DEVICE, METHOD, AND PROGRAM FOR PREDICTING OXYGEN STORAGE CAPACITY OF Ce-Zr-BASED COMPOSITE OXIDE AFTER HEAT TREATMENT

Abstract

To provide a device, a method, and a program that accurately predict oxygen storage capacity of a Ce-Zr-based composite oxide containing additive elements after a heat treatment.SOLUTION: A prediction device 1 for estimating an oxygen storage capacity of a Ce-Zr-M0-based composite oxide (hereinafter referred to as oxide) after a heat treatment includes: a first learned model 2; a second learned model 3; a first prediction unit 4 which uses the first learned model 2 and information about first characteristics of the oxide to acquire information about whether the oxide undergoes phase separation due to the heat treatment; a determination unit 5 which determines whether the information acquired by the first prediction unit 4 is information indicating that the oxide does not undergo phase separation due to the heat treatment; and a second prediction unit 6 which acquires information about the oxygen storage capacity after the heat treatment of the oxide by using the second learned model 3 and the information about second characteristics of the oxide when the information acquired by the first prediction unit 4 is the information indicating that the oxide does not undergo phase separation due to the heat treatment.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to an apparatus, a method and a program for predicting the oxygen storage capacity of a Ce-Zr based composite oxide after heat treatment. [Background technology]

[0002] Exhaust gas emitted from internal combustion engines of automobiles, motorcycles, etc. contains harmful components such as hydrocarbons (HC), carbon monoxide (CO), nitrogen oxides (NOx), etc. Three-way catalysts, which have catalytic activity to oxidize HC to water and carbon dioxide, oxidize CO to carbon dioxide, and reduce NOx to nitrogen, are used as exhaust gas purification catalysts to purify and detoxify these harmful components.

[0003] In order to mitigate fluctuations in the oxygen concentration in exhaust gas and efficiently purify HC, CO, NOx, etc., materials having oxygen storage capacity (OSC), such as Ce-Zr composite oxides, are used as constituent materials of exhaust gas purification catalysts.

[0004] The oxygen storage capacity is the capacity to store oxygen when the oxygen concentration in the exhaust gas is high and to release oxygen when the oxygen concentration in the exhaust gas is low.

[0005] The Ce-Zr composite oxide is an oxide containing Ce and Zr elements. The Ce-Zr composite oxide has an oxygen storage capacity, and reduces the fluctuation of the oxygen concentration in the exhaust gas, thereby widening the operating window of the catalytically active component. Since the exhaust gas purification catalyst is exposed to a high-temperature environment, it is preferable that the Ce-Zr composite oxide has a high oxygen storage capacity even after being exposed to the high-temperature environment.

[0006] The Ce element in the Ce-Zr composite oxide contributes to the absorption and release of oxygen in the Ce-Zr composite oxide through the change in valence between trivalent and tetravalent. The Zr in the Ce-Zr composite oxide contributes to improving the heat resistance of the Ce-Zr composite oxide and stabilizing the structure of the Ce-Zr composite oxide.

[0007] When Ce-Zr based composite oxides contain additive elements (elements other than Ce, Zr, and O), the additive elements affect various performance properties of the Ce-Zr based composite oxides. Therefore, attempts have been made to add additive elements to Ce-Zr based composite oxides in order to improve various performance properties, particularly the oxygen storage capacity after heat treatment.

[0008] If it were possible to predict the oxygen storage capacity of a Ce-Zr-based composite oxide containing an additive element after heat treatment, it would be possible to efficiently search for a desirable Ce-Zr-based composite oxide.

[0009] Patent Document 1 describes an apparatus, a method, and a program for predicting material properties of a target material. [Prior art documents] [Patent documents]

[0010] [Patent Document 1] Patent No. 7190615 Summary of the Invention [Problem to be solved by the invention]

[0011] In the conventional technology, it is difficult to accurately predict the oxygen storage capacity of a Ce-Zr based composite oxide containing an additive element after heat treatment.

[0012] Therefore, an object of the present invention is to provide an apparatus, method and program capable of accurately predicting the oxygen storage capacity of a Ce-Zr based composite oxide containing an additive element after heat treatment. [Means for solving the problem]

[0013] The present invention provides the following apparatus, method, and program. [1] An apparatus for predicting the oxygen storage capacity of a Ce-Zr-M0-based composite oxide (wherein M0 represents an additive element) after heat treatment, comprising: a first trained model generated by machine learning using information regarding a first characteristic of each composite oxide in a first composite oxide group as an explanatory variable and information regarding whether each composite oxide in the first composite oxide group undergoes phase separation upon heat treatment as a target variable, wherein the first composite oxide group includes a plurality of Ce-Zr-M1-based composite oxides that do not undergo phase separation upon heat treatment and a plurality of Ce-Zr-M1-based composite oxides that undergo phase separation upon heat treatment, each M1 independently representing an additive element; a second trained model generated by machine learning using information on a second property of each complex oxide in a second complex oxide group as an explanatory variable and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment as a target variable, wherein the second complex oxide group includes a plurality of Ce-Zr-M2-based complex oxides that do not undergo phase separation upon heat treatment, but does not include material properties of Ce-Zr-M2-based complex oxides that undergo phase separation upon heat treatment, and each M2 independently represents an additive element; a first prediction unit that obtains information regarding whether or not the Ce-Zr-M0-based composite oxide undergoes phase separation by a heat treatment, using the first trained model and information regarding a first characteristic of the Ce-Zr-M0-based composite oxide; a determination unit that determines whether or not the information acquired by the first prediction unit is information indicating that the Ce-Zr-M0-based composite oxide will not undergo phase separation by heat treatment; and a second prediction unit that, when the information acquired by the first prediction unit is information indicating that the Ce-Zr-M0-based composite oxide does not undergo phase separation due to heat treatment, acquires information regarding the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after heat treatment, using the second trained model and information regarding a second characteristic of the Ce-Zr-M0-based composite oxide. An apparatus comprising: [2] The apparatus of [1], wherein the first characteristic includes at least one selected from the group consisting of atomic radius, covalent bond radius, van der Waals radius, metallic bond radius, and molar volume. [3] The device according to [1] or [2], wherein the second property includes at least one selected from the group consisting of density, heat capacity, and electronegativity. [4] A method for predicting the oxygen storage capacity of a Ce-Zr-M0-based composite oxide (wherein M0 represents an additive element) after heat treatment, comprising: Follow these steps: (a) preparing a first trained model generated by machine learning using information regarding a first characteristic of each composite oxide in a first composite oxide group as an explanatory variable and information regarding whether each composite oxide in the first composite oxide group undergoes phase separation upon heat treatment as a target variable, wherein the first composite oxide group includes a plurality of Ce-Zr-M1-based composite oxides that do not undergo phase separation upon heat treatment and a plurality of Ce-Zr-M1-based composite oxides that undergo phase separation upon heat treatment, each M1 independently representing an additive element; (b) preparing a second trained model generated by machine learning using information about a second characteristic of each complex oxide in a second complex oxide group as an explanatory variable and information about the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment as a target variable, wherein the second complex oxide group includes a plurality of Ce-Zr-M2-based complex oxides that do not undergo phase separation upon heat treatment, but do not include material properties of Ce-Zr-M2-based complex oxides that undergo phase separation upon heat treatment, and each M2 independently represents an additive element; (c) acquiring information on whether or not the Ce-Zr-M0-based composite oxide undergoes phase separation by heat treatment, using the first trained model and information on a first characteristic of the Ce-Zr-M0-based composite oxide; (d) determining whether the information obtained in step (c) is information indicating that the Ce-Zr-M0-based composite oxide does not undergo phase separation by heat treatment; and (e) when the information acquired in step (c) is information indicating that the Ce-Zr-M0-based composite oxide does not undergo phase separation by heat treatment, acquiring information regarding the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after heat treatment using the second trained model and information regarding a second characteristic of the Ce-Zr-M0-based composite oxide. The method includes: [5] The method according to [4], wherein the first characteristic includes at least one selected from the group consisting of atomic radius, covalent bond radius, van der Waals radius, metallic bond radius, and molar volume. [6] The method according to [4] or [5], wherein the second property includes at least one selected from the group consisting of density, heat capacity, and electronegativity. [7] A program for predicting the oxygen storage capacity of a Ce-Zr-M0-based composite oxide (wherein M0 represents an additive element) after heat treatment, comprising: Follow these steps: (a) preparing a first trained model generated by machine learning using information regarding a first characteristic of each composite oxide in a first composite oxide group as an explanatory variable and information regarding whether each composite oxide in the first composite oxide group undergoes phase separation upon heat treatment as a target variable, wherein the first composite oxide group includes a plurality of Ce-Zr-M1-based composite oxides that do not undergo phase separation upon heat treatment and a plurality of Ce-Zr-M1-based composite oxides that undergo phase separation upon heat treatment, each M1 independently representing an additive element; (b) preparing a second trained model generated by machine learning using information about a second characteristic of each complex oxide in a second complex oxide group as an explanatory variable and information about the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment as a target variable, wherein the second complex oxide group includes a plurality of Ce-Zr-M2-based complex oxides that do not undergo phase separation upon heat treatment, but do not include material properties of Ce-Zr-M2-based complex oxides that undergo phase separation upon heat treatment, and each M2 independently represents an additive element; (c) acquiring information on whether or not the Ce-Zr-M0-based composite oxide undergoes phase separation by heat treatment, using the first trained model and information on a first characteristic of the Ce-Zr-M0-based composite oxide; (d) determining whether the information obtained in step (c) is information indicating that the Ce-Zr-M0-based composite oxide does not undergo phase separation by heat treatment; and (e) when the information acquired in step (c) is information indicating that the Ce-Zr-M0-based composite oxide does not undergo phase separation by heat treatment, acquiring information regarding the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after heat treatment using the second trained model and information regarding a second characteristic of the Ce-Zr-M0-based composite oxide. A program that causes a computer to execute the following. [8] The program according to [7], wherein the first characteristic includes at least one selected from the group consisting of atomic radius, covalent bond radius, van der Waals radius, metallic bond radius, and molar volume. [9] The program according to [7] or [8], wherein the second property includes at least one selected from the group consisting of density, heat capacity, and electronegativity. Effect of the Invention

[0014] According to the present invention, there are provided an apparatus, a method and a program capable of accurately predicting the oxygen storage capacity of a Ce-Zr based composite oxide containing an additive element after heat treatment. [Brief description of the drawings]

[0015] [Figure 1] FIG. 1 is a diagram showing an example of a hardware configuration of a computer that constitutes a prediction device according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram illustrating an example of a functional configuration of a prediction device according to an embodiment of the present invention. [Diagram 3] FIG. 3 is a flowchart of a processing procedure of the prediction device according to one embodiment of the present invention. [Figure 4] FIG. 4 is a flowchart of a processing procedure of the first model generating unit according to one embodiment of the present invention. [Diagram 5] FIG. 5 is a flowchart of a processing procedure of the second model generation unit according to one embodiment of the present invention. [Figure 6]FIG. 6 is a diagram showing second characteristic values ​​(108 types of characteristic values ​​selected from the characteristic values ​​recorded in XenonPy) of each complex oxide in the second complex oxide group used in Example 1 and Comparative Example 1. [Figure 7] FIG. 7 is a diagram showing an example of the correspondence between the weighted average value of density ("ave:density"), the weighted average value of electronegativity ("ave:en_ghosh") and the weighted average value of heat capacity ("ave:heat_capacity_molar"), among the second characteristic values ​​of each complex oxide in the second complex oxide group, and the OSC value (actual measured value) after heat treatment of each complex oxide in the second complex oxide group. [Figure 8] FIG. 8 is a diagram showing an example of the correspondence between the first characteristic values ​​(atomic radius of M element, covalent bond radius of M element, van der Waals radius of M element, and molar volume of M element) of each complex oxide in the first complex oxide group and information on whether or not each complex oxide in the first complex oxide group has undergone phase separation by heat treatment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] Hereinafter, embodiments of the present invention will be described. Two or more embodiments can be combined, and combinations of two or more embodiments are also included in the present invention.

[0017] <Terminology explanation> The following describes terms used in this specification. The following explanations of terms apply throughout this specification unless otherwise specified.

[0018] <Complex oxide> Below, we will explain Ce-Zr-M based composite oxides and Ce-Zr based composite oxides that do not contain additive elements. The following explanation of Ce-Zr-M based composite oxides applies to Ce-Zr-M0 based composite oxides, Ce-Zr-M1 based composite oxides, and Ce-Zr-M2 based composite oxides, unless otherwise specified. When applied to Ce-Zr-M0 based composite oxides, "M" is replaced with "M0", when applied to Ce-Zr-M1 based composite oxides, "M" is replaced with "M1", and when applied to Ce-Zr-M2 based composite oxides, "M" is replaced with "M2".

[0019] The Ce-Zr-M based composite oxide is an oxide containing the elements Ce, Zr and M.

[0020] In the Ce-Zr-M based composite oxide, M represents an additive element. The "additive element" refers to an element other than Ce, Zr, and O.

[0021] The Ce-Zr based composite oxide containing no additional element is an oxide containing the Ce element and the Zr element but not the M element.

[0022] The type of M element can be appropriately selected in consideration of the oxygen storage capacity, heat resistance, etc. of the Ce-Zr-M based composite oxide. The M element may be composed of one kind of element, or may be composed of two or more and n or less kinds of elements. That is, the Ce-Zr-M based composite oxide may contain one kind of additive element, or may contain two or more and n or less kinds of additive elements. n is not particularly limited as long as it is an integer of 2 or more. In one embodiment, n is an integer of 2 or more and 5 or less. In another embodiment, n is an integer of 2 or more and 4 or less.

[0023] The M element can be selected from, for example, rare earth elements other than Ce, alkaline earth metal elements, transition elements of Groups 4 to 11, and other metal elements (metal elements other than rare earth elements, alkaline earth metal elements, and transition elements of Groups 4 to 11).

[0024] Examples of rare earth elements other than Ce include Sc, Y, La, Pr, Nd, Pm, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, and Lu.

[0025] Examples of the alkaline earth metal elements include Mg, Ca, Sr, and Ba.

[0026] Examples of the transition elements of Groups 4 to 11 include Mn, Fe, Co, Ni, Cu, Nb, and Pd.

[0027] Examples of other metal elements include B, Al, Si, P, and S.

[0028] The content R11 of Ce element in the elements other than O element in the Ce-Zr-M based composite oxide is, for example, 5 mol% to 50 mol%, preferably 10 mol% to 45 mol%, and more preferably 15 mol% to 40 mol%. R11 is calculated based on the following formula. The unit of the amount of substance is mole. In the following formula, "amount of substance of M element" means the amount of substance of one element when M element is composed of one element, and means the total amount of substance of two or more elements when M element is composed of two or more elements. R11 = amount of Ce element / (amount of Ce element + amount of Zr element + amount of M element) x 100

[0029] The content R12 of Ce element in the elements other than O element in the Ce-Zr based composite oxide not including the additive element is, for example, 5 mol% to 50 mol%, preferably 10 mol% to 45 mol%, more preferably 15 mol% to 40 mol%. R12 is calculated based on the following formula. The unit of the amount of substance is mole. R12 = amount of Ce element substance / (amount of Ce element substance + amount of Zr element substance) x 100

[0030] The content R21 of Zr element in the elements other than O element in the Ce-Zr-M based composite oxide is, for example, 20 mol% or more and 95 mol% or less, preferably 35 mol% or more and 90 mol% or less, and more preferably 50 mol% or more and 84 mol% or less. R21 is calculated based on the following formula. The unit of the amount of substance is mole. In the following formula, "amount of substance of M element" has the same meaning as above. R21 = amount of Zr element / (amount of Ce element + amount of Zr + amount of M element) x 100

[0031] The content R22 of Zr element in the elements other than O element in the Ce-Zr based composite oxide not including the additive element is, for example, 50 mol % or more and 95 mol % or less, preferably 55 mol % or more and 90 mol % or less, and more preferably 60 mol % or more and 85 mol % or less. R22 is calculated based on the following formula. The unit of the amount of substance is mole. R22 = amount of Zr element / (amount of Ce element + amount of Zr element) x 100

[0032] The content R3 of the M element in the elements other than the O element in the Ce-Zr-M based composite oxide is, for example, 0.1 mol% to 30 mol%, preferably 0.5 mol% to 20 mol%, more preferably 1 mol% to 10 mol%. When the M element is composed of one element, R3 means the content of the one element, and when the M element is composed of two or more elements, R3 means the total content of the two or more elements. R3 is calculated based on the following formula. The unit of the amount of substance is mole. In the following formula, "amount of substance of the M element" has the same meaning as above. R3 = amount of substance of M element / (amount of substance of Ce element + amount of substance of Zr + amount of substance of M element) x 100

[0033] When the M element is composed of two or more elements (first element, second element, ..., and n element (n has the same meaning as above)), the contents of the first element, the second element, ..., and n element in the elements other than the O element in the Ce-Zr-M composite oxide are hereinafter referred to as "R31", "R32", ..., and "R3 n " R31, R32, ... and R3 nR31, R32, ... and R3 are each, for example, 0.1 mol % or more, preferably 0.5 mol % or more, and more preferably 1 mol % or more. n The upper limits of R31, R32, ... and R3 n The total of the k-th element (k=1, 2, . . . or n) is adjusted to be, for example, 30 mol % or less, preferably 20 mol % or less, and more preferably 10 mol % or less. k is calculated based on the following formula. The unit of substance amount is mole. In the following formula, "amount of substance of M element" has the same meaning as above. R3 k = amount of substance of kth element / (amount of substance of Ce element + amount of substance of Zr element + amount of substance of M element) x 100

[0034] The Ce element may form a solid solution phase (e.g., a solid solution phase of CeO2 and ZrO2, a solid solution phase of CeO2 and an oxide of the M element, a solid solution phase of CeO2, ZrO2 and an oxide of the M element, etc.), or may form a single phase which is a crystalline phase or an amorphous phase (e.g., a CeO2 single phase), or may form both a solid solution phase and a single phase, but it is preferable that at least a portion of the Ce element forms a solid solution phase.

[0035] The Zr element may form a solid solution phase (e.g., a solid solution phase of CeO2 and ZrO2, a solid solution phase of ZrO2 and an oxide of the M element, a solid solution phase of CeO2, ZrO2 and an oxide of the M element, etc.), or may form a single phase which is a crystalline phase or an amorphous phase (e.g., a ZrO2 single phase), or may form both a solid solution phase and a single phase, but it is preferable that at least a portion of Zr forms a solid solution phase.

[0036] The M element may form a solid solution phase (e.g., a solid solution phase of CeO2 and an oxide of the M element, a solid solution phase of ZrO2 and an oxide of the M element, a solid solution phase of CeO2, ZrO2 and an oxide of the M element, etc.), or may form a single phase which is a crystalline phase or an amorphous phase (e.g., a single phase of an oxide of the M element), or may form both a solid solution phase and a single phase, but it is preferable that at least a part of the M element forms a solid solution phase.

[0037] When the composition of the complex oxide is known, the content of each element in the complex oxide in terms of oxide can be determined from the composition of the complex oxide. When the composition of the complex oxide is not known, the content of each element in the complex oxide can be measured by a conventional method such as inductively coupled plasma optical emission spectroscopy (ICP), X-ray fluorescence spectroscopy (XRF), or scanning electron microscope-energy dispersive X-ray spectroscopy (SEM-EDX).

[0038] The Ce-Zr-based composite oxide and the Ce-Zr-M-based composite oxide containing no additive elements are preferably prepared by dissolving a water-soluble salt of Ce, a water-soluble salt of Zr, and optionally a water-soluble salt of M in water to prepare an aqueous solution, adding a base to the prepared aqueous solution, separating the resulting precipitate from the aqueous solution, and then drying and calcining the resulting solution.

[0039] <Heat treatment> The treatment temperature in the heat treatment is, for example, 800° C. or more and 1200° C. or less. In one embodiment, the treatment temperature in the heat treatment is, for example, 900° C. or more and 1100° C. or less.

[0040] The treatment time in the heat treatment is, for example, 3 hours or more and 50 hours or less. In one embodiment, the treatment time in the heat treatment is 10 hours or more and 30 hours or less.

[0041] The heat treatment can be carried out, for example, under hydrothermal conditions. Examples of heat treatments carried out under hydrothermal conditions include heat treatments carried out under an atmosphere containing 0.50 vol% O2 gas, 10 vol% water vapor, and N2 as a balance gas. The heat treatment can also be carried out under conditions in which the atmosphere is circulated.

[0042] <Phase separation> Whether or not a certain composite oxide undergoes phase separation by heat treatment can be determined by the X-ray diffraction pattern of the composite oxide after heat treatment. Specifically, it is as follows. When one peak is observed at 2θ=27 to 32° in the X-ray diffraction pattern of the composite oxide after heat treatment, it can be determined that no phase separation has occurred in the composite oxide after heat treatment (i.e., the composite oxide does not undergo phase separation by heat treatment). On the other hand, when two or more peaks are observed at 2θ=27 to 32° in the X-ray diffraction pattern of the composite oxide after heat treatment, it can be determined that phase separation has occurred in the composite oxide after heat treatment (i.e., the composite oxide undergoes phase separation by heat treatment).

[0043] The X-ray diffraction pattern can be obtained by X-ray diffraction (XRD) using CuKα radiation. For example, CuKα1 radiation can be used as the CuKα radiation. XRD can be performed, for example, using a commercially available X-ray diffraction device (for example, MiniFlex600 manufactured by Rigaku Corporation). XRD can be performed, for example, according to the conditions described in the examples.

[0044] <Prediction device> A prediction device 1 according to one embodiment of the present invention is a device for predicting the oxygen storage capacity of a Ce-Zr-M0 based composite oxide after heat treatment.

[0045] The Ce-Zr-M0 composite oxide is a composite oxide whose oxygen storage capacity after heat treatment should be predicted. The explanation of "heat treatment" is as above.

[0046] The oxygen storage capacity of Ce-Zr-M0 based composite oxides after heat treatment is generally unknown.

[0047] In the Ce-Zr-M0-based composite oxide, M0 represents an additive element. The explanation of the "additive element" is as described above.

[0048] The prediction device 1 is composed of one or more computers. When the prediction device 1 is composed of two or more computers, these computers are connected via a communication network such as the Internet or an intranet to logically construct a single prediction device 1.

[0049] The prediction device 1 will be described below with reference to FIGS.

[0050] FIG. 1 is a diagram illustrating an example of a hardware configuration of a computer 100 that constitutes the prediction device 1. As shown in FIG.

[0051] 1, the computer 100 includes a processor 101, a main memory unit 102, an auxiliary memory unit 103, an input device 104, an output device 105, a communication interface (communication I / F) 106, and an external interface (external I / F) 107. Each piece of hardware of the computer 100 is connected to each other via a bus line.

[0052] The processor 101 is an arithmetic device such as a central processing unit (CPU) or a graphics processing unit (GPU). The processor 101 is an arithmetic device that realizes control, functions, and the like of the computer 100 by reading programs, data, and the like from storage devices such as the main storage unit 102, the auxiliary storage unit 103, and the recording medium 108, and executing processing. The main storage unit 102 is composed of, for example, a random access memory (RAM) and a read only memory (ROM). The RAM is a volatile semiconductor memory (storage device) that temporarily holds programs, data, and the like. The ROM is a non-volatile semiconductor memory (storage device) that can hold programs, data, and the like even when the power is turned off. The auxiliary storage unit 103 is, for example, a storage device (storage device) such as a hard disk drive (HDD), a solid state drive (SSD), a flash memory, and the like. The input device 104 is, for example, a keyboard, a mouse, a touch panel, and the like. The output device 105 is a monitor, a touch panel display, and the like. The communication I / F 106 is an interface for connecting the computer 100 to a communication network such as the Internet or an intranet. The computer 100 is connected to a communication network such as the Internet or an intranet via the communication I / F 106 so as to be able to perform data communication. The external I / F 107 is an interface with an external device such as a recording medium 108. The computer 100 can read from and write to the recording medium 108 via the external I / F 107. Examples of the recording medium 108 include a flexible disk, a CD (Compact Disc), a DVD (Digital Versatile Disk), a Secure Digital memory card (SD memory card), and a Universal Serial Bus (USB) memory card.

[0053] Each functional element of the prediction device 1 is realized by causing the processor 101 to execute a program. The processor 101 operates the input device 104, the output device 105, etc. in accordance with the program, and reads and writes data in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, etc. Programs, data, databases, etc. required for processing are stored in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, etc. The programs will be described later.

[0054] FIG. 2 is a diagram illustrating an example of a functional configuration of the prediction device 1. As shown in FIG.

[0055] As shown in FIG. 2, the prediction device 1 includes a first trained model 2, a second trained model 3, a first prediction unit 4, a determination unit 5, and a second prediction unit 6.

[0056] As shown in FIG. 2, the prediction device 1 may include a first model generation unit 7 and a second model generation unit 8.

[0057] The first trained model 2, the second trained model 3, the first prediction unit 4, the second prediction unit 6, the first model generation unit 7 and the second model generation unit 8 may be constructed on the same computer 100, or some of them may be constructed on different computers 100.

[0058] <First trained model> The first trained model 2 is stored, for example, in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, etc.

[0059] The first trained model 2 is a prediction model generated by machine learning with information about the first characteristic of each complex oxide in the first complex oxide group as an explanatory variable and information about whether each complex oxide in the first complex oxide group undergoes phase separation by heat treatment as an objective variable. That is, the first trained model 2 is a prediction model that acquires, by machine learning, the correspondence between information about the first characteristic of each complex oxide in the first complex oxide group and information about whether each complex oxide in the first complex oxide group undergoes phase separation by heat treatment.

[0060] The first composite oxide group is composed of known Ce-Zr-based composite oxides for which information on the first characteristic is known and information on whether or not phase separation occurs by heat treatment is known.

[0061] The first composite oxide group includes a plurality of Ce-Zr-M1 composite oxides that do not undergo phase separation by heat treatment, and a plurality of Ce-Zr-M1 composite oxides that undergo phase separation by heat treatment. The explanations regarding "heat treatment" and "phase separation" are as described above.

[0062] The first composite oxide group may include a Ce-Zr composite oxide that does not include an additive element. The explanation regarding the "Ce-Zr composite oxide that does not include an additive element" is as above. The Ce-Zr composite oxide that does not include an additive element may be one that does not cause phase separation by heat treatment, or one that causes phase separation by heat treatment.

[0063] From the viewpoint of improving the prediction accuracy of the first trained model 2, it is preferable that the first composite oxide group is composed of a plurality of Ce-Zr-M1-based composite oxides that do not undergo phase separation upon heat treatment and a plurality of Ce-Zr-M1-based composite oxides that undergo phase separation upon heat treatment.

[0064] In the Ce-Zr-M1-based composite oxide, each M1 independently represents an additive element. The explanation of the "additive element" is as described above.

[0065] From the viewpoint of improving the prediction accuracy of the first trained model 2, the number of the multiple Ce-Zr-M1 based composite oxides that do not undergo phase separation by heat treatment and are included in the first composite oxide group is preferably 5 or more, more preferably 10 or more, and even more preferably 50 or more. The upper limit is not particularly limited, and can be appropriately adjusted in consideration of the time required to generate the first trained model 2, etc. The upper limit is, for example, 100,000 or less.

[0066] From the viewpoint of improving the prediction accuracy of the first trained model 2, the number of the multiple Ce-Zr-M1-based composite oxides that undergo phase separation by heat treatment and are included in the first composite oxide group is preferably 5 or more, more preferably 10 or more, and even more preferably 50 or more. The upper limit is not particularly limited, and can be appropriately adjusted in consideration of the time required to generate the first trained model 2, etc. The upper limit is, for example, 100,000 or less.

[0067] From the standpoint of improving the prediction accuracy by the first trained model 2, the ratio of the number of Ce-Zr-M1-based composite oxides that do not undergo phase separation by heat treatment to the total number of Ce-Zr-M1-based composite oxides that do not undergo phase separation by heat treatment in the first composite oxide group is preferably 30% or more and 70% or less, more preferably 40% or more and 60% or less.

[0068] Examples of the first characteristic include atomic radius, covalent bond radius, van der Waals radius, metallic bond radius, and molar volume. The atomic radius is the metallic bond radius for metal elements, the covalent bond radius for base metal elements other than rare gases, and the van der Waals radius for rare gases. The covalent bond radius may be the single bond distance, the double bond distance, or the triple bond distance. The molar volume of a certain element is, for example, the volume of the element when the element itself is dissolved under certain conditions (e.g., 22°C and 1.0×10 5 This is the bulk volume occupied by 1 mole of the monomer when it takes on its most stable phase (under conditions of 100 MPa and 1000 Pa).

[0069] From the viewpoint of improving the prediction accuracy by the first trained model 2, it is preferable that the first characteristic includes at least one selected from the group consisting of atomic radius, covalent bond radius, van der Waals radius, metallic bond radius, and molar volume.

[0070] As the atomic radius, the atomic radius by Rahm, etc. may be used. As the covalent bond radius, the covalent bond radius by Cerdero, the single bond distance of the covalent bond radius by Pyykko, the double bond distance of the covalent bond radius by Pyykko, the triple bond distance of the covalent bond radius by Pyykko, the covalent bond radius by Slater, etc. may be used. As the van der Waals radius, the van der Waals radius by Alvarez, the van der Waals radius of MM3, the van der Waals radius of UFF, etc. may be used.

[0071] The information about the first characteristic of each complex oxide in the first complex oxide group is, for example, a quantified value of the first characteristic of each complex oxide in the first complex oxide group. Hereinafter, the quantified value of the first characteristic is referred to as the "first characteristic value." The first characteristic value of each complex oxide may be an actually measured value or a calculated value. The atomic radius, covalent bond radius, van der Waals radius, and metallic bond radius are each in a unit of, for example, Å (angstrom), and the molar volume is each in a unit of, for example, m 3 / mol.

[0072] Hereinafter, a first characteristic value of the Ce-Zr-M1 based composite oxide in which the M1 element is composed of one kind of element will be described.

[0073] In one embodiment, the first characteristic value of the Ce-Zr-M1 based composite oxide is a weighted average of the first characteristic value of the Ce element, the first characteristic value of the Zr element, and the first characteristic value of the M1 element, and the weighted average is calculated based on the following formula: Weighted average value = ((Ce element content R11 × first characteristic value of Ce element) + (Zr element content R21 × first characteristic value of Zr element) + (M1 element content R3 × first characteristic value of M1 element)) / (Ce element content R11 + Zr element content R21 + M1 element content R3)

[0074] In another embodiment, the first characteristic value of the Ce-Zr-M1 based composite oxide is a variance of the first characteristic value of the Ce element, the first characteristic value of the Zr element, and the first characteristic value of the M1 element, and the variance is calculated based on the following formula: Variance value = (Ce element content R11 × (first characteristic value of Ce element - weighted average value) 2 + Zr element content R21 × (first characteristic value of Zr element - weighted average value) 2 + Content of M1 element R3 × (first characteristic value of M1 element - weighted average value) 2 ) / (Ce element content R11 + Zr element content R21 + M1 element content R3) In the above formula, the "weighted average value" represents a weighted average value of the first characteristic value of the Ce element, the first characteristic value of the Zr element, and the first characteristic value of the M1 element.

[0075] In still another embodiment, the first characteristic value of the Ce-Zr-M1 based composite oxide is a first characteristic value of the M1 element.

[0076] The first characteristic value of the Ce-Zr-M1 based composite oxide in which the M1 element is composed of one kind of element is preferably the first characteristic value of the M1 element.

[0077] Hereinafter, a first characteristic value of a Ce-Zr-M1 based composite oxide in which the M1 element is composed of two or more elements (a first element, a second element, ..., and an nth element (n has the same meaning as above)) will be described.

[0078] In one embodiment, the first characteristic value of the Ce-Zr-M1 based composite oxide is a weighted average of the first characteristic value of the Ce element, the first characteristic value of the Zr element, the first characteristic value of the first element, the first characteristic value of the second element, ..., and the first characteristic value of the n element. The weighted average is calculated based on the following formula. Weighted average value = ((Ce element content R11 × first characteristic value of Ce element) + (Zr element content R21 × first characteristic value of Zr element) + (1st element content R31 × first characteristic value of 1st element) + (2nd element content R32 × first characteristic value of 2nd element) + ··· + (nth element content R3 n × 1st characteristic value of nth element) / (Content of Ce element R11 + Content of Zr element R21 + Content of 1st element R31 + Content of 2nd element R32 + ··· + Content of nth element R3 n )

[0079] In another embodiment, the first characteristic value of the Ce-Zr-M1 based composite oxide is a variance of the first characteristic value of the Ce element, the first characteristic value of the Zr element, the first characteristic value of the first element, the first characteristic value of the second element, ..., and the first characteristic value of the n element. The variance is calculated based on the following formula. Variance value = (Ce element content R11 × (first characteristic value of Ce element - weighted average value) 2 + Zr element content R21 × (first characteristic value of Zr element - weighted average value) 2 + Content of the first element R31 × (first characteristic value of the first element - weighted average value) 2 + Content of second element R32 × (first characteristic value of second element - weighted average value) 2 +···+ Content of nth element R3 n × (first characteristic value of nth element - weighted average value) 2 ) / (Ce element content R11 + Zr element content R21 + 1st element content R31 + 2nd element content R32 + ··· + nth element content R3 n ) In the above formula, the “weighted average value” represents the weighted average value of the first characteristic value of the Ce element, the first characteristic value of the Zr element, the first characteristic value of the first element, the first characteristic value of the second element, ..., and the first characteristic value of the n element.

[0080] In yet another embodiment, the first characteristic value of the Ce-Zr-M1 based composite oxide is a weighted average of the first characteristic value of the first element, the first characteristic value of the second element, ..., and the first characteristic value of the n element, and the weighted average is calculated based on the following formula: Weighted average value = ((content of first element R31 × first characteristic value of first element) + (content of second element R32 × first characteristic value of second element) + + (content of nth element R3 n × first characteristic value of nth element) / (content of first element R31 + content of second element R32 + + content of nth element R3 n )

[0081] In still another embodiment, the first characteristic value of the Ce-Zr-M1 based composite oxide is a variance of the first characteristic value of the first element, the first characteristic value of the second element, ..., and the first characteristic value of the n element, and the variance is calculated based on the following formula: Variance value = (content of first element R31 × (first characteristic value of first element - weighted average value) 2 + Content of second element R32 × (first characteristic value of second element - weighted average value) 2 +···+ Content of nth element R3 n × (first characteristic value of nth element - weighted average value) 2 ) / (Content of 1st element R31 + Content of 2nd element R32 + ··· + Content of nth element R3 n ) In the above formula, the "weighted average value" represents a weighted average value of the first characteristic value of the first element, the first characteristic value of the second element, . . . and the first characteristic value of the nth element.

[0082] The first characteristic value of a Ce-Zr-M1-based composite oxide in which the M1 element is composed of two or more elements is preferably a weighted average of the first characteristic value of the first element, the first characteristic value of the second element, ..., and the first characteristic value of the nth element, or a variance of the first characteristic value of the first element, the first characteristic value of the second element, ..., and the first characteristic value of the nth element.

[0083] Hereinafter, a first characteristic value of the Ce—Zr based composite oxide containing no additive element will be described.

[0084] In one embodiment, the first characteristic value of the Ce-Zr based composite oxide not including the additive element is a weighted average of the first characteristic value of the Ce element and the first characteristic value of the Zr element, and the weighted average is calculated based on the following formula: Weighted average value = ((Ce element content R12 × first characteristic value of Ce element) + (Zr element content R22 × first characteristic value of Zr element)) / (Ce element content R12 + Zr element content R22)

[0085] In another embodiment, the first characteristic value of the Ce-Zr based composite oxide not including the additive element is a variance value of the first characteristic value of the Ce element and the first characteristic value of the Zr element, and the variance value is calculated based on the following formula: Variance value = (Ce element content R12 × (first characteristic value of Ce element - weighted average value) 2 + Zr element content R22 × (first characteristic value of Zr element - weighted average value)2 ) / (Ce element content R12 + Zr element content R22) In the above formula, the "weighted average value" represents the weighted average value of the first characteristic value of the Ce element and the first characteristic value of the Zr element.

[0086] The information on whether or not each composite oxide in the first composite oxide group undergoes phase separation upon heat treatment is, for example, a value corresponding to whether or not each composite oxide in the first composite oxide group undergoes phase separation upon heat treatment. In one embodiment, the value is 0 when phase separation occurs, and the value is 1 when phase separation does not occur. In another embodiment, the value is 1 when phase separation occurs, and the value is 0 when phase separation does not occur.

[0087] <Second trained model> The second trained model 3 is stored, for example, in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, etc.

[0088] The second trained model 3 is a prediction model generated by machine learning with information on the second characteristic of each complex oxide in the second complex oxide group as an explanatory variable and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment as a target variable. That is, the second trained model 3 is a prediction model that acquires, by machine learning, the correspondence between information on the second characteristic of each complex oxide in the second complex oxide group and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment.

[0089] The second composite oxide group is composed of known Ce-Zr-based composite oxides for which information on the second property is known and information on the oxygen storage capacity after heat treatment is known.

[0090] The second composite oxide group includes a plurality of Ce-Zr-M2 composite oxides that do not undergo phase separation upon heat treatment, but does not include Ce-Zr-M2 composite oxides that undergo phase separation upon heat treatment. The explanations regarding "heat treatment" and "phase separation" are as described above.

[0091] The second composite oxide group may contain a Ce-Zr composite oxide that does not contain an additive element as long as it does not cause phase separation by heat treatment. The explanation regarding the "Ce-Zr composite oxide that does not contain an additive element" is as described above.

[0092] Since the second complex oxide group includes a plurality of Ce-Zr-M2-based complex oxides that do not undergo phase separation upon heat treatment, but does not include any Ce-Zr-M2-based complex oxides that undergo phase separation upon heat treatment, the prediction accuracy by the second trained model 3 can be improved.

[0093] From the viewpoint of improving the prediction accuracy of the second trained model 3, it is preferable that the second composite oxide group is composed of a plurality of Ce-Zr-M2-based composite oxides that do not undergo phase separation upon heat treatment.

[0094] In the Ce-Zr-M2-based composite oxide, each M2 independently represents an additive element. The explanation of the "additive element" is as described above.

[0095] From the viewpoint of improving the prediction accuracy of the second trained model 3, the number of the multiple Ce-Zr-M2 based composite oxides that do not undergo phase separation by heat treatment and are included in the second composite oxide group is preferably 5 or more, more preferably 10 or more, and even more preferably 50 or more. The upper limit is not particularly limited, and can be appropriately adjusted taking into consideration the time required to generate the second trained model 3, etc. The upper limit is, for example, 100,000 or less.

[0096] The multiple Ce-Zr-M2 composite oxides contained in the second composite oxide group, which do not undergo phase separation by heat treatment, may be completely identical to the multiple Ce-Zr-M1 composite oxides contained in the first composite oxide group, which do not undergo phase separation by heat treatment, or may be partially identical (i.e., partially different), or may be completely different, but it is preferable that they are completely identical or partially identical, and it is more preferable that they are completely identical.

[0097] Examples of the second characteristic include density, heat capacity, and electronegativity. The heat capacity may be mass heat capacity or molar heat capacity. The density and heat capacity of a certain element are, for example, the density and heat capacity of the element under a certain condition (e.g., 22°C and 1.0×10 5 These are the density and heat capacity of the substance when it is in its most stable phase (under conditions of 100 MPa and 100 MPa Pa).

[0098] From the viewpoint of improving the prediction accuracy by the second trained model 3, it is preferable that the second property includes at least one selected from the group consisting of density, heat capacity, and electronegativity.

[0099] As the electronegativity, the Allen electronegativity, the Ghosh electronegativity, the Pauling electronegativity, or the like may be used.

[0100] The information on the second property of each complex oxide in the second complex oxide group is, for example, a quantified value of the second property of each complex oxide in the second complex oxide group. Hereinafter, the quantified value of the second property is referred to as the "second property value." The second property value of each complex oxide may be an actually measured value or a calculated value. The unit of density is, for example, kg / m 3 and the units of heat capacity are, for example, J / K / kg or J / K / mol. Electronegativity is usually a dimensionless quantity.

[0101] Hereinafter, a second characteristic value of the Ce-Zr-M2 based composite oxide in which the M2 element is composed of one kind of element will be described.

[0102] In one embodiment, the second characteristic value of the Ce-Zr-M2 based composite oxide is a weighted average of the second characteristic value of the Ce element, the second characteristic value of the Zr element, and the second characteristic value of the M2 element, and the weighted average is calculated based on the following formula: Weighted average value = ((Ce element content R11 × second characteristic value of Ce element) + (Zr element content R21 × second characteristic value of Zr element) + (M2 element content R3 × second characteristic value of M2 element)) / (Ce element content R11 + Zr element content R21 + M2 element content R3)

[0103] In another embodiment, the second characteristic value of the Ce-Zr-M2 based composite oxide is a variance of the second characteristic value of the Ce element, the second characteristic value of the Zr element, and the second characteristic value of the M2 element, and the variance is calculated based on the following formula: Variance value = (Ce element content R11 × (Ce element second characteristic value - weighted average value) 2 + Zr element content R21 × (Zr element second characteristic value - weighted average value) 2 + Content of M2 element R3 × (second characteristic value of M2 element - weighted average value) 2 ) / (Ce element content R11 + Zr element content R21 + M2 element content R3) In the above formula, the "weighted average value" represents a weighted average value of the second characteristic value of the Ce element, the second characteristic value of the Zr element, and the second characteristic value of the M2 element.

[0104] In yet another embodiment, the second characteristic value of the Ce-Zr-M2 based composite oxide is a second characteristic value of the M2 element.

[0105] When the M2 element is composed of one type of element, the second characteristic value of the Ce-Zr-M2 composite oxide is preferably a weighted average value of the second characteristic value of the Ce element, the second characteristic value of the Zr element, and the second characteristic value of the M2 element.

[0106] Hereinafter, a second characteristic value of the Ce-Zr-M2 based composite oxide in which the M2 element is composed of two or more elements (a first element, a second element, ..., and an nth element (n has the same meaning as above)) will be described.

[0107] In one embodiment, the second characteristic value of the Ce-Zr-M2 based composite oxide is a weighted average of the second characteristic value of the Ce element, the second characteristic value of the Zr element, the second characteristic value of the first element, the second characteristic value of the second element, ..., and the second characteristic value of the n element. The weighted average is calculated based on the following formula. Weighted average value = ((Ce element content R11 × second characteristic value of Ce element) + (Zr element content R21 × second characteristic value of Zr element) + (1st element content R31 × second characteristic value of 1st element) + (2nd element content R32 × second characteristic value of 2nd element) + ··· + (nth element content R3 n × second characteristic value of nth element) / (Content of Ce element R11 + Content of Zr element R21 + Content of first element R31 + Content of second element R32 + ··· + Content of nth element R3 n )

[0108] In another embodiment, the second characteristic value of the Ce-Zr-M2 based composite oxide is a variance of the second characteristic value of the Ce element, the second characteristic value of the Zr element, the second characteristic value of the first element, the second characteristic value of the second element, ..., and the second characteristic value of the n element. The variance is calculated based on the following formula. Variance value = (Ce element content R11 × (Ce element second characteristic value - weighted average value) 2 + Zr element content R21 × (Zr element second characteristic value - weighted average value) 2 + Content of the first element R31 × (second characteristic value of the first element - weighted average value) 2 + Content of second element R32 × (second characteristic value of second element - weighted average value) 2 +···+ Content of nth element R3 n × (2nd characteristic value of nth element - weighted average value) 2 ) / (Ce element content R11 + Zr element content R21 + 1st element content R31 + 2nd element content R32 + ··· + nth element content R3 n ) In the above formula, the “weighted average value” represents the weighted average value of the second characteristic value of the Ce element, the second characteristic value of the Zr element, the second characteristic value of the first element, the second characteristic value of the second element, ..., and the second characteristic value of the n element.

[0109] In yet another embodiment, the second characteristic value of the Ce-Zr-M2 based composite oxide is a weighted average of the second characteristic value of the first element, the second characteristic value of the second element, ..., and the second characteristic value of the n element, and the weighted average is calculated based on the following formula: Weighted average value = ((content of first element R31 × second characteristic value of first element) + (content of second element R32 × second characteristic value of second element) + ··· + (content of nth element R3 n × second characteristic value of nth element) / (content of first element R31 + content of second element R32 + + content of nth element R3 n )

[0110] In still another embodiment, the second characteristic value of the Ce-Zr-M2 based composite oxide is a variance of the second characteristic value of the first element, the second characteristic value of the second element, ..., and the second characteristic value of the n element, and the variance is calculated based on the following formula: Variance value = (content of first element R31 × (second characteristic value of first element - weighted average value) 2 + Content of second element R32 × (second characteristic value of second element - weighted average value) 2 +···+ Content of nth element R3 n × (2nd characteristic value of nth element - weighted average value) 2 ) / (Content of 1st element R31 + Content of 2nd element R32 + ··· + Content of nth element R3 n ) In the above formula, the "weighted average value" represents a weighted average value of the second characteristic value of the first element, the second characteristic value of the second element, . . . and the second characteristic value of the nth element.

[0111] Hereinafter, the second characteristic value of the Ce—Zr based composite oxide containing no additive element will be described.

[0112] In one embodiment, the second characteristic value of the Ce-Zr based composite oxide not including the additive element is a weighted average of the second characteristic value of the Ce element and the second characteristic value of the Zr element, and the weighted average is calculated based on the following formula: Weighted average value = ((Ce element content R12 × second characteristic value of Ce element) + (Zr element content R22 × second characteristic value of Zr element)) / (Ce element content R12 + Zr element content R22)

[0113] In another embodiment, the second characteristic value of the Ce-Zr based composite oxide not including the additive element is a variance value of the second characteristic value of the Ce element and the second characteristic value of the Zr element, and the variance value is calculated based on the following formula. Dispersion value = (Ce element content R12 × (Ce element second characteristic value - arithmetic mean value) 2 + Zr element content R22 × (second characteristic value of Zr element - arithmetic mean value) 2 ) / (Ce element content R12 + Zr element content R22) In the above formula, the "weighted average value" represents the weighted average value of the second characteristic value of the Ce element and the second characteristic value of the Zr element.

[0114] The information on the oxygen storage capacity of each complex oxide in the second complex oxide group after the heat treatment is, for example, a quantified value of the oxygen storage capacity of each complex oxide in the second complex oxide group after the heat treatment. Hereinafter, the quantified value of the oxygen storage capacity is referred to as the "OSC value". The OSC value may be an actual measured value or a calculated value. The OSC value is preferably an actual measured value. The unit of the OSC value is, for example, μmol / g. "μmol / g" means the OSC value per unit mass of the complex oxide. The OSC value can be measured, for example, according to the method described in the Examples.

[0115] <First Prediction Section> The first prediction unit 4 uses the first trained model 2 and information on the first characteristic of the Ce-Zr-M0-based composite oxide to obtain information on whether or not the Ce-Zr-M0-based composite oxide will undergo phase separation due to heat treatment. Specifically, the first prediction unit 4 obtains information on the first characteristic of the Ce-Zr-M0-based composite oxide and inputs the obtained information to the first trained model 2. Based on the input information, the first trained model 2 predicts whether or not the Ce-Zr-M0-based composite oxide will undergo phase separation due to heat treatment, and outputs information on whether or not the Ce-Zr-M0-based composite oxide will undergo phase separation due to heat treatment. The first prediction unit 4 obtains information output from the first trained model 2.

[0116] The first prediction unit 4 obtains information on whether or not phase separation occurs in the Ce—Zr—M0 based composite oxide by heat treatment, and then provides the obtained information to the determination unit 5 .

[0117] The information on the first characteristic of the Ce-Zr-M0 based composite oxide corresponds to the information on the first characteristic of each composite oxide in the first composite oxide group. The first characteristic has been described above. When the information on the first characteristic of each composite oxide in the first composite oxide group includes information on a certain characteristic (e.g., at least one selected from the group consisting of atomic radius, covalent bond radius, van der Waals radius, metallic bond radius, and molar volume), the information on the first characteristic of the Ce-Zr-M0 based composite oxide includes information on that characteristic.

[0118] In one embodiment, the information on the first characteristic of the Ce-Zr-M0 based composite oxide is input by a user via the input device 104, and the first prediction unit 4 receives the information on the first characteristic of the Ce-Zr-M0 based composite oxide input by the user via the input device 104, thereby acquiring the information on the first characteristic of the Ce-Zr-M0 based composite oxide.

[0119] In another embodiment, information on the first characteristic of the Ce-Zr-M0 based composite oxide is stored in the main memory unit 102 or the auxiliary memory unit 103, and the first prediction unit 4 acquires the information on the first characteristic of the Ce-Zr-M0 based composite oxide from the main memory unit 102 or the auxiliary memory unit 103 in accordance with an instruction or condition input by a user via the input device 104.

[0120] In still another embodiment, information about the first characteristic of the Ce-Zr-M0-based composite oxide is stored in an external device connected to the prediction device 1 via a communication network, and the first prediction unit 4 acquires the information about the first characteristic of the Ce-Zr-M0-based composite oxide from the external device via the communication network in accordance with an instruction or condition input by a user via the input device 104.

[0121] In still another embodiment, the information on the first characteristic of the Ce-Zr-M0 based composite oxide is stored in a recording medium 108, and the first prediction unit 4 acquires the information on the first characteristic of the Ce-Zr-M0 based composite oxide from the recording medium 108 in accordance with an instruction or condition input by a user via the input device 104.

[0122] <Judgment part> The judgment unit 5 judges whether or not the information acquired by the first prediction unit 4 (information provided by the first prediction unit 4) is information indicating that the Ce-Zr-M0-based composite oxide will not undergo phase separation by heat treatment.

[0123] The determination unit 5 obtains, as a determination result, information that the Ce-Zr-M0 based composite oxide does not undergo phase separation by heat treatment or information that the Ce-Zr-M0 based composite oxide undergoes phase separation by heat treatment.

[0124] A case in which the information acquired by the first prediction unit 4 (information provided by the first prediction unit 4) indicates that the Ce-Zr-M0 based composite oxide will not undergo phase separation due to heat treatment will be referred to as a "case in which the judgment result is YES", and a case in which the information acquired by the first prediction unit 4 (information provided by the first prediction unit 4) indicates that the Ce-Zr-M0 based composite oxide will undergo phase separation due to heat treatment will be referred to as a "case in which the judgment result is NO".

[0125] When the result of the determination is YES, the determining unit 5 provides the second predicting unit 6 with information to the effect that the Ce-Zr-M0 based composite oxide will not undergo phase separation due to the heat treatment.

[0126] If the determination result is NO, the determination unit 5 may provide the second prediction unit 6 with information to the effect that phase separation will occur in the Ce-Zr-M0 based composite oxide by heat treatment, but it is preferable not to provide such information.

[0127] When the determination result is NO, the determination unit 5 may provide information to the output device 105 that the Ce-Zr-M0 based composite oxide will undergo phase separation due to the heat treatment, and the output device 105 may output information that the Ce-Zr-M0 based composite oxide will undergo phase separation due to the heat treatment.

[0128] <Second Prediction Section> When the judgment result is YES (i.e., when the judgment unit 5 provides the second prediction unit 6 with information that the Ce-Zr-M0-based composite oxide does not undergo phase separation by heat treatment), the second prediction unit 6 uses the second trained model 3 and information on the second characteristic of the Ce-Zr-M0-based composite oxide to acquire information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment. Specifically, it is as follows. The second prediction unit 6 acquires information on the second characteristic of the Ce-Zr-M0-based composite oxide and inputs the acquired information to the second trained model 3. The second trained model 3 predicts the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment based on the input information, and outputs information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment. The second prediction unit 6 acquires the information output from the second trained model 3.

[0129] The information on the second characteristic of the Ce-Zr-M0 based composite oxide corresponds to the information on the second characteristic of each composite oxide in the second composite oxide group. The second characteristic is as described above. When the information on the second characteristic of each composite oxide in the second composite oxide group includes information on a certain characteristic (e.g., at least one selected from the group consisting of density, heat capacity, and electronegativity), the information on the second characteristic of the Ce-Zr-M0 based composite oxide includes information on that characteristic.

[0130] In one embodiment, the information on the second characteristic of the Ce-Zr-M0 based composite oxide is input by a user via the input device 104, and the second prediction unit 6 acquires the information on the second characteristic of the Ce-Zr-M0 based composite oxide by accepting the information on the second characteristic of the Ce-Zr-M0 based composite oxide input by the user via the input device 104.

[0131] In another embodiment, information on the second characteristic of the Ce-Zr-M0 based composite oxide is stored in the main memory unit 102 or the auxiliary memory unit 103, and the second prediction unit 6 acquires the information on the second characteristic of the Ce-Zr-M0 based composite oxide from the main memory unit 102 or the auxiliary memory unit 103 in accordance with an instruction or condition input by a user via the input device 104.

[0132] In still another embodiment, information regarding the second property of the Ce-Zr-M0-based composite oxide is stored in an external device connected to the prediction device 1 via a communication network, and the second prediction unit 6 acquires the information regarding the second property of the Ce-Zr-M0-based composite oxide from the external device via the communication network in accordance with an instruction or condition input by a user via the input device 104.

[0133] In still another embodiment, the information on the second property of the Ce-Zr-M0 based composite oxide is stored in a recording medium 108, and the second prediction unit 6 acquires the information on the second property of the Ce-Zr-M0 based composite oxide from the recording medium 108 in accordance with an instruction or condition input by a user via the input device 104.

[0134] After acquiring the information regarding the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment, the second prediction unit 6 may provide the information regarding the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment to the output device 105, together with information that the Ce-Zr-M0 based composite oxide does not undergo phase separation due to the heat treatment, as necessary, and the output device 105 may output the information regarding the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment, together with information that the Ce-Zr-M0 based composite oxide does not undergo phase separation due to the heat treatment, as necessary.

[0135] After acquiring the information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment, the second prediction unit 6 may acquire predetermined information based on the information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment, and may provide the predetermined information to the output device 105 together with information that the Ce-Zr-M0-based composite oxide does not undergo phase separation due to the heat treatment, as necessary, and the output device 105 may output the predetermined information together with information that the Ce-Zr-M0-based composite oxide does not undergo phase separation due to the heat treatment, as necessary. Examples of the predetermined information include information on whether the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment satisfies a predetermined standard (for example, whether the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment is equal to or greater than a predetermined standard value, or which of the preset categories the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment belongs to), etc.

[0136] <First model generation part> As shown in FIG. 2, the first model generation unit 7 includes a first data acquisition unit 71 and a first learning unit 72.

[0137] The first model generation unit 7 creates a first training dataset, and generates a first trained model 2 by machine learning using the first training dataset.

[0138] Information regarding the first characteristic of each complex oxide in the first complex oxide group and information regarding whether or not each complex oxide in the first complex oxide group undergoes phase separation by heat treatment are associated with each other and stored in, for example, the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, or another external device.

[0139] In one embodiment, the first data acquisition unit 71 acquires a first training data set including information on the first property of each complex oxide in the first complex oxide group and information on whether each complex oxide in the first complex oxide group undergoes phase separation by heat treatment from the main memory unit 102 or the auxiliary memory unit 103. The first data acquisition unit 71 may acquire the first training data set from the main memory unit 102 or the auxiliary memory unit 103 in accordance with an instruction or condition input by a user via the input device 104.

[0140] In another embodiment, the first data acquiring unit 71 acquires, via the communication network, from an external device connected to the prediction device 1 via the communication network, a first training data set including information on the first property of each complex oxide in the first complex oxide group and information on whether each complex oxide in the first complex oxide group undergoes phase separation by heat treatment. The first data acquiring unit 71 may acquire the first training data set via the communication network from the external device connected to the prediction device 1 via the communication network, in accordance with an instruction or condition input by a user via the input device 104.

[0141] In yet another embodiment, the first data acquiring unit 71 acquires a first training data set including information on a first property of each complex oxide in the first complex oxide group and information on whether each complex oxide in the first complex oxide group undergoes phase separation by heat treatment from the recording medium 108. The first data acquiring unit 71 may acquire the first training data set from the recording medium 108 in accordance with an instruction or condition input by a user via the input device 104.

[0142] The first learning dataset acquired by the first data acquisition unit 71 is used as a learning dataset (also called a teacher dataset) in which information regarding the first characteristic of each complex oxide in the first complex oxide group is used as an explanatory variable, and information regarding whether each complex oxide in the first complex oxide group undergoes phase separation upon heat treatment is used as a target variable.

[0143] From the viewpoint of improving the prediction accuracy of the first trained model 2, the number of pairs of information on the first characteristic of the Ce-Zr-M1-based composite oxide that does not undergo phase separation by heat treatment and information that the Ce-Zr-M1-based composite oxide does not undergo phase separation by heat treatment, which are included in the first training data set, is preferably 5 or more, more preferably 10 or more, and even more preferably 50 or more. The upper limit is not particularly limited, and can be appropriately adjusted in consideration of the time required to generate the first trained model 2, etc. The upper limit is, for example, 100,000 or less.

[0144] From the viewpoint of improving the prediction accuracy of the first trained model 2, the number of pairs of information on the first characteristic of the Ce-Zr-M1-based composite oxide that undergoes phase separation by heat treatment and information that the Ce-Zr-M1-based composite oxide undergoes phase separation by heat treatment, which are included in the first training data set, is preferably 5 or more, more preferably 10 or more, and even more preferably 50 or more. The upper limit is not particularly limited, and can be appropriately adjusted in consideration of the time required to generate the first trained model 2, etc. The upper limit is, for example, 100,000 or less.

[0145] From the standpoint of improving the prediction accuracy by the first trained model 2, the ratio of the number of Ce-Zr-M1-based composite oxides that do not undergo phase separation by heat treatment to the total number of Ce-Zr-M1-based composite oxides that do not undergo phase separation by heat treatment in the first composite oxide group is preferably 30% or more and 70% or less, more preferably 40% or more and 60% or less.

[0146] The first learning unit 72 performs machine learning using the first learning data set acquired by the first data acquisition unit 71, and generates a first trained model 2.

[0147] Examples of machine learning methods performed by the first learning unit 72 include logistic regression, random forest, simple linear regression, Ridge regression, Lasso regression, Elastic Net regression, general additive model, rule fit regression, gradient boosting tree, extra tree, support vector regression, Gaussian process regression, k-nearest neighbor regression, kernel ridge regression, and neural network.

[0148] The generated first trained model 2 is stored, for example, in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, etc.

[0149] <Second model generation part> As shown in FIG. 2, the second model generation unit 8 includes a second data acquisition unit 81 and a second learning unit .

[0150] The second model generation unit 8 creates a second training data set, and generates a second trained model 3 by machine learning using the second training data set.

[0151] Information regarding the second property of each complex oxide in the second complex oxide group and information regarding the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment are associated with each other and stored, for example, in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, or other external device.

[0152] In one embodiment, the second data acquisition unit 81 acquires a second training data set including information on the second property of each complex oxide in the second complex oxide group and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment from the main memory unit 102 or the auxiliary memory unit 103. The second data acquisition unit 81 may acquire the second training data set from the main memory unit 102 or the auxiliary memory unit 103 in accordance with an instruction or condition input by a user via the input device 104.

[0153] In another embodiment, the second data acquisition unit 81 acquires, via the communication network, a second training data set including information on the second property of each complex oxide in the second complex oxide group and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment from an external device connected to the prediction device 1 via the communication network. The second data acquisition unit 81 may acquire the second training data set via the communication network from an external device connected to the prediction device 1 via the communication network in accordance with an instruction or condition input by a user via the input device 104.

[0154] In yet another embodiment, the second data acquisition unit 81 acquires a second training data set including information on the second property of each complex oxide in the second complex oxide group and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment from the recording medium 108. The second data acquisition unit 81 may acquire the second training data set from the recording medium 108 in accordance with an instruction or condition input by a user via the input device 104.

[0155] The second learning dataset acquired by the second data acquisition unit 81 is used as a learning dataset (also called a teacher dataset) in which information regarding the second characteristic of each complex oxide in the second complex oxide group is used as an explanatory variable, and information regarding the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment is used as a target variable.

[0156] From the viewpoint of improving the prediction accuracy by the second trained model 3, the number of pairs of information on the second property of the Ce-Zr-M2-based composite oxide that does not undergo phase separation by heat treatment and information on the oxygen storage capacity of the Ce-Zr-M2-based composite oxide after heat treatment, which are included in the second training data set, is preferably 5 or more, more preferably 10 or more, and even more preferably 50 or more. The upper limit is not particularly limited, and can be appropriately adjusted in consideration of the time required to generate the second trained model 3, etc. The upper limit is, for example, 100,000 or less.

[0157] The second learning unit 82 performs machine learning using the second learning data set acquired by the second data acquisition unit 81, and generates a second trained model 3.

[0158] Examples of machine learning methods performed by the second learning unit 82 include logistic regression, random forest, simple linear regression, Ridge regression, Lasso regression, Elastic Net regression, general additive model, rule fit regression, gradient boosting tree, extra tree, support vector regression, Gaussian process regression, k-nearest neighbor regression, kernel ridge regression, and neural network.

[0159] The generated second trained model 3 is stored, for example, in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, etc.

[0160] <Prediction device processing> Hereinafter, the processing procedure of the prediction device 1 will be described with reference to FIG.

[0161] 3 is a flowchart of the processing procedure of the prediction device 1. Each process is executed by the processor 101 of the computer 100. The processor 101 executes a program to function as the first prediction unit 4, the determination unit 5, and the second prediction unit 6, and as the first model generation unit 7 and / or the second model generation unit 8 as necessary. When some of the first prediction unit 4, the determination unit 5, the second prediction unit 6, the first model generation unit 7, and the second model generation unit 8 are constructed on a certain computer 100 and the remaining units are constructed on another computer 100, some of the processes may be executed by the processor 101 of the certain computer 100, and the remaining processes may be executed by the processor 101 of the other computer 100.

[0162] In step S11, a first trained model 2 is prepared. The first trained model 2 may be generated by the first model generation unit 7, or may be generated by other means. In step S11, the first model generation unit 7 may generate the first trained model 2.

[0163] In step S12, the second trained model 3 is prepared. The second trained model 3 may be generated by the second model generation unit 8, or may be generated by other means. In step S12, the second model generation unit 8 may generate the second trained model 3.

[0164] Of steps S11 and S12, one may be executed before the other, or both may be executed in parallel.

[0165] In step S13, the first prediction unit 4 uses the first trained model 2 and information on the first characteristic of the Ce-Zr-M0-based composite oxide to obtain information on whether or not the Ce-Zr-M0-based composite oxide will undergo phase separation due to heat treatment. Specifically, the first prediction unit 4 obtains information on the first characteristic of the Ce-Zr-M0-based composite oxide and inputs the obtained information to the first trained model 2. Based on the input information, the first trained model 2 predicts whether or not the Ce-Zr-M0-based composite oxide will undergo phase separation due to heat treatment, and outputs information on whether or not the Ce-Zr-M0-based composite oxide will undergo phase separation due to heat treatment. The first prediction unit 4 obtains the information output from the first trained model 2.

[0166] The first prediction unit 4 obtains information on whether or not phase separation occurs in the Ce—Zr—M0 based composite oxide by heat treatment, and then provides the obtained information to the determination unit 5 .

[0167] In one embodiment, the information on the first characteristic of the Ce-Zr-M0 based composite oxide is input by a user via the input device 104, and the first prediction unit 4 receives the information on the first characteristic of the Ce-Zr-M0 based composite oxide input by the user via the input device 104, thereby acquiring the information on the first characteristic of the Ce-Zr-M0 based composite oxide.

[0168] In another embodiment, information on the first characteristic of the Ce-Zr-M0 based composite oxide is stored in the main memory unit 102 or the auxiliary memory unit 103, and the first prediction unit 4 acquires the information on the first characteristic of the Ce-Zr-M0 based composite oxide from the main memory unit 102 or the auxiliary memory unit 103 in accordance with an instruction or condition input by a user via the input device 104.

[0169] In still another embodiment, information about the first characteristic of the Ce-Zr-M0-based composite oxide is stored in an external device connected to the prediction device 1 via a communication network, and the first prediction unit 4 acquires the information about the first characteristic of the Ce-Zr-M0-based composite oxide from the external device via the communication network in accordance with an instruction or condition input by a user via the input device 104.

[0170] In still another embodiment, the information on the first characteristic of the Ce-Zr-M0 based composite oxide is stored in a recording medium 108, and the first prediction unit 4 acquires the information on the first characteristic of the Ce-Zr-M0 based composite oxide from the recording medium 108 in accordance with an instruction or condition input by a user via the input device 104.

[0171] In step S14, the judgment unit 5 judges whether or not the information acquired by the first prediction unit 4 (information provided by the first prediction unit 4) is information indicating that the Ce-Zr-M0-based composite oxide does not undergo phase separation by heat treatment.

[0172] The determination unit 5 obtains, as a determination result, information that the Ce-Zr-M0 based composite oxide does not undergo phase separation by heat treatment or information that the Ce-Zr-M0 based composite oxide undergoes phase separation by heat treatment.

[0173] A case in which the information acquired by the first prediction unit 4 (information provided by the first prediction unit 4) indicates that the Ce-Zr-M0 based composite oxide will not undergo phase separation due to heat treatment will be referred to as a "case in which the judgment result is YES", and a case in which the information acquired by the first prediction unit 4 (information provided by the first prediction unit 4) indicates that the Ce-Zr-M0 based composite oxide will undergo phase separation due to heat treatment will be referred to as a "case in which the judgment result is NO".

[0174] When the result of the determination is YES, the determining unit 5 provides the second predicting unit 6 with information to the effect that the Ce-Zr-M0 based composite oxide will not undergo phase separation due to the heat treatment.

[0175] When the judgment result is YES (i.e., when the judgment unit 5 provides the second prediction unit 6 with information that the Ce-Zr-M0-based composite oxide does not undergo phase separation by heat treatment), in step S15, the second prediction unit 6 uses the second trained model 3 and information on the second characteristic of the Ce-Zr-M0-based composite oxide to acquire information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment. Specifically, it is as follows. The second prediction unit 6 acquires information on the second characteristic of the Ce-Zr-M0-based composite oxide and inputs the acquired information to the second trained model 3. The second trained model 3 predicts the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment based on the input information, and outputs information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment. The second prediction unit 6 acquires the information output from the second trained model 3.

[0176] In one embodiment, the information on the second characteristic of the Ce-Zr-M0 based composite oxide is input by a user via the input device 104, and the second prediction unit 6 acquires the information on the second characteristic of the Ce-Zr-M0 based composite oxide by accepting the information on the second characteristic of the Ce-Zr-M0 based composite oxide input by the user via the input device 104.

[0177] In another embodiment, information on the second characteristic of the Ce-Zr-M0 based composite oxide is stored in the main memory unit 102 or the auxiliary memory unit 103, and the second prediction unit 6 acquires the information on the second characteristic of the Ce-Zr-M0 based composite oxide from the main memory unit 102 or the auxiliary memory unit 103 in accordance with an instruction or condition input by a user via the input device 104.

[0178] In still another embodiment, information regarding the second property of the Ce-Zr-M0-based composite oxide is stored in an external device connected to the prediction device 1 via a communication network, and the second prediction unit 6 acquires the information regarding the second property of the Ce-Zr-M0-based composite oxide from the external device via the communication network in accordance with an instruction or condition input by a user via the input device 104.

[0179] In still another embodiment, the information on the second property of the Ce-Zr-M0 based composite oxide is stored in a recording medium 108, and the second prediction unit 6 acquires the information on the second property of the Ce-Zr-M0 based composite oxide from the recording medium 108 in accordance with an instruction or condition input by a user via the input device 104.

[0180] After acquiring the information regarding the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment in step S15, the second prediction unit 6 may provide the information regarding the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment to the output device 105, together with information that the Ce-Zr-M0 based composite oxide does not undergo phase separation due to the heat treatment, as necessary, and the output device 105 may output the information regarding the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment, together with information that the Ce-Zr-M0 based composite oxide does not undergo phase separation due to the heat treatment, as necessary.

[0181] After acquiring information on the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment in step S15, the second prediction unit 6 acquires predetermined information based on the information on the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment, and may provide the predetermined information to the output device 105, if necessary, together with information that the Ce-Zr-M0 based composite oxide does not undergo phase separation due to the heat treatment, and the output device 105 may output the predetermined information, if necessary, together with information that the Ce-Zr-M0 based composite oxide does not undergo phase separation due to the heat treatment. The predetermined information is as described above.

[0182] If the determination result is NO, the determination unit 5 may provide the second prediction unit 6 with information to the effect that phase separation will occur in the Ce-Zr-M0 based composite oxide by heat treatment, but it is preferable not to provide such information.

[0183] When the determination result is NO, the determination unit 5 may provide information to the output device 105 that the Ce-Zr-M0 based composite oxide will undergo phase separation due to the heat treatment, and the output device 105 may output information that the Ce-Zr-M0 based composite oxide will undergo phase separation due to the heat treatment.

[0184] <Processing of the first model generation unit> Hereinafter, the processing procedure of the first model generating unit 7 will be described with reference to FIG.

[0185] 4 is a flowchart of the processing procedure of the first model generating unit 7. Each step is executed by the processor 101 of the computer 100. The processor 101 functions as the first model generating unit 7 by executing a program.

[0186] In step S21, the first data acquisition unit 71 acquires a first learning data set including information on a first characteristic of each complex oxide in the first complex oxide group and information on whether each complex oxide in the first complex oxide group undergoes phase separation upon heat treatment.

[0187] Information regarding the first characteristic of each complex oxide in the first complex oxide group and information regarding whether or not each complex oxide in the first complex oxide group undergoes phase separation by heat treatment are associated with each other and stored in, for example, the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, or another external device.

[0188] In one embodiment, the first data acquisition unit 71 acquires a first training data set including information on the first property of each complex oxide in the first complex oxide group and information on whether each complex oxide in the first complex oxide group undergoes phase separation by heat treatment from the main memory unit 102 or the auxiliary memory unit 103. The first data acquisition unit 71 may acquire the first training data set from the main memory unit 102 or the auxiliary memory unit 103 in accordance with an instruction or condition input by a user via the input device 104.

[0189] In another embodiment, the first data acquiring unit 71 acquires, via the communication network, from an external device connected to the prediction device 1 via the communication network, a first training data set including information on the first property of each complex oxide in the first complex oxide group and information on whether each complex oxide in the first complex oxide group undergoes phase separation by heat treatment. The first data acquiring unit 71 may acquire the first training data set via the communication network from the external device connected to the prediction device 1 via the communication network, in accordance with an instruction or condition input by a user via the input device 104.

[0190] In yet another embodiment, the first data acquiring unit 71 acquires a first training data set including information on a first property of each complex oxide in the first complex oxide group and information on whether each complex oxide in the first complex oxide group undergoes phase separation by heat treatment from the recording medium 108. The first data acquiring unit 71 may acquire the first training data set from the recording medium 108 in accordance with an instruction or condition input by a user via the input device 104.

[0191] In step S22, the first learning unit 72 performs machine learning using the first learning data set acquired by the first data acquisition unit 71, and generates a first trained model 2.

[0192] The generated first trained model 2 is stored, for example, in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, etc.

[0193] <Processing of the second model generation section> Hereinafter, the processing procedure of the second model generating unit 8 will be described with reference to FIG.

[0194] 5 is a flowchart of the processing procedure of the second model generating unit 8. Each step is executed by the processor 101 of the computer 100. The processor 101 functions as the second model generating unit 8 by executing a program.

[0195] In step S31, the second data acquisition unit 81 acquires a second learning data set including information on the second characteristic of each complex oxide in the second complex oxide group and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment.

[0196] Information regarding the second property of each complex oxide in the second complex oxide group and information regarding the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment are associated with each other and stored, for example, in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, or other external device.

[0197] In one embodiment, the second data acquisition unit 81 acquires a second training data set including information on the second property of each complex oxide in the second complex oxide group and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment from the main memory unit 102 or the auxiliary memory unit 103. The second data acquisition unit 81 may acquire the second training data set from the main memory unit 102 or the auxiliary memory unit 103 in accordance with an instruction or condition input by a user via the input device 104.

[0198] In another embodiment, the second data acquisition unit 81 acquires, via the communication network, a second training data set including information on the second property of each complex oxide in the second complex oxide group and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment from an external device connected to the prediction device 1 via the communication network. The second data acquisition unit 81 may acquire the second training data set via the communication network from an external device connected to the prediction device 1 via the communication network in accordance with an instruction or condition input by a user via the input device 104.

[0199] In yet another embodiment, the second data acquisition unit 81 acquires a second training data set including information on the second property of each complex oxide in the second complex oxide group and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after heat treatment from the recording medium 108. The second data acquisition unit 81 may acquire the second training data set from the recording medium 108 in accordance with an instruction or condition input by a user via the input device 104.

[0200] In step S32, the second learning unit 82 performs machine learning using the second learning data set acquired by the second data acquisition unit 81, and generates a second trained model 3.

[0201] The generated second trained model 3 is stored, for example, in the main memory unit 102, the auxiliary memory unit 103, the recording medium 108, etc.

[0202] <Program> The program for making the computer 100 function as the prediction device 1 (i.e., the program for making the computer 100 execute steps S11 to S15, and steps S21 to S22 and / or steps S31 to S32 as necessary) includes program code for making the processor 101 of the computer 100 function as the first prediction unit 4, the determination unit 5, and the second prediction unit 6, and as necessary, the first model generation unit 7 and / or the second model generation unit 8. The program may be provided after being recorded on a computer-readable recording medium. The computer-readable recording medium is, for example, a tangible (i.e., non-volatile) recording medium such as a CD-ROM, a DVD-ROM, or a semiconductor memory. Alternatively, the program may be provided via a communication network as a data signal superimposed on a carrier wave. The provided program is stored, for example, in the main memory unit 102, the auxiliary memory unit 103, or the like. The processor 101 reads out the program from the main memory unit 102, the auxiliary memory unit 103, or the like and executes it, thereby realizing each of the above-mentioned functional elements.

[0203] Effect When the Ce-Zr-M0 based composite oxide undergoes phase separation upon heat treatment, there is a low correlation between the information on the second property of the Ce-Zr-M0 based composite oxide and the information on the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after heat treatment. In contrast, when the Ce-Zr-M0 based composite oxide does not undergo phase separation upon heat treatment, there is a high correlation between the information on the second property of the Ce-Zr-M0 based composite oxide and the information on the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after heat treatment. Therefore, in the present invention, the first trained model and information on the first characteristic of the Ce-Zr-M0 composite oxide are used to obtain information on whether or not the Ce-Zr-M0 composite oxide undergoes phase separation upon heat treatment, and it is determined whether or not the obtained information is information on the Ce-Zr-M0 composite oxide not undergoing phase separation upon heat treatment. If the obtained information is information on the Ce-Zr-M0 composite oxide not undergoing phase separation upon heat treatment, the second trained model and information on the second characteristic of the Ce-Zr-M0 composite oxide are used to obtain information on the oxygen storage capacity of the Ce-Zr-M0 composite oxide after heat treatment. The thus obtained information on the oxygen storage capacity of the Ce-Zr-M0 composite oxide after heat treatment sufficiently reflects the actual state of the oxygen storage capacity of the Ce-Zr-M0 composite oxide after heat treatment. That is, according to the present invention, the oxygen storage capacity of the Ce-Zr-M0 composite oxide after heat treatment can be accurately predicted.

[0204] <<Variations>> As described above, when the determination result is NO, it is preferable that the determination unit 5 not provide the second prediction unit 6 with information to the effect that the Ce-Zr-M0 based composite oxide will undergo phase separation due to heat treatment. This is because, when the Ce-Zr-M0 based composite oxide will undergo phase separation due to heat treatment, there is a low correlation between the information on the second property of the Ce-Zr-M0 based composite oxide and the information on the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after heat treatment.

[0205] However, if the judgment result is NO, the judgment unit 5 may provide the second prediction unit 6 with information that the Ce-Zr-M0-based composite oxide will undergo phase separation due to heat treatment. When the judgment unit 5 provides the second prediction unit 6 with information that the Ce-Zr-M0-based composite oxide will undergo phase separation due to heat treatment, the second prediction unit 6 may acquire information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment using the second trained model 3 and information on the second characteristic of the Ce-Zr-M0-based composite oxide. There is a high possibility that the information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment acquired in this manner does not sufficiently reflect the actual state of the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment. Therefore, for the purpose of alerting the user, it is preferable that the information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment is provided to the user together with information that the Ce-Zr-M0-based composite oxide will undergo phase separation due to the heat treatment. For example, after acquiring the information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment, the second prediction unit 6 may provide the information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment together with information that the Ce-Zr-M0-based composite oxide will undergo phase separation due to the heat treatment to the output device 105, and the output device 105 may output the information on the oxygen storage capacity of the Ce-Zr-M0-based composite oxide after the heat treatment together with information that the Ce-Zr-M0-based composite oxide will undergo phase separation due to the heat treatment. Alternatively, the second prediction unit 6 may obtain information on the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment, and then obtain predetermined information based on the information on the oxygen storage capacity of the Ce-Zr-M0 based composite oxide after the heat treatment, and provide the predetermined information to the output device 105 together with information that the Ce-Zr-M0 based composite oxide will undergo phase separation due to the heat treatment, and the output device 105 may output the predetermined information together with information that the Ce-Zr-M0 based composite oxide will undergo phase separation due to the heat treatment. The predetermined information is as described above. EXAMPLES

[0206] <Sample preparation> A total of 17 kinds of additive elements were selected: 3 kinds from alkaline earth metal elements, 5 kinds from transition elements of Groups 4 to 11, 7 kinds from rare earth elements, and 2 kinds from other metal elements (metal elements other than alkaline earth metal elements, transition elements of Groups 4 to 11, and rare earth elements). Using the selected 17 kinds of additive elements, 17 kinds of Ce-Zr-M composite oxides (wherein, each M independently represents an additive element) each containing one different kind of additive element were prepared. Specifically, the process is as follows. A water-soluble salt of cerium, a water-soluble salt of zirconium, and a water-soluble salt of an additive element were dissolved in water to prepare an aqueous solution, a base was added to the prepared aqueous solution, and the resulting precipitate was filtered, dried, and fired to prepare the 17 kinds of Ce-Zr-M composite oxides.

[0207] In all of the 17 types of Ce-Zr-M based composite oxides, R11=33 mol%, R21=62 mol%, and R3=5 mol%. The explanations for R11, R21, and R3 are as above.

[0208] <Heat treatment> The above 17 types of Ce-Zr-M based composite oxides were used as Samples No. 1 to No. 17. Each sample was in powder form.

[0209] Each sample was subjected to a heat treatment at 1000°C for 25 hours under hydrothermal conditions (specifically, under conditions in which an atmosphere containing 0.50 vol% O2 gas, 10 vol% water vapor, and N2 gas as the balance gas was circulated).

[0210] <Determining the presence or absence of phase separation> The X-ray diffraction patterns of each sample after the heat treatment were obtained by X-ray diffraction (XRD). XRD was performed under the following conditions. X-ray diffraction equipment: Rigaku MiniFlex600 X-ray source:CuKα Operation axis: 2θ / θ Measurement method: Continuous Counting unit: cps Starting angle: 5° End angle: 80° Sampling width: 0.02° Scan speed: 7° / min Voltage: 15kV Current: 100mA

[0211] Whether or not phase separation occurred in each sample after the heat treatment was judged based on the X-ray diffraction pattern of each sample after the heat treatment. Specifically, it is as follows. When one peak was observed at 2θ=27 to 32° in the X-ray diffraction pattern of the sample after the heat treatment, it was judged that no phase separation occurred in the sample after the heat treatment. On the other hand, when two or more peaks were observed at 2θ=27 to 32° in the X-ray diffraction pattern of the sample after the heat treatment, it was judged that phase separation occurred in the sample after the heat treatment.

[0212] <OSC value measurement> The OSC value of each sample after the heat treatment was measured by the CO pulse method. Specifically, it is as follows. The CO pulse method was performed using a temperature-programmed reduction treatment device (TPR device) manufactured by Hemmy Slide Rule Co., Ltd. The sample after the heat treatment was heated to 800°C under He gas flow, and the sample was pretreated by holding the temperature for 40 minutes, and then cooled to 500°C. Next, while holding the sample at 500°C, O2 gas was injected in four pulses to perform an oxidation treatment, and then a test gas containing CO was injected in 15 pulses to measure the OSC value (μmol / g) per unit mass of each sample at 500°C from the total amount of CO gas consumed (μmol). The total amount of CO gas consumed was measured using an infrared gas concentration measurement device (Shimadzu Corporation, CGT-7000). The test gas was prepared by diluting CO gas to 25% by volume in He gas.

[0213] The OSC value (μmol / g) per unit mass of each sample was calculated based on the following formula. OSC value per unit mass of each sample (μmol / g) = total amount of CO gas consumed (μmol) / Ce content in each sample converted to CeO2 (g)

[0214] Table 1 shows the results of determining whether or not there was phase separation and the measurement results of the OSC value.

[0215] [Table 1]

[0216] Example 1 and Comparative Example 1 In Example 1, samples No. 2, No. 3, and No. 9 to 12, which did not undergo phase separation by heat treatment, were used as the second composite oxide group. In Comparative Example 1, samples No. 2, No. 3, and No. 9 to 12, which did not undergo phase separation by heat treatment, and samples No. 1, No. 4 to 8, No. 13, and No. 14, which did undergo phase separation by heat treatment, were used as the second composite oxide group.

[0217] In both Example 1 and Comparative Example 1, 108 types of characteristic values ​​selected from the characteristic values ​​recorded in XenonPy were used as the second characteristic values ​​of each complex oxide in the second complex oxide group.

[0218] The 108 characteristic values ​​used are shown in Figure 6. In Figure 6, "ave" represents the weighted average of the characteristic values ​​of Ce, Zr, and M, and "var" represents the variance of the characteristic values ​​of Ce, Zr, and M. In Figure 6, "atomic_radius" is the atomic radius, "atomic_radius_rahm" is the atomic radius according to Rahm, "covalent_radius_cordero" is the covalent radius according to Cerdero, "covalent_radius_pyykko" is the single bond distance of the covalent radius according to Pyykko, "covalent_radius_pyykko_double" is the double bond distance of the covalent radius according to Pyykko, "covalent_radius_pyykko_triple" is the triple bond distance of the covalent radius according to Pyykko, "covalent_radius_slater" is the covalent radius according to Slater, and "vdw_radius" is the fan der Waals radius, "vdw_radius_alvarez" is the van der Waals radius according to Alvarez, "vdw_radius_mm3" is the van der Waals radius in MM3, "vdw_radius_uff" is the van der Waals radius in UFF, "molar_volume" is the molar volume, "density" is the density, "heat_capacity_mass" is the mass heat capacity, "heat_capacity_molar" is the molar heat capacity, "electron_negativity" is the electronegativity, "en_allen" is the electronegativity according to Allen, "en_ghosh" is the electronegativity according to Ghosh, and "en_pauling" is the electronegativity according to Pauling.

[0219] The second model generation unit 8 was caused to generate the second trained model 3. Specifically, the second learning unit 82 was caused to perform machine learning using a second training data set in which the second characteristic value of each complex oxide in the second complex oxide group was used as an explanatory variable and the OSC value (actual value) after heat treatment of each complex oxide in the second complex oxide group was used as an explanatory variable, thereby generating the second trained model 3. Random Forest was used as the machine learning method performed by the second learning unit 82.

[0220] The second trained model 3 generated by the second model generating unit 8 was made to predict the OSC value of each complex oxide in the second complex oxide group after heat treatment based on the second characteristic value of each complex oxide in the second complex oxide group. The correspondence between the OSC value (actual value) of each complex oxide in the second complex oxide group after heat treatment and the OSC value (predicted value) of each complex oxide in the second complex oxide group after heat treatment was analyzed. In Example 1, the Q2 value, which is an index of the prediction accuracy of the second trained model 3, was 0.429. In Comparative Example 1, the Q2 value, which is an index of the prediction accuracy of the second trained model 3, was 0.048. The Q2 value is a value between 0 and 1. A Q2 value of 0.4 or more is sufficient for practical use.

[0221] Regarding the second characteristic values ​​of each complex oxide in the second complex oxide group, the weighted average value of density ("ave:density"), the weighted average value of electronegativity ("ave:en_ghosh"), and the weighted average value of heat capacity ("ave:heat_capacity_molar"), the correspondence relationship with the OSC value (actual value) after heat treatment of each complex oxide in the second complex oxide group is shown in Figure 7. As shown in Figure 7, the weighted average value of density ("ave:density"), the weighted average value of electronegativity ("ave:en_ghosh"), and the weighted average value of heat capacity ("ave:heat_capacity_molar") each had a high correlation with the OSC value (actual value) after heat treatment of each complex oxide in the second complex oxide group.

[0222] Example 2 (1) Generation of the second trained model In Example 2, the weighted average value of the density ("ave:density"), which had the highest correlation coefficient and was ranked in the top 54 among the second characteristics used in Example 1, was used as the second characteristic value. The other conditions were the same as in Example 1.

[0223] (2) Generation of the first trained model In Example 2, Samples No. 2, No. 3, and No. 9 to 12 in which phase separation did not occur by heat treatment, and Samples No. 1, No. 4 to 8, No. 13, and No. 14 in which phase separation occurred by heat treatment were used as the first composite oxide group.

[0224] In Example 2, the atomic radius of the M element ("atomic_radius"), the covalent bond radius of the M element ("covalent_radius_pyykko"), the van der Waals radius of the M element ("vdw_radius"), and the molar volume of the M element ("molar_volume"), selected from the property values ​​included in XenonPy, were used as the first property values ​​of each complex oxide in the first complex oxide group.

[0225] The first model generation unit 7 was made to generate the first trained model 2. Specifically, the first characteristic value of each complex oxide in the first complex oxide group was used as an explanatory variable, and the first learning unit 72 was made to perform machine learning using a first learning dataset in which information on whether or not each complex oxide in the first complex oxide group has undergone phase separation due to heat treatment was used as an explanatory variable, thereby generating the first trained model 2. As information on whether or not each complex oxide in the first complex oxide group has undergone phase separation due to heat treatment, a value corresponding to whether or not each complex oxide in the first complex oxide group has undergone phase separation due to heat treatment (a value of 0 when phase separation has occurred, and a value of 1 when phase separation has not occurred) was used. Logistic regression was used as the machine learning method performed by the first learning unit 72.

[0226] The correspondence between the first characteristic value of each complex oxide in the first complex oxide group (atomic radius of M element, covalent bond radius of M element, van der Waals radius of M element, and molar volume of M element) and information on whether each complex oxide in the first complex oxide group has undergone phase separation by heat treatment is shown in Fig. 8. The horizontal axis in Fig. 8 represents the standardized value obtained by dividing the first characteristic value of each complex oxide in the first complex oxide group (atomic radius of M element, covalent bond radius of M element, van der Waals radius of M element, and molar volume of M element) by the minimum value of the first characteristic values ​​of all complex oxides in the first complex oxide group (minimum atomic radius of M element, minimum covalent bond radius of M element, minimum van der Waals radius of M element, and minimum molar volume of M element). As shown in Fig. 8, there was a high correlation between the first characteristic value (atomic radius of M element, covalent radius of M element, van der Waals radius of M element, and molar volume of M element) of each complex oxide in the first complex oxide group and information on whether or not each complex oxide in the first complex oxide group has undergone phase separation by heat treatment. Note that the covalent radius shown in Fig. 8 is "covalent_radius_pyykko", but the other covalent radii also had a high correlation with information on whether or not each complex oxide in the first complex oxide group has undergone phase separation by heat treatment.

[0227] (3) Model validation Using the first trained model 2 and the first characteristic values ​​(atomic radius of element M, covalent bond radius of element M, van der Waals radius of element M, and molar volume of element M) of each of sample Nos. 15 to 17, information was obtained as to whether or not each of sample Nos. 15 to 17 would undergo phase separation upon heat treatment. The obtained information was information indicating that each of sample Nos. 15 to 17 would not undergo phase separation upon heat treatment. In other words, the first trained model 2 predicted that each of sample Nos. 15 to 17 would not undergo phase separation upon heat treatment. The prediction results reflected the actual conditions of sample Nos. 15 to 17.

[0228] Using the second trained model 3 and the second characteristic value (weighted average value of density) of each of Samples No. 15 to 17, the OSC value (predicted value) after heat treatment of each of Samples No. 15 to 17 was obtained. The obtained OSC value (predicted value) was 181 μmol / g for Sample No. 15, 154 μmol / g for Sample No. 16, and 238 μmol / g for Sample No. 17, which fully reflected the actual state of Samples No. 15 to 17 (OSC value (actual value) shown in Table 1). [Explanation of symbols]

[0229] 1. Prediction device 2. First trained model 3. Second trained model 4. First prediction section 5... Judgment section 6. Second prediction section 7. First model generation part 71...First data acquisition unit 72...First Learning Section 8. Second model generation part 81...Second data acquisition unit 82...Second Learning Section

Claims

1. Ce-Zr-M ０ Based composite oxide (wherein, M ０ represents an additive element.) after heat treatment, a first trained model generated by machine learning using information about a first characteristic of each composite oxide in a first composite oxide group as an explanatory variable and information about whether each composite oxide in the first composite oxide group undergoes phase separation upon heat treatment as a target variable, wherein the first composite oxide group is a plurality of Ce—Zr—M １ The composite oxide and the multiple Ce-Zr-M oxides that undergo phase separation by heat treatment. １ and M-based composite oxide. １ each independently represents an additive element; a second trained model generated by machine learning using information about a second characteristic of each composite oxide in a second composite oxide group as an explanatory variable and information about the oxygen storage capacity of each composite oxide in the second composite oxide group after a heat treatment as a target variable, wherein the second composite oxide group is a plurality of Ce-Zr-M ２ Ce-Zr-M composite oxides are included, but phase separation occurs upon heat treatment. ２ The material properties of M-based composite oxides are not included. ２ each independently represents an additive element; The first trained model and the Ce—Zr-M ０ Using information on the first property of the Ce-Zr-M system composite oxide, ０ a first prediction unit that acquires information regarding whether or not the composite oxide undergoes phase separation upon heat treatment; The information acquired by the first prediction unit is ０ A determination unit that determines whether the information indicates that the composite oxide will not undergo phase separation by heat treatment; and The information acquired by the first prediction unit is ０ When the information indicates that the Ce-Zr-M system composite oxide does not undergo phase separation by heat treatment, ０ Using information on the second property of the Ce-Zr-M system composite oxide, ０ A second prediction unit for acquiring information on the oxygen storage capacity of the composite oxide after heat treatment. An apparatus comprising:

2. 2. The apparatus of claim 1, wherein the first characteristic comprises at least one selected from the group consisting of an atomic radius, a covalent bond radius, a van der Waals radius, a metallic bond radius, and a molar volume.

3. The device of claim 1 or 2, wherein the second property comprises at least one selected from the group consisting of density, heat capacity, and electronegativity.

4. Ce-Zr-M ０ Based composite oxide (wherein, M ０ represents an additive element.) after heat treatment, comprising the steps of: Steps below: (a) preparing a first trained model generated by machine learning using information on a first characteristic of each composite oxide of a first composite oxide group as an explanatory variable and information on whether each composite oxide of the first composite oxide group undergoes phase separation by heat treatment as an objective variable, wherein the first composite oxide group is a plurality of Ce-Zr-M １ The composite oxide and the multiple Ce-Zr-M oxides that undergo phase separation by heat treatment. １ and M-based composite oxide. １ each independently represents an additive element; (b) preparing a second trained model generated by machine learning using information on the second characteristic of each complex oxide in a second complex oxide group as an explanatory variable and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after a heat treatment as a target variable, wherein the second complex oxide group is a plurality of Ce-Zr-M oxides that do not undergo phase separation by a heat treatment. ２ Ce-Zr-M composite oxides are included, but phase separation occurs upon heat treatment. ２ The material properties of M-based composite oxides are not included. ２ each independently represents an additive element; (c) The first trained model and the Ce—Zr-M ０ Using information on the first property of the Ce-Zr-M system composite oxide, ０ acquiring information on whether the composite oxide undergoes phase separation upon heat treatment; (d) The information obtained in step (c) is ０ A step of determining whether the information indicates that the composite oxide does not undergo phase separation by heat treatment; and (e) The information obtained in step (c) is ０ When the information indicates that the Ce-Zr-M system composite oxide does not undergo phase separation by heat treatment, ０ Using information on the second property of the Ce-Zr-M system composite oxide, ０ Obtaining information regarding the oxygen storage capacity of the composite oxide after heat treatment. The method includes:

5. 5. The method of claim 4, wherein the first characteristic comprises at least one selected from the group consisting of atomic radius, covalent bond radius, van der Waals radius, metallic bond radius, and molar volume.

6. The method of claim 4 or 5, wherein the second property comprises at least one selected from the group consisting of density, heat capacity, and electronegativity.

7. Ce-Zr-M ０ Based composite oxide (wherein, M ０ represents an additive element.) after heat treatment, Steps below: (a) preparing a first trained model generated by machine learning using information on a first characteristic of each composite oxide of a first composite oxide group as an explanatory variable and information on whether each composite oxide of the first composite oxide group undergoes phase separation by heat treatment as an objective variable, wherein the first composite oxide group is a plurality of Ce-Zr-M １ The composite oxide and the multiple Ce-Zr-M oxides that undergo phase separation by heat treatment. １ and M-based composite oxide. １ each independently represents an additive element; (b) preparing a second trained model generated by machine learning using information on the second characteristic of each complex oxide in a second complex oxide group as an explanatory variable and information on the oxygen storage capacity of each complex oxide in the second complex oxide group after a heat treatment as a target variable, wherein the second complex oxide group is a plurality of Ce-Zr-M oxides that do not undergo phase separation by a heat treatment. ２ Ce-Zr-M composite oxides are included, but phase separation occurs upon heat treatment. ２ The material properties of M-based composite oxides are not included. ２ each independently represents an additive element; (c) The first trained model and the Ce—Zr-M ０ Using information on the first property of the Ce-Zr-M system composite oxide, ０ acquiring information on whether the composite oxide undergoes phase separation upon heat treatment; (d) The information obtained in step (c) is ０ A step of determining whether the information indicates that the composite oxide does not undergo phase separation by heat treatment; and (e) The information obtained in step (c) is ０ When the information indicates that the Ce-Zr-M system composite oxide does not undergo phase separation by heat treatment, ０ Using information on the second property of the Ce-Zr-M system composite oxide, ０ Obtaining information regarding the oxygen storage capacity of the composite oxide after heat treatment. A program that causes a computer to execute the following.

8. The program of claim 7 , wherein the first property includes at least one selected from the group consisting of an atomic radius, a covalent bond radius, a van der Waals radius, a metallic bond radius, and a molar volume.

9. The program according to claim 7 or 8, wherein the second property includes at least one selected from the group consisting of density, heat capacity, and electronegativity.

Images