recorded on a plurality of recording media.
[0120] (Computer-readable recording medium) The computer-readable recording medium disclosed herein records the crystal analysis program disclosed herein. The computer-readable recording medium disclosed herein is not particularly limited and can be appropriately selected depending on the purpose. Examples of the recording medium include an internal hard disk, an external hard disk, a CD-ROM, a DVD-ROM, an MO disk, and a USB memory. Furthermore, the computer-readable recording medium disclosed herein may be a plurality of recording media on which the crystal analysis program disclosed herein is recorded in a divided form for each arbitrary process.
[0121] An example of the technology disclosed herein will be described in further detail below using an example of the device configuration and a flow chart. FIG. 18 shows an example of the hardware configuration of the crystal analysis apparatus disclosed in the present application. In the crystal analysis apparatus 100, for example, a control unit 101, a main memory device 102, an auxiliary memory device 103, an I / O interface 104, a communication interface 105, an input device 106, an output device 107, and a display device 108 are connected via a system bus 109.
[0122] The control unit 101 performs calculations (arithmetic operations, comparison operations, calculations of the simulated annealing method, etc.), and controls the operation of hardware and software, etc. The control unit 101 may be, for example, a CPU (Central Processing Unit), or may be a part of an annealing machine used in the simulated annealing method, or may be a combination of these. The control unit 101 realizes various functions by, for example, executing a program (such as the crystal analysis program disclosed herein) loaded into the main storage device 102 or the like. The processes performed by the graph creation unit and the analysis unit in the crystal analysis apparatus disclosed herein can be performed by the control unit 101, for example.
[0123] The main memory device 102 stores various programs and also stores data necessary for executing the various programs. As the main memory device 102, for example, one having at least one of a Read Only Memory (ROM) and a Random Access Memory (RAM) can be used. The ROM stores various programs such as a BIOS (Basic Input / Output System), etc. The ROM is not particularly limited and can be appropriately selected according to the purpose, and examples thereof include a mask ROM and a PROM (Programmable ROM). The RAM functions as a working area in which various programs stored in, for example, the ROM or the auxiliary storage device 103 are deployed when they are executed by the control unit 101. There are no particular limitations on the RAM, and it can be appropriately selected depending on the purpose, and examples of the RAM include DRAM (Dynamic Random Access Memory) and SRAM (Static Random Access Memory).
[0124] The auxiliary storage device 103 is not particularly limited as long as it can store various information, and can be appropriately selected according to the purpose, and examples of the auxiliary storage device 103 include a solid state drive (SSD), a hard disk drive (HDD), etc. The auxiliary storage device 103 may also be a portable storage device such as a CD drive, a DVD drive, or a BD (Blu-ray (registered trademark) Disc) drive. Furthermore, the crystal analysis program disclosed herein is, for example, stored in the auxiliary storage device 103, loaded into the RAM (main memory) of the main storage device 102, and executed by the control unit 101.
[0125] The I / O interface 104 is an interface for connecting various external devices. The I / O interface 104 enables input and output of data from, for example, a CD-ROM (Compact Disc ROM), a DVD-ROM (Digital Versatile Disk ROM), an MO disk (Magneto-Optical disk), a USB memory (Universal Serial Bus (USB) flash drive), and the like.
[0126] There are no particular limitations on the communication interface 105, and any known interface may be used as appropriate, such as a wireless or wired communication device. The input device 106 is not particularly limited as long as it can accept various requests and information input to the structure exploration device 100, and any known device can be used as appropriate, such as a keyboard, a mouse, a touch panel, a microphone, etc. In addition, when the input device 106 is a touch panel (touch display), the input device 106 can also serve as the display device 108.
[0127] There are no particular limitations on the output device 107, and any known device may be used as appropriate, such as a printer. There are no particular limitations on the display device 108, and any known display device may be used as appropriate, such as a liquid crystal display or an organic EL display.
[0128] FIG. 19 shows another example of the hardware configuration of the crystal analysis apparatus disclosed in the present application. 19, crystal analysis apparatus 100 is divided into computer 200 that performs processes such as graph creation and crystal analysis, and annealing machine 300 that optimizes the Ising model formula (ground state search). In the example shown in FIG. 19, computer 200 and annealing machine 300 in crystal analysis apparatus 100 are connected by network 400. In the example shown in FIG. 19, for example, a CPU or the like can be used as the control unit 101a in the computer 200, and a device specialized for annealing can be used as the control unit 101b in the annealing machine 300.
[0129] In the example shown in FIG. 19, for example, a computer 200 performs a process of creating a graph, a process of analyzing a crystal, and various settings for defining an Ising model formula. Then, the weight (w ij ) and bias (b i ) is transmitted from the computer 200 to the annealing machine 300 via the network 400. Next, the annealing machine 300 anneals the received weights (w ij) and bias (b i ) to find the minimum value of the Ising model formula and the state of the bit that gives the minimum value. The annealing machine 300 then transmits the found minimum value of the Ising model formula and the state of the bit that gives the minimum value from the annealing machine 300 to the computer 200 via the network 400. Next, the computer 200 analyzes the crystal based on the state of the bit that gives the minimum value to the received Ising model equation.
[0130] FIG. 20 shows an example of the functional configuration of the crystal analysis apparatus disclosed in the present application. As shown in FIG. 20, crystal analysis device 100 includes a communication function section 120, an input function section 130, an output function section 140, a display function section 150, a storage function section 160, and a control function section 170.
[0131] The communication function unit 120 transmits and receives various data to and from an external device, for example. The communication function unit 120 may receive graph data, bias and weight data in the Ising model equation, and the like, from an external device, for example. The input function unit 130, for example, receives various instructions for the crystal analysis apparatus 100. The input function unit 130 may also receive input of, for example, graph data, bias and weight data in the Ising model formula, and the like. The output function unit 140 outputs, for example, data on the results of crystal analysis by printing. The display function unit 150 displays, for example, data on the results of crystal analysis on a display. The storage function unit 160 stores, for example, various programs, graph data, bias and weight data in the Ising model formula, and the like.
[0132] The control function unit 170 includes a graph creation unit 171 and an analysis unit 172 . The graph creation unit 171 performs, for example, a process of creating a graph in which one repeating unit lattice in a crystal is expanded to the repeating unit lattice and an adjacent repeating unit lattice adjacent to the one repeating unit lattice. Furthermore, when the number of intra-lattice nodes, which are data on anionic atoms bonded to cationic atoms in one repeating unit lattice in an ionic crystal, is n, the graph creation unit 171 performs a process of making the number of intra-lattice nodes, which are data on anionic atoms, n-1 or less. The analysis unit 172 performs processing such as analysis of the crystal (ionic crystal) using the graph created by the graph creation unit 171. More specifically, the analysis unit 172 performs processing such as calculating the similarity between the first crystal and the second crystal based on the first graph and the second graph, for example.
[0133] FIG. 21 shows an example of a flowchart for analyzing ionic crystals using an example of the technology disclosed in the present application.
[0134] First, the graph creation unit 171 acquires crystal structure data from a database (S1). Here, the database from which the crystal structure data is acquired is not particularly limited and can be appropriately selected depending on the purpose, and examples thereof include the Inorganic Crystal Structure Database (ICSD). The number and types of crystal structure data to be acquired are not particularly limited and can be appropriately selected depending on the purpose. For example, only crystal structure data of a chemical substance having a certain specific element may be acquired. Next, the graph creation unit 171 creates a graph based on the acquired crystal structure data (S2). Note that the flow of creating the graph in S2 will be described in detail later. Then, the graph creation unit 171 performs analysis of the crystal using the created graph (S3). As an example of the analysis of the crystal, there is the calculation of the similarity described above. The calculation of the similarity is performed, for example, by expressing the maximum independent set problem of the conflict graph in an Ising model formula and solving it with an annealing machine or the like. The calculation of the similarity can be performed, for example, by referring to the following non-patent literature. Non-patent document: Maritza Hernandez, Arman Zaribafiyan, Maliheh Aramon, Mohammad Naghibi “A Novel Graph-based Approach for Determining Molecular Similarity”. arXiv:1601.06693(https: / / arxiv.org / pdf / 1601.06693.pdf) In this manner, it is possible to perform an evaluation of the similarity of crystal structures, which is one type of crystal structure analysis.
[0135] 22 to 26 show an example of a flowchart for creating a graph by a method of treating a repeating crystal structure as a loop structure, using an example of the technology disclosed in the present application.
[0136] 22, the graph creation unit 171 determines whether the crystal to be analyzed is an ionic crystal (S100). In S100, if the graph creation unit 171 determines that the crystal to be analyzed is an ionic crystal, the process proceeds to S101, and if the graph creation unit 171 determines that the crystal to be analyzed is not an ionic crystal, the process ends. Next, the graph creation unit 171 reads the crystal structure data (S101). Next, the graph creation unit 171 calculates the relative coordinates of all atoms in the single unit lattice (S102). More specifically, in S102, the graph creation unit 171 calculates the relative coordinates of all atoms in the single unit lattice to obtain the relative positional relationship of all atoms.
[0137] Then, the graph creating unit 171 sets all atoms in the single unit lattice as nodes of the graph (S103). Moreover, the nodes prepared in S103 are nodes within the lattice in the repeating unit lattice. Here, S104 and S116 mean that the processes from S105 to S115 described later are performed (repeated) for all atoms in a single unit lattice that correspond to a certain atom A. Similarly, S105 and S108 mean that the processes from S106 to S107 described later are performed (repeated) for all combinations of a certain atom (atom A) and another atom B. Furthermore, the graph creation unit 171 judges whether or not the distance between A and B is appropriate as a bond distance for a combination (A, B) of two atoms, a certain atom (atom A) and another atom (atom B) other than the certain atom, among all atoms in a single unit lattice (S106). Then, if the distance between A and B is appropriate as a bond distance, the graph creation unit 171 bonds node A and node B (S107). That is, in S107, the graph creation unit 171 provides an edge corresponding to node A and node B. Furthermore, the edge provided in S107 is an intra-lattice edge in the repeating unit lattice. Moreover, if the distance between A and B is not appropriate as the bond distance, the graph creation unit 171 shifts the process to S108.
[0138] Next, in FIG. 23, S109 and S115 mean that the processes from S110 to S114 described below are performed (repeated) for all combinations of a certain atom (atom A) and another atom B'. Then, the graph creation unit 171 judges whether or not a bond equivalent to the bond A-B' already exists between an atom (atom A) in a single unit lattice and an arbitrary atom B' in an adjacent unit lattice (S110). If a bond equivalent to the bond A-B' does not yet exist in S110, the graph creation unit 171 judges whether or not the distance between A and B' is appropriate as a bond distance (S111).
[0139] Furthermore, if the distance between A and B' is appropriate as a bond distance, the graph creation unit 171 judges whether or not a node corresponding to B' already exists (S112). Thereafter, if the graph creation unit 171 judges that a node corresponding to B' does not yet exist, it adds atom B' as an extended node (S113). Next, after adding atom B' as an extended node, the graph creation unit 171 bonds node A and node B' (S114). That is, in S114, the graph creation unit 171 provides an edge corresponding to node A and node B'. Moreover, if the graph creation unit 171 determines in S112 that a node corresponding to B' already exists, it moves the process to S114 and joins node A and node B'. Here, when the graph creation unit 171 determines in S110 that there is no bond equivalent to the bond A-B', when it determines in S111 that the distance between A and B' is not appropriate as a bond distance, or when it has performed the processing of S114, it shifts the processing to S115 so that the processing of S110 to S114 is performed (repeated) for all combinations of a certain atom (atom A) and another atom B'.
[0140] Next, in FIG. 24, S117 and S120 mean that the processes of S118 and S119, which will be described later, are performed (repeated) for all of the extended nodes. Then, the graph creation unit 171 identifies a node A' in the unit lattice that is equivalent to the extended node B' (S118). Next, graph creation unit 171 regards extended node B' as node A' (S119). That is, in S119, graph creation unit 171 regards extended node B' as node A', and thereby sets the edge corresponding to the node A-node A' as a loop edge.
[0141] 25, S121 and S126 mean that the processes of S122 to S125 described later are performed (repeated) in all of the cation nodes A'. Similarly, S122 and S125 mean that the processes of S123 and S124 described later are performed (repeated) in all of the cation nodes B' other than the cation node A'. The graph creation unit 171 calculates n C′ is calculated (S123). Next, the graph creation unit 171 calculates the bond order between the cation node A′ and the cation node B′ by n C′ That is, in S124, the graph creation unit 171 specifies the bond order between the cation node A' and the cation node B' based on the number of the anion nodes C'.
[0142] Next, the graph creation unit 171 deletes bonds (edges) between all cation nodes and anion nodes (S127). Furthermore, the graph creation unit 171 deletes all anion nodes (S128). That is, in S128, the graph creation unit 171 sets the number of nodes in the lattice, which are data of anionic atoms, to 0. By doing so, in the examples shown in FIGS. 22 to 26, the number of nodes included in the repeating unit lattice can be reduced, and the number of bits required when analyzing an ionic crystal can be further reduced.
[0143] 26, the graph creation unit 171 calculates the total number of nodes (m) (S129). That is, in S129, the graph creation unit 171 counts the number of cation nodes in the unit lattice to identify the total number of nodes (m). Next, the graph creation unit 171 sets the ideal number of nodes in the graph as n, calculates the value of n / m, specifies the number obtained by rounding off the value of n / m, and integrally multiplies and copies the graph by the specified number (S130). That is, in S130, the graph creation unit 171 integrally multiplies the graph so as to approach the ideal number of nodes n in the graph. Note that the ideal number of nodes n in the graph can be appropriately selected depending on the ionic crystal to be targeted.
[0144] Next, the graph creation unit 171 outputs the graph data (S131). Then, when the graph creation unit 171 has completed outputting the graph data, it ends the process of creating the graph.
[0145] 27 to 29 show an example of a flowchart for creating a graph by a method of expanding and handling a unit lattice using an expansion node, using an example of the technology disclosed in the present application. In the method of expanding and handling a unit lattice using an expansion node, the same processes can be performed from S100 to S128 in the flow described with reference to FIGS. 22 to 26, and therefore the description thereof will be omitted.
[0146] In the method of expanding and handling a unit lattice using an expansion node, the graph creation unit 171 calculates the total number of nodes (m) in the same manner as S129 in Fig. 26, as shown in Fig. 27 (S229). Next, that is, in S129, the graph creation unit 171 counts the number of cation nodes in the unit lattice to identify the total number of nodes (m). Next, as shown in FIG. 27, similar to S130 in FIG. 26, the graph creation unit 171 sets the ideal number of nodes in the graph to n, calculates the value of n / m, determines the number obtained by rounding off the value of n / m, and duplicates the graph by an integer multiple the determined number of times (S230).
[0147] In FIG. 28, S231 and S234 mean that the processes of S232 and S233, which will be described later, are performed (repeated) for all of the catenary nodes A'. The graph creation unit 171 judges whether or not there is a node that is in a bond relationship with the atom corresponding to A' among the atoms of the adjacent unit lattice that is not the subject of the graph (S232). In S232, if the graph creation unit 171 judges that there is a node that is in a bond relationship with the atom corresponding to A', it shifts the process to S233, and if it judges that there is no node that is in a bond relationship with the atom corresponding to A', it shifts the process to S234. Next, the graph creation unit 171 designates the atom determined in S232 to be a node in a bond relationship with the atom corresponding to A' as node B', and adds it to list B''. Next, the graph creation unit 171 deletes duplicate elements from among the elements (nodes) of list B'' (S235). Next, the graph creation unit 171 adds the nodes included in list B'' to the graph (S236).
[0148] In FIG. 29, S237 and S239 mean that the process of S238 described below is performed (repeated) for all of the nodes B' included in the list B''. For each node C' in the graph that has a connection relationship with node B', the graph creation unit 171 adds a connection (edge) between node B' and node C' (S238). Next, the graph creation unit 171 outputs the graph data (S240). Then, when the graph creation unit 171 has completed outputting the graph data, it ends the process of creating the graph.
[0149] 21 to 29, the process flow in the example of the technology disclosed herein has been described in a specific order, but in the technology disclosed herein, the order of each step can be changed as appropriate within the scope of technical feasibility. Also, in the technology disclosed herein, multiple steps can be performed together within the scope of technical feasibility.
[0150] An example of an annealing method and an annealing machine will be described below. Simulated annealing is a method for finding a solution probabilistically using random values and superposition of quantum bits. In the following, we will use the problem of minimizing the value of the evaluation function to be optimized as an example, and refer to the value of the evaluation function as energy. To maximize the value of the evaluation function, we simply change the sign of the evaluation function.
[0151] First, starting from the initial state where each variable is assigned one of the discrete values, a state closest to the current state (combination of variable values) is selected (for example, a state where only one variable is changed), and the state transition is considered. The change in energy for that state transition is calculated, and depending on that value, a probabilistic decision is made as to whether to adopt that state transition and change the state, or not adopt it and maintain the original state. If the probability of adopting a state transition that decreases energy is selected to be greater than that of an energy increase, then on average, the state will change in the direction of decreasing energy, and it is expected that the state will transition to a more appropriate state over time. As a result, it is possible to ultimately obtain an optimal solution or an approximate solution that gives energy close to the optimal value. If we were to deterministically accept a state when its energy decreases and reject it when its energy increases, the change in energy would be a monotonically decreasing energy over time, but once a local optimum is reached, no further change would occur. As mentioned above, there are a very large number of local optimums in discrete optimization problems, so the state will almost certainly be trapped in a local optimum that is not very close to the optimum value. Therefore, when solving discrete optimization problems, it is important to probabilistically decide whether or not to accept a state.
[0152] In the simulated annealing method, it has been proven that if the adoption (admission) probability of a state transition is determined as follows, the state will reach the optimal solution in the limit of infinity in time (number of iterations). The following describes in a step-by-step manner how to find an optimal solution using simulated annealing.
[0153] (1) For the energy change (energy decrease) value (-ΔE) accompanying a state transition, the allowable probability p of that state transition is determined by one of the following functions f().
[0154]
number
[0155]
number
[0156]
number
[0157] Here, T is a parameter called the temperature value, which can be varied, for example, as follows:
[0158] (2) The temperature value T is logarithmically decreased with the number of iterations t as expressed by the following equation:
[0159]
number
[0160] Here, T0 is the initial temperature value and should be set sufficiently large depending on the problem. When using the admissible probability expressed by equation (1), if a steady state is reached after sufficient iterations, the occupation probability of each state follows the Boltzmann distribution for thermal equilibrium in thermodynamics. Furthermore, since the probability of occupying a low-energy state increases when the temperature is gradually lowered from a high temperature, it is believed that a low-energy state can be obtained when the temperature is sufficiently lowered. This process is very similar to the state change that occurs when a material is annealed, so this method is called the annealing method (or quasi-annealing method). The fact that state transitions that increase energy occur probabilistically corresponds to thermal excitation in physics.
[0161] An example of the functional configuration of an annealing machine that performs the simulated annealing method is shown in Fig. 30. Note that, although the following description will also discuss a case in which multiple state transition candidates are generated, the basic simulated annealing method generates transition candidates one by one.
[0162] The annealing machine 300 has a state storage unit 111 that stores the current state S (values of a plurality of state variables). The annealing machine 300 also has an energy calculation unit 112 that calculates an energy change value {-ΔEi} of each state transition when a state transition from the current state S occurs due to a change in any of the values of the plurality of state variables. The annealing machine 300 also has a temperature control unit 113 that controls a temperature value T, and a transition control unit 114 for controlling the state change. The annealing machine 300 can be a part of the structure searching device 100 described above.
[0163] The transition control unit 114 probabilistically determines whether or not to accept one of a plurality of state transitions based on the temperature value T, the energy change value {-ΔEi}, and a random value, depending on the relative relationship between the energy change value {-ΔEi} and the thermal excitation energy.
[0164] Here, the transition control unit 114 has a candidate generation unit 114a that generates state transition candidates, and a permission / prohibition determination unit 114b that probabilistically determines whether or not to permit a state transition for each candidate based on the energy change value {-ΔEi} and the temperature value T. Furthermore, the transition control unit 114 has a transition determination unit 114c that determines which candidate to adopt from the permitted candidates, and a random number generation unit 114d that generates random variables.
[0165] The operation of the annealing machine 300 in one iteration is as follows. First, the candidate generation unit 114a generates one or more candidates (candidate number {Ni}) for state transition from the current state S held in the state holding unit 111 to the next state. Next, the energy calculation unit 112 calculates the energy change value {-ΔEi} for each of the state transition candidates using the current state S and the state transition candidates. The possibility determination unit 114b uses the temperature value T generated by the temperature control unit 113 and the random variable (random value value) generated by the random number generation unit 114d to permit the state transition with the permissible probability of the above formula (1) according to the energy change value {-ΔEi} of each state transition. Then, the possibility determination unit 114b outputs the possibility {fi} of each state transition. If there are multiple permitted state transitions, the transition determination unit 114c randomly selects one of them using a random value. Then, the transition determination unit 114c outputs the transition number N of the selected state transition and the possibility f of the transition. If there is a permitted state transition, the value of the state variable stored in the state holding unit 111 is updated according to the adopted state transition.
[0166] Starting from the initial state, the above iterations are repeated while lowering the temperature value by the temperature control unit 113, and the operation ends when a certain number of iterations is reached or an end determination condition is met, such as the energy falling below a certain value. The answer output by the annealing machine 300 is the state at the end.
[0167] 30 may be realized by using, for example, a semiconductor integrated circuit. For example, the transition control unit 114 may include a random number generating circuit that functions as the random number generating unit 114d, a comparison circuit that functions as at least a part of the feasibility determining unit 114b, a noise table (described later), and the like.
[0168] Regarding the transition control unit 114 shown in FIG. 30, a mechanism for permitting a state transition with the permissible probability expressed by equation (1) will be described in further detail.
[0169] A circuit that outputs 1 with probability p and 0 with probability (1 - p) has two inputs, A and B. It can be realized by inputting probability p to input A of a comparator that outputs 1 when A > B and 0 when A < B, and inputting a uniform random number taking values in the interval [0, 1) to input B. Therefore, if the value of probability p calculated using equation (1) based on the energy change value and temperature value T is input to input A of this comparator, the above function can be realized.
[0170] That is, when f is the function used in equation (1) and u is a uniform random number taking values in the interval [0, 1), the above function can be realized by a circuit that outputs 1 when f(ΔE / T) is greater than u.
[0171] Also, the same function as the above can be realized even if the following transformation is performed. Applying the same monotonically increasing function to two numbers does not change their magnitude relationship. Therefore, the output does not change even if the same monotonically increasing function is applied to the two inputs of the comparator. As this monotonically increasing function, the inverse function f -1 is adopted, and it can be seen that a circuit can be formed that outputs 1 when -ΔE / T is greater than f -1 (u). Furthermore, since the temperature value T is positive, it can be seen that a circuit that outputs 1 when -ΔE is greater than Tf -1 (u) is sufficient. The transition control unit 114 in FIG. 30 is a conversion table for realizing the inverse function f -1 (u), and may include a noise table that outputs the values of the following function for the discretized input in the interval [0, 1).
[0172]
Number
[0173]
Number
[0174] Fig. 31 is a diagram showing an example of an operation flow of transition control unit 114. The operation flow shown in Fig. 31 includes a step (S0001) of selecting one state transition as a candidate, a step (S0002) of determining whether or not the state transition is possible by comparing the product of an energy change value, a temperature value, and a random value for the state transition, and a step (S0003) of adopting the state transition if it is possible, and not adopting it if it is not possible. EXAMPLES
[0175] An embodiment of the technology disclosed in the present application will be described below, but the technology disclosed in the present application is not limited to this embodiment.
[0176] Example 1 As Example 1, the similarity in ionic crystals was calculated using an example of the crystal analysis apparatus disclosed in the present application. The calculation of the similarity of the ionic crystals in Example 1 was performed using a crystal analysis device having a hardware configuration as shown in Fig. 19, following the flow of the flowcharts in Fig. 21 to Fig. 26. Note that in Example 1, S129 and S130 in Fig. 26 were not performed (the graph was not multiplied by an integer), and a graph was created to calculate the similarity.
[0177] More specifically, in Example 1, the graph was expanded using a method of treating the repeating structure of the ionic crystal as a loop structure, and the following formula (1) was minimized by the annealing method using a Digital Annealer (registered trademark) as an annealing machine.
number
[0178] Furthermore, in the first embodiment, the similarity was calculated for the obtained maximum independent set by using the following formula (2).
number
[0179] Here, in Example 1, the similarity was calculated for ionic crystal A, whose repeating unit lattice contains 76 atoms, and ionic crystal B, whose repeating unit lattice contains 96 atoms. The structure of the repeating unit lattice of ionic crystal A is shown in FIG. 15A, and the structure of the repeating unit lattice of ionic crystal B is shown in FIG. 15B. Here, as shown in Figure 15A, ionic crystal A has 8 lithium (Li), 4 manganese (Mn), 16 phosphorus (P), and 48 oxygen (O) atoms in the unit lattice, totaling 76 atoms. Similarly, as shown in Figure 15B, ionic crystal B has 16 lithium (Li), 8 manganese (Mn), 16 phosphorus (P), and 56 oxygen (O) atoms in the unit lattice, totaling 96 atoms. Ionic crystal A is a crystal of "ICSD#253225" with the composition formula "Li2MnP2O7", and ionic crystal B is a crystal of "ICSD#248405" with the composition formula "Li2Mn(PO3)4".
[0180] In Example 1, a conflict graph was created after deleting 48 oxygen atom nodes contained in the repeating unit lattice of ionic crystal A and deleting 56 oxygen atom nodes contained in the repeating unit lattice of ionic crystal B. That is, in Example 1, 48 oxygen atom nodes were deleted from ionic crystal A, which had 76 nodes, to reduce the number of nodes to 28, and 56 oxygen atom nodes were deleted from ionic crystal B, which had 96 nodes, to reduce the number of nodes to 40.
[0181] FIG. 32 shows a summary of the results of Example 1 and Comparative Example 1. In the first embodiment, the number of nodes in the conflict graph (the number of bits required for calculation) is "8×16=128" for lithium nodes, "4×8=32" for manganese nodes, and "16×16=256" for phosphorus nodes. Therefore, in the first embodiment, the number of nodes in the conflict graph is "128+32+256=416". On the other hand, when the similarity was calculated without deleting the nodes of the oxygen atoms contained in the unit lattices of ionic crystal A and ionic crystal B (comparative example to Example 1; Comparative Example 1), the number of nodes required to calculate the similarity was "3104", as described above, which was a much larger number than in Example 1.
[0182] Furthermore, in Example 1, the number of atoms in the common substructure of ionic crystal A and ionic crystal B was "18", and the similarity based on the above formula (2) was calculated to be "0.546". On the other hand, in the comparative example (Comparative Example 1) for Example 1, the number of atoms in the common substructure was "55", and the similarity based on the above formula (2) was calculated to be "0.648". From these results, it was confirmed that the difference in similarity between Example 1 and Comparative Example 1 was about 0.1, which is a relatively close value.
[0183] Example 2 In Example 2, the similarity between ionic crystal A and ionic crystal B was calculated in the same manner as in Example 1, except that S129 and S130 in Figure 26 were performed and the graph was multiplied by an integer before calculating the similarity. Specifically, for ionic crystal A and ionic crystal B, in which all oxygen atom nodes had been removed, in order to approximate the number of nodes contained in the unit lattice of each crystal, the unit lattice of ionic crystal A was doubled to make the number of nodes "80", and the unit lattice of ionic crystal B was tripled to make it "84", and the similarity was calculated.
[0184] FIG. 33 shows a summary of the results of Example 2. In Example 2, for ionic crystal A and ionic crystal B, the number of bits required for calculation was "2496", the number of atoms in the common substructure was "43", and the similarity based on the above formula (2) was calculated to be "0.525". This similarity is calculated by aligning the number of nodes of the unit lattices in each crystal, and therefore can be considered to have higher accuracy than the similarity in Example 2. In addition, in Example 2, as described above, the number of bits required for the calculation was "2496", and the similarity could be calculated with a smaller number of bits than in Comparative Example 1, in which the number of nodes of the unit lattice in each crystal was made uniform and the oxygen atoms were not deleted.
[0185] Example 3 In Example 3, the ionic crystals to be calculated for similarity were crystals (six types) whose composition formula was Li2MP2O7 (M = Fe, Co, Mn, Ba), and the similarity was calculated in the same manner as in Example 1, except that the oxygen atom nodes were deleted. As a comparative example corresponding to Example 3, the similarity was calculated without deleting the oxygen atom nodes (Comparative Example 2).
[0186] FIG. 34 shows the relationship between the type of ionic crystal and the calculated similarity in Example 3 and Comparative Example 2. Here, among the six types of ionic crystals shown in FIG. 34, the crystals of "M=Ba" (ICSD#107724 and ICSD#41433) are crystals having a different type of structure from the other crystals. 34, in Example 3, the similarity between crystals in which M is any of Fe, Co, and Mn is high, and the similarity between crystals in which M is Ba is also high. On the other hand, the similarity between crystals in which M is any of Fe, Co, and Mn and crystals in which M is Ba is low, and it is understood that the crystal structures are correctly classified according to their types.
[0187] FIG. 35 shows the relationship between the type of ionic crystal and the number of bits required for calculating the similarity in Example 3 and Comparative Example 2. 35, the number of bits required for calculating the similarity was 656 or more and 832 or less in Example 3, and was 3792 or more and 3968 or less in Comparative Example 2. From this, it was confirmed that in Example 3, the number of bits required for calculation can be reduced, and a highly accurate similarity that can be correctly classified according to the type of crystal structure can be calculated.
[0188] Example 4 In Example 4, the ionic crystals to be calculated for similarity were crystals represented by the respective composition formulas of CsCl, KCl, NaCl, NaF, LiF, and LiCl, and the similarity was calculated in the same manner as in Example 3, except that halogen atoms (Cl or F), which are an example of anionic atoms, were deleted. In the ionic crystals used in Example 4, CsCl is ICSD#257256, KCl is ICSD#28938, NaCl is ICSD#28048, NaF is ICSD#29128, LiF is ICSD#44879, and LiCl is ICSD#52235. In Example 4, two lattices were graphed in the a-axis, b-axis, and c-axis directions for CsCl (ICSD#257256), and two lattices were graphed in the a-axis and b-axis directions for the other crystals, and one lattice was graphed in the c-axis direction for the other crystals. The number of nodes in each graph was 8.
[0189] FIG. 36 shows the relationship between the type of ionic crystal and the calculated similarity in Example 4. Here, among the six types of ionic crystals shown in FIG. 36, CsCl (ICSD#257256) has a simple cubic lattice, and the other crystals have face-centered cubic lattices. As shown in Figure 36, in Example 4, the similarity between crystals of face-centered cubic lattice is high, while the similarity between crystals of simple cubic lattice and crystals of face-centered cubic lattice is low, indicating that the crystals were correctly classified according to the type of crystal structure.
[0190] The following supplementary notes are further disclosed regarding the above embodiment. (Appendix 1) In the crystal, a graph creation step of creating a graph in which the one repeating unit cell, which includes an intra-lattice node indicating data of an atom in the one repeating unit cell and an intra-lattice edge indicating data of a chemical bond between two atoms in the one repeating unit cell, is expanded to the one repeating unit cell and an adjacent repeating unit cell adjacent to the one repeating unit cell; an analysis step of analyzing the crystal using the graph; Including, In the graph creation step, when the crystal is an ionic crystal, when the number of nodes in the lattice which are data on anionic atoms bonded to a cationic atom in one repeating unit lattice in the ionic crystal is n, the number of nodes in the lattice which are data on the anionic atoms is n-1 or less. A crystal analysis method comprising: (Appendix 2) 2. The crystal analysis method according to claim 1, wherein in the graph creation step, the number of nodes in the lattice which are data on the anionic atoms is set to 0. (Appendix 3) 3. The crystal analysis method according to claim 2, wherein in the graph creation step, the number of nodes in the lattice, which are data of the anionic atoms, is set to 0, and when the cationic atoms in one repeating unit lattice are regarded as being directly chemically bonded to each other, the bond order of the chemical bonds between the cationic atoms is set to the number of chemical bonds between the cationic atoms via the anionic atoms, which are regarded as not existing. (Appendix 4) The graph created in the graph creation step is an intra-lattice node representing data of an atom in a repeating unit cell; and an intra-lattice edge representing data of a chemical bond between two atoms in the repeating unit cell; The data includes an extended node, which is data of an atom in an adjacent repeating unit lattice adjacent to the one repeating unit lattice and is chemically bonded to the atom in the one repeating unit lattice, and an extended edge, which is data of a chemical bond between an atom corresponding to the intra-lattice node in the one repeating lattice unit and an atom corresponding to the extended node. 4. A crystal analysis method according to any one of claims 1 to 3. (Appendix 5) The graph created in the graph creation step is an intra-lattice node representing data of an atom in a repeating unit cell; and an intra-lattice edge representing data of a chemical bond between two atoms in the repeating unit cell; a loop graph including an atom Y in the one repeating unit lattice corresponding to the atom Y' in an adjacent repeating unit lattice adjacent to the one repeating unit lattice, and a loop edge showing data of a virtual chemical bond between the atom X and the atom Y in the one repeating unit lattice, the atom Y' being chemically bonded to the atom X in the one repeating unit lattice, 4. A crystal analysis method according to any one of claims 1 to 3. (Appendix 6) In the graph creation step, a first graph for a first crystal and a second graph for a second crystal are created; In the analyzing step, a similarity between the first crystal and the second crystal is calculated based on the first graph and the second graph. 6. A crystal analysis method according to any one of claims 1 to 5. (Appendix 7) 7. The crystal analysis method according to claim 6, wherein the calculation of the similarity in the analysis step is performed after at least one of the one repeating unit lattice for the first graph and the one repeating unit lattice for the second graph is integer-multiplied so that the numbers of nodes in the lattice in the first graph and the second graph are approximate to each other. (Appendix 8) 8. The crystal analysis method according to claim 6, wherein the calculation of the similarity in the analysis step is performed by searching for a maximum independent set of a conflict graph between the first graph and the second graph using the following formula (1):
number
number
number
number
number
number
[0191] 10X Crystal 10 Unit Cell 100 Crystal analyzer 101 Control section 102 Main storage 103 Auxiliary storage device 104 I / O Interface 105 Communication Interface 106 Input Devices 107 Output Device 108 Display device 109 Bus 120 Communication function unit 130 Input function unit 140 Output function section 150 Display function section 160 Memory function unit 170 Control Function Section 171 Graph Creation Department 172 Analysis Department
Claims
1. In the crystal, a graph creation step of creating a graph in which the one repeating unit cell, which includes an intra-lattice node indicating data of an atom in the one repeating unit cell and an intra-lattice edge indicating data of a chemical bond between two atoms in the one repeating unit cell, is expanded to the one repeating unit cell and an adjacent repeating unit cell adjacent to the one repeating unit cell; an analysis step of analyzing the crystal using the graph; [0033] In the graph creation step, when the crystal is an ionic crystal, when the number of nodes in the lattice which are data on anionic atoms bonded to a cationic atom in one repeating unit lattice in the ionic crystal is n, the number of nodes in the lattice which are data on the anionic atoms is n-1 or less. A crystal analysis method comprising:
2. 2. The crystallography method according to claim 1, wherein in said graph creation step, the number of said intra-lattice nodes which are data of said anionic atoms is set to 0.
3. 3. The crystal analysis method according to claim 2, wherein in the graph creation step, when the number of nodes in the lattice, which are data of the anionic atoms, is set to 0, and the cationic atoms in one repeating unit lattice are regarded as being directly chemically bonded to each other, the bond order of the chemical bonds between the cationic atoms is set to the number of chemical bonds between the cationic atoms via the anionic atoms that are regarded as not existing.
4. The graph created in the graph creation step is the intra-lattice nodes representing data of atoms in the one repeating unit cell, and the intra-lattice edges representing data of chemical bonds between two atoms in the one repeating unit cell; The data includes an extended node, which is data of an atom in an adjacent repeating unit lattice adjacent to the one repeating unit lattice and is chemically bonded to the atom in the one repeating unit lattice, and an extended edge, which is data of a chemical bond between the atom corresponding to the intra-lattice node in the one repeating lattice unit and the atom corresponding to the extended node. The crystal analysis method according to any one of claims 1 to 3.
5. The graph created in the graph creation step is the intra-lattice nodes representing data of atoms in the one repeating unit cell, and the intra-lattice edges representing data of chemical bonds between two atoms in the one repeating unit cell; a loop graph including an atom X in the one repeating lattice unit and an atom Y' chemically bonded to the atom X, the atom Y in the one repeating unit lattice corresponding to the atom Y' in an adjacent repeating unit lattice adjacent to the one repeating unit lattice, and a loop edge showing data of a virtual chemical bond between the atom X and the atom Y in the one repeating unit lattice, The crystal analysis method according to any one of claims 1 to 3.
6. In the graph creation step, a first graph for a first crystal and a second graph for a second crystal are created; In the analyzing step, a similarity between the first crystal and the second crystal is calculated based on the first graph and the second graph. The crystal analysis method according to any one of claims 1 to 5.
7. 7. The crystal analysis method according to claim 6, wherein the calculation of the similarity in the analysis step is performed after at least one of the one repeating unit lattice for the first graph and the one repeating unit lattice for the second graph is integer-multiplied so that the numbers of intra-lattice nodes in the first graph and the second graph are approximate to each other.
8. 8. The crystal analysis method according to claim 6, wherein the calculation of the similarity in the analysis step is performed by searching for a maximum independent set of a conflict graph between the first graph and the second graph using the following formula (1): [0010] However, in the formula (1), E is a Hamiltonian, minimizing E means searching for the maximal independent set, Said c' ij is a coefficient representing a weight for the bond degree between the i-th node and the j-th node in the conflict graph of the first graph and the second graph, The x i is a binary variable that indicates that the i-th node is 0 or 1, The x j is a binary variable representing that the j-th node is 0 or 1, The α and β are positive numbers.
9. 9. The crystallography method according to claim 8, wherein the search for the maximal independent set is performed by minimizing a Hamiltonian in the formula (1) by a simulated annealing method.
10. 10. The crystal analysis method according to claim 6, wherein the similarity in the analysis step is calculated for the largest independent set found by using the following formula (2): [0025] In the above formula (2), The S(G A , G B ) represents the similarity between the first crystal represented as the first graph and the second crystal represented as the second graph, and is expressed as 0 to 1, with the closer to 1 the similarity is, the higher the similarity is; The V A represents the total number of node atoms in the first crystal represented as the first graph, The V c A represents the number of node atoms included in a maximum independent set of a conflict graph among the node atoms in the first crystal represented as the first graph, The V B represents the total number of node atoms in the second crystal represented as the second graph, The V c B represents the number of node atoms included in a maximum independent set of a conflict graph among the node atoms in the second crystal represented as the second graph, The δ is a number from 0 to 1.
11. In the crystal, a graph creating unit that creates a graph in which the one repeating unit lattice, which includes an intra-lattice node that indicates data of an atom in the one repeating unit lattice and an intra-lattice edge that indicates data of a chemical bond between two atoms in the one repeating unit lattice, is expanded to an adjacent repeating unit lattice adjacent to the one repeating unit lattice; an analysis unit that performs crystal analysis using the graph; having the graph creation unit, when the crystal is an ionic crystal, sets the number of the intra-lattice nodes which are data of anionic atoms bonded to a cationic atom in one repeating unit lattice in the ionic crystal to n-1 or less when the number of the intra-lattice nodes which are data of the anionic atoms is n; A crystal analysis apparatus characterized by:
12. In the crystal, a graph creation step of creating a graph in which the one repeating unit cell, which includes an intra-lattice node indicating data of an atom in the one repeating unit cell and an intra-lattice edge indicating data of a chemical bond between two atoms in the one repeating unit cell, is expanded to the one repeating unit cell and an adjacent repeating unit cell adjacent to the one repeating unit cell; an analysis step of analyzing the crystal using the graph; Including, In the graph creation step, when the crystal is an ionic crystal, when the number of nodes in the lattice which are data on anionic atoms bonded to a cationic atom in one repeating unit lattice in the ionic crystal is n, the number of nodes in the lattice which are data on the anionic atoms is n-1 or less. A crystal analysis program characterized by causing a computer to carry out processing.
Citation Information
Patent Citations
Device, method, and program for predicting material properties
JP2020194488A
