Catalyst search method

Abstract

To provide a catalyst search method capable of efficiently searching a combination of elements which have preferable characteristics.SOLUTION: There is provided a catalyst search method which is a method for searching a catalyst which is formed of a two-elements alloy for reducing activation energy to a specific chemical reaction, and the method comprises: an analyzing step for, to a plurality of two-elements alloys, analyzing the activation energy of each two-elements alloy, by first principle calculation; a prediction model constructing step for constructing a prediction model in which a material basic physical value of the two elements forming each two-elements alloy is set to an explanatory variable, and the activation energy analyzed in the analyzing step is set to an objective variable; a reverse analyzing step for calculating the material basic physical value for minimizing the activation energy, using the prediction model; and a two-elements selecting step for, on the basis of the material basic physical value obtained in the reverse analyzing step, selecting two elements which exist really.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method for searching for catalysts. [Background technology]

[0002] For example, nitrogen oxide (NOx) in gases emitted from automobile engines is decomposed using catalysts whose main components are platinum (Pt), rhodium (Rh), palladium (Pd), etc., and is released into the atmosphere as N2. All of these catalysts are rare metals, and there are concerns about their high prices and resource depletion. For this reason, the development of new catalysts, especially those made from two-element alloys, is attracting attention.

[0003] The search for catalysts consisting of two-element alloys has generally been carried out through trial and error, in which two elements are selected based on proven materials and prototypes are repeatedly evaluated. However, in recent years, evaluation of the catalytic activity of alloy elements based on first-principles calculations has also been used (Non-Patent Document 1).

[0004] In the catalytic activity evaluation, once the alloy type of the catalyst is specified, the activation energy when that catalyst is used for a certain chemical reaction can be calculated. If this activation energy is reduced, it is understood that the catalyst is effective. [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] Machine-learning prediction of the d-band center for metals and bimetals, Ichigaku Takigawa, Ken-ichi Shimizu, and Satoru Takakusagi, Royal Soc. Chem. Adv., 6, 52587-52595 (2016) Summary of the Invention [Problem to be solved by the invention]

[0006] Although the above catalytic activity evaluation is more efficient than repeated prototype evaluation, it still requires trial and error to identify the elements. Because there are a huge number of combinations of two elements and the calculation load is heavy, it is difficult to find the combination of elements with the best properties.

[0007] The present invention has been made in light of the above-mentioned circumstances, and aims to provide a method for searching for catalysts that can efficiently search for combinations of elements with good properties. [Means for solving the problem]

[0008] A catalyst searching method according to one embodiment of the present invention is a method for searching for a catalyst consisting of a two-element alloy that reduces the activation energy for a specific chemical reaction, and includes an analysis step of analyzing the activation energy of each two-element alloy using first-principles calculations for a plurality of two-element alloys; a prediction model construction step of constructing a prediction model in which the basic material properties of the two elements that constitute each of the two-element alloys are used as explanatory variables and the activation energy analyzed in the analysis step is used as the target variable; an inverse analysis step of calculating the basic material properties that minimize the activation energy using the prediction model; and a two-element selection step of selecting two real elements based on the basic material properties obtained in the inverse analysis step.

[0009] In this catalyst search method, the activation energy of each of a plurality of binary alloys is analyzed using first-principles calculations, and then a predictive model is constructed based on the obtained results, with the activation energy as the objective variable. In this catalyst search method, the predictive model is used to perform inverse analysis to calculate the fundamental material properties that minimize the activation energy, and two real elements that are close to the fundamental material properties that minimize the activation energy are selected, thereby making it possible to efficiently search for combinations of elements with good properties.

[0010] In the above analysis step, it is preferable to assume that the two-element alloy is an atomic cluster consisting of X host atoms and Y guest atoms (X and Y are natural numbers satisfying the condition 10≧X≧Y). By assuming an atomic cluster in this way in the above analysis step, the calculation scale can be significantly reduced. This makes it possible to increase the number of two-element alloys analyzed in the above analysis step while suppressing an increase in calculation time, thereby improving the prediction accuracy of the prediction model constructed in the prediction model construction step.

[0011] The atomic cluster may be a four-atom cluster where X = 3 and Y = 1. In the analysis step, by assuming a four-atom cluster consisting of three host atoms and one guest atom as the two-element alloy, it is possible to further reduce the calculation scale while preventing a decrease in analytical accuracy.

[0012] Preferably, the two-element selection step selects a plurality of two-element alloys, and the method further includes a verification step of verifying the activation energies of the selected plurality of two-element alloys. By providing the verification step of verifying the activation energies of the two-element alloys in this manner, the appropriateness of the selected two-element alloys can be verified, and when selecting a plurality of two-element alloys, a combination of elements with good properties can be more reliably selected.

[0013] The reaction rate of the chemical reaction may be determined by adsorption reaction. The catalyst searching method functions particularly well when the reaction rate of the chemical reaction is determined by adsorption reaction.

[0014] The chemical reaction may be a gas reduction reaction, and the catalyst searching method functions particularly well for gas reduction reactions.

[0015] The gas may be nitrogen oxide (NOx). By using the catalyst search method, it is possible to efficiently search for a catalyst that decomposes nitrogen oxide (NOx) contained in gas emitted from an automobile engine for which an alternative catalyst is needed, for example.

[0016] Here, the term "two-element alloy" includes not only alloys composed of only two elements, but also alloys containing other elements as unavoidable impurities, or other elements added to the alloy to a degree that does not impair the properties of the two-element alloy, for example, for the purpose of stabilizing the alloy. The mass content of the other elements is 0.05 times or less, preferably 0.03 times or less, the sum of the masses of the two elements constituting the two-element alloy. [Effects of the Invention]

[0017] As described above, the catalyst searching method of the present invention can efficiently search for combinations of elements with good properties. [Brief explanation of the drawings]

[0018] [Figure 1] FIG. 1 is a flow chart showing a method for searching for a catalyst according to one embodiment of the present invention. [Figure 2] FIG. 2 is a schematic diagram for explaining the atomic cluster used in the analysis step of FIG. [Figure 3] FIG. 3 is a graph showing the regression characteristics in the examples. [Figure 4] FIG. 4 is a schematic diagram showing the configuration of a test device used for evaluating catalytic reactivity in the examples. [Figure 5] FIG. 5 is a graph showing the NO decomposition rate of the catalyst in the examples. DETAILED DESCRIPTION OF THE INVENTION

[0019] Hereinafter, a method for searching for a catalyst according to one embodiment of the present invention will be described with reference to the drawings.

[0020] A method for searching for a catalyst according to one embodiment of the present invention is a method for searching for a catalyst made of a two-element alloy that reduces the activation energy for a specific chemical reaction.

[0021] <Chemical Reaction> The above-mentioned chemical reaction is not particularly limited as long as it is a chemical reaction in which the catalyst functions effectively, and it can be applied to various phenomena such as oxidation-reduction reactions and corrosion reactions.

[0022] It is preferable that the reaction rate of the chemical reaction is adsorption-reaction-rate-limited. The catalyst search method functions particularly well when the reaction rate of the chemical reaction is adsorption-reaction-rate-limited. Conversely, it is preferable that the reaction rate of the chemical reaction is not transport-rate-limited.

[0023] The chemical reaction may be a gas reduction reaction. The catalyst search method works particularly well for gas reduction reactions. In particular, the gas may be nitrogen oxide (NOx). Taking nitrogen monoxide (NO) as a typical example, NO is said to be decomposed through the following process: (1) Adsorption of NO onto the catalyst surface (2) NO → N + O (3) NO + N → NO (4) NO → N + O (5) N+N→N2 In other words, (1) NO is adsorbed onto the catalyst surface and (2) decomposed into N and O atoms. The decomposed N atoms are reduced to N either by (3) reacting with NO to form (4) NO or by (5) reacting with each other. In this reaction, the rate-determining step is the adsorption reaction (2), so the effect of the catalyst is easily apparent. Therefore, by using this catalyst search method, it is possible to efficiently search for catalysts that decompose nitrogen oxides (NOx) contained in gases emitted from automobile engines, for which alternative catalysts are needed, for example.

[0024] As shown in FIG. 1, the catalyst search method includes an analysis step S1, a prediction model construction step S2, an inverse analysis step S3, a two-element selection step S4, and a verification step S5.

[0025] <Analysis process> In the analysis step S1, the activation energy of each of a plurality of two-element alloys is analyzed by first-principles calculation.

[0026] A binary alloy is an alloy consisting of two elements, and is determined by specifying the two elements. There are a huge number of ways to select these two elements. In the analysis step S1, several binary alloys are selected from these and the activation energy of each is analyzed.

[0027] The lower limit of the number of binary alloys to be analyzed is preferably 100, more preferably 150. On the other hand, the upper limit of the number of binary alloys to be analyzed is preferably 500, more preferably 300. If the number of binary alloys to be analyzed is less than the above lower limit, sufficient prediction accuracy may not be ensured in the prediction model construction step S2 described below. Conversely, if the number of binary alloys to be analyzed exceeds the above upper limit, the effect of improving prediction accuracy relative to the analysis time tends to saturate.

[0028] The two elements should be selected so that they include a variety of elements, and in particular so that the ionization potential is not biased. In other words, it is desirable to select elements with a moderate combination of low ionization potentials, high ionization potentials, or low and high ionization potentials, and to include a wide variety of elements.

[0029] The first-principles calculation is a known calculation method, and for example, the method described in Non-Patent Document 1 can be used.

[0030] While analysis is typically performed using atomic clusters consisting of several hundred atoms on a support, the catalyst search method assumes an atomic cluster consisting of X host atoms and Y guest atoms (X and Y are natural numbers satisfying the condition 10≧X≧Y) as a binary alloy. The inventors have found that it is particularly effective to assume a four-atom cluster 1 (tetrahedral structure) consisting of three host atoms 11 and one guest atom 12 arranged on a support X, as shown in Figure 2. In the catalyst search method, highly accurate activation energy values ​​are not necessarily required in the analysis step S1. In the prediction model construction step S2, a method with sufficient accuracy to maintain correlation with basic material properties is practical. The inventors discovered that a cluster with as few as four atoms is practical and established a four-atom model for analysis. Furthermore, by assuming a four-atom cluster 1 in the analysis step S1, the calculation scale can be significantly reduced. This makes it possible to increase the number of types of two-element alloys analyzed in the analysis step S1 while suppressing an increase in calculation time, thereby improving the prediction accuracy of the prediction model constructed in the prediction model construction step S2.

[0031] When performing first-principles calculations using the four-atom cluster 1, for example, electronic state calculations based on DFT (Density Functional Theory) are performed, and the activation energy can be estimated from the initial and final states using the Bell-Evans-Volanyi rule.

[0032] When a four-atom cluster 1 is used, the analytical results may differ depending on which of the two elements is used as the host atom 11. The results obtained by interchanging the host atom 11 and the guest atom 12 are treated as independent results.

[0033] <Prediction model construction process> In the prediction model construction step S2, a prediction model is constructed in which the basic material property values ​​of the two elements constituting each of the two-element alloys are used as explanatory variables and the activation energy analyzed in the analysis step is used as a response variable.

[0034] The basic material properties preferably include both the physical properties of each element as a single atom and as a crystal. Examples of the former include group, period, ionization potential, and electronegativity, while examples of the latter include lattice constant, enthalpy of fusion, density, and Wigner-Seitz cell. Among these, it is preferable to include electronegativity, ionization potential, enthalpy of fusion, and density.

[0035] It is preferable to construct a regression model as the above-mentioned prediction model. Examples of the regression model include elastic net, random forest, support vector, and artificial neural network (ANN). Among them, ANN, which has good regression characteristics, is preferable. In other words, it is preferable to construct the above-mentioned prediction model based on machine learning.

[0036] <Back analysis process> In the inverse analysis step S3, the basic material property values ​​that minimize the activation energy are calculated using the prediction model.

[0037] Genetic algorithms, simulated annealing, Bayesian optimization, and other methods can be used to search for fundamental material properties that minimize activation energy, with Bayesian optimization being the most preferred. In particular, when a regression model is constructed using machine learning, the contents of the model itself are a black box. Bayesian optimization is a search method well suited to such black box optimization.

[0038] In addition, even among basic material properties, if there is no or only a small effect on the activation energy, the value is not specified and any value may be used.

[0039] <2 element selection process> In the two-element selection step S4, two real elements are selected based on the basic material property values ​​obtained in the inverse analysis step S3.

[0040] In this two-element selection step S4, a similarity evaluation is performed to select two elements similar to the basic material property values ​​obtained in the inverse analysis step S3, i.e., a specific two-element alloy. Examples of the similarity evaluation include a method using Euclidean distance and cosine similarity.

[0041] The basic material property values ​​of the two elements selected in this two-element selection step S4 do not necessarily match the basic material property values ​​obtained in the inverse analysis step S3. Also, it is possible that an alloy cannot be made with the two selected elements. For this reason, multiple two-element alloys are selected in the two-element selection step S4.

[0042] The lower limit of the number of binary alloys to be selected is preferably 10, more preferably 15. On the other hand, the upper limit of the number of binary alloys to be selected is preferably 50, more preferably 35. If the number of binary alloys to be selected is less than the lower limit, there is a risk that no candidate binary alloys will exist after the verification step S5. Conversely, if the number of binary alloys to be selected exceeds the upper limit, there is a risk that the verification step S5 will take too long.

[0043] Furthermore, when calculation is performed in the analysis step S1 assuming a four-atom cluster, it is preferable to select three to seven types of host atoms 11 and three to seven types of guest atoms 12 for the selected host atoms 11. The number of types of guest atoms 12 may be different for each selected host atom 11.

[0044] In the verification step S5, the activation energies of the selected two-element alloys are verified. By further providing the verification step S5 for verifying the activation energies of the two-element alloys, the validity of the selected two-element alloys can be verified, and when selecting two-element alloys, a combination of elements with good properties can be more reliably selected.

[0045] In the verification step S5, in addition to a method of measuring the activation energy itself, a method of using other parameters that serve as indicators of the activation energy may be adopted. For example, in the case of NO, verification can be performed using the NO decomposition rate.

[0046] In the verification step S5, it is first examined whether the selected two-element alloy can actually be produced. Alloys that cannot actually be produced are excluded from the selection at this stage. This examination may be carried out from a chemical standpoint, or may be judged by actually producing the alloy.

[0047] To determine which binary alloys can be produced, it is advisable to actually produce binary alloys and measure the activation energy of the target chemical reaction. By doing so, it is possible to reliably select the most suitable binary alloy from among multiple binary alloys.

[0048] The activation energy may be calculated by first-principles calculation. Calculation by calculation can reduce the time required to select a two-element alloy compared to actual measurement. The calculation may be performed using a four-atom cluster 1, or a conventional analytical method using an atomic cluster consisting of several hundred atoms on a support X may be used. When a four-atom cluster 1 is used, selection can be performed in a short time. On the other hand, when a conventional analytical method is used, the calculation accuracy is high, so more accurate selection is possible.

[0049] <Advantages> In this catalyst search method, the activation energy of each of a plurality of binary alloys is analyzed using first-principles calculations, and then a predictive model is constructed based on the obtained results, with the activation energy as the objective variable. In this catalyst search method, the predictive model is used to perform inverse analysis to calculate the fundamental material properties that minimize the activation energy, and two real elements that are close to the fundamental material properties that minimize the activation energy are selected, thereby making it possible to efficiently search for combinations of elements with good properties.

[0050] [Other embodiments] The present invention is not limited to the above-described embodiment.

[0051] In the above embodiment, the verification step is described, but the verification step can be omitted. The two-element alloy selected in the two-element selection step can be used as a catalyst without going through the verification step.

[0052] In the above embodiment, a case where multiple two-element alloys are selected in the two-element selection step has been described, but the present invention also contemplates a case where only one two-element alloy is selected. In this case, the verification step may be omitted. On the other hand, if the verification step is performed, it is advisable to verify whether the activation energy of the selected two-element alloy is equivalent to the predicted desired value. [Example]

[0053] The present invention will be described in more detail below with reference to examples, but the present invention is not limited to these examples.

[0054] The effect of the present invention was demonstrated using a direct reduction catalyst for nitric oxide (NO).

[0055] Approximately 260 combinations of 43 binary alloys (1,849 combinations of binary alloys) from groups 3 to 7 of the periodic table (Mg to Nd) were selected. Combinations that could not stably adsorb NO were excluded.

[0056] (Analysis process) Focusing on the adsorption of NO onto catalyst particles and the dissociation of NO, the activation energies of the above 260 combinations were analyzed. Specifically, electronic state calculations based on DFT (Density Functional Theory) were performed, and the activation energies were estimated from the initial and final states using the Bell-Evans-Volanyi law. The (111) surface of zirconia ZrO2 was used as the support.

[0057] (Predictive model construction process) A regression model (prediction model) was constructed using ANN to predict the activation energy for the reduction of NO molecules when the two-element alloy is used as a catalyst, given the basic material properties of each element. The basic material properties used were electronegativity, ionization potential, enthalpy of fusion, and density.

[0058] To verify the validity of the prediction model, the analyzed data was divided into training data and validation data in a 7:3 ratio, and tolerance verification was performed on the training data using the k-fold method. The results are shown in Figure 3. Both the training data and validation data were analyzed using R 2 The value exceeded 0.9, demonstrating excellent properties.

[0059] (Back analysis process) Next, the basic material properties that minimize the activation energy were calculated using the above prediction model. The inverse analysis employed the Bayesian optimization method. As a result, it was found that elements with high electron donating properties could be candidates.

[0060] (2 element selection process) As an example, when one of the atoms is Nb, the activation energies are sorted out and the results show that NbW has the lowest activation energy, NbSb has the highest, and NbGe is somewhere in between. The smaller the activation energy, the more effective the catalyst.

[0061] (Verification process) We actually produced catalyst prototypes using Nb-based alloys that were candidates in the above two element selection process, and measured the NO decomposition rate.

[0062] First, activated alumina (spherical diameter 2 mm, average pore diameter 48 angstroms) was used as a support, and an aqueous solution of Nb compound was impregnated and supported, and then dried at 115°C. This Nb-supported activated alumina was impregnated and supported with an aqueous solution of Sb compound, and then dried at 115°C. After calcination at 600°C for 3 hours under air flow, it was cooled while air was still flowing. In this way, an NbSb catalyst was produced.

[0063] NbW and NbGe catalysts were prepared in a similar manner.

[0064] The prototype catalyst was evaluated using a test apparatus 2 shown in Fig. 4. The test apparatus 2 includes an NO cylinder 21, an N2 cylinder 22, a reaction tube 23, and a gas analyzer 24.

[0065] The catalyst 25 is filled between a pair of wire meshes 23a provided in the reaction tube 23. The reaction tube 23 is provided with a thermocouple 23b capable of measuring the temperature of the catalyst 25.

[0066] A pressure reducing valve 26 for adjusting the supply pressure of the gas and a mass flow controller 27 for adjusting the flow rate are provided downstream of the NO cylinder 21 and the N cylinder 22, respectively. The NO gas and N gas, the pressure and flow rate of which have been adjusted by the pressure reducing valve 26 and the mass flow controller 27, are mixed and supplied to the reaction tube 23. In this test, the N gas was heated to 600°C, and then the N gas was adjusted to have an NO concentration of 1000 ppm and supplied to the reaction tube 23.

[0067] The concentration of NO gas that passed through the catalyst was measured by the gas analyzer 24. The results are shown in Figure 5. It can be seen that the NO decomposition rate was NbW>NbGe>NbSb as predicted. [Industrial Applicability]

[0068] The catalyst search method of the present invention makes it possible to efficiently search for combinations of elements with good properties. [Explanation of symbols]

[0069] 1 4-atom cluster 11 Host atoms 12 guest atoms 2. Test equipment 21 NO Cylinder 22 N2 cylinder 23 Reaction tube 23a Wire mesh 23b Thermocouple 24 Gas analyzer 25 Catalyst 26 Pressure reducing valve 27 Mass flow controller X carrier

Claims

1. A method for searching for a catalyst consisting of a two-element alloy that reduces the activation energy for a chemical reaction in which the catalyst functions effectively, comprising: an analysis step of analyzing the activation energy of each of a plurality of two-element alloys by first-principles calculation; a prediction model construction step of constructing a prediction model using basic material property values ​​of the two elements constituting each of the two-element alloys as explanatory variables and the activation energy analyzed in the analysis step as a response variable; an inverse analysis step of calculating basic material properties that minimize the activation energy using the prediction model; a two-element selection step of selecting two elements similar to the basic material property values ​​from 43 elements in groups 3 to 7 of the periodic table by similarity evaluation using Euclidean distance and cosine similarity based on the basic material property values ​​obtained in the inverse analysis step; Equipped with the chemical reaction is a reduction reaction of a gas, which includes adsorption and dissociation of a gas on the catalyst, and the reaction rate of the chemical reaction is determined by the adsorption reaction rate; The method for searching for a catalyst wherein the gas is nitrogen oxide (NOx).

2. 2. The catalyst searching method according to claim 1, wherein in the analyzing step, an atomic cluster consisting of X host atoms and Y guest atoms (X and Y are natural numbers satisfying the condition 10≧X≧Y) is assumed as the two-element alloy.

3. 3. The method for searching for a catalyst according to claim 2, wherein the atomic cluster is a four-atom cluster in which X=3 and Y=1.

4. In the two-element selection step, a plurality of two-element alloys are selected, 4. The method for searching for a catalyst according to claim 1, further comprising a verification step of verifying the activation energies of the selected plurality of binary alloys.

Images