CN-117198415-B - Method for high-throughput screening MOFs (metal-oxide-semiconductor field effect transistors) catalytic carbon dioxide cycloaddition catalyst

CN117198415BCN 117198415 BCN117198415 BCN 117198415BCN-117198415-B

Abstract

The invention relates to the technical field of chemical MOFs catalysis, and discloses a method for high-throughput screening of MOFs-catalyzed carbon dioxide cycloaddition catalysts. The method solves the technical problem that high-flux material screening is difficult to realize in the field of catalysis in MOFs material design by reasonable feature descriptor selection, reasonable machine learning algorithm selection and evaluation model. Specifically, in the invention, a machine learning model is established by using a machine learning program based on Density Functional Theory (DFT) calculation and a low-cost Monte Carlo (GCMC) simulation to obtain MOFs electronic and structural properties and a descriptor combined with reaction conditions, and a TOF value of a catalyst is selected as a classification standard to establish a high-flux screening model. The high flux in the process is mainly realized in that the calculation cost is reduced through reasonable descriptor selection, so that the method can realize large-scale screening and ensure the calculation accuracy. In addition, the application of the machine learning technology also provides a data analysis scheme for MOFs catalyst design.

Inventors

LI JIANRONG
BAI XUEFENG
ZHANG XIN
XIE YABO

Assignees

北京工业大学

Dates

Publication Date: 20260505
Application Date: 20230913

Claims (6)

1. A method for high-throughput screening of MOFs catalyzed carbon dioxide cycloaddition catalyst is characterized in that a machine learning model is established by using a machine learning program based on DFT calculation and a low-cost GCMC simulation to obtain MOFs structure and electronic properties and a descriptor combined with reaction conditions, TOF values of the catalyst are selected as classification standards, a high-throughput screening model is established, in addition, the application of the machine learning technology also provides a data analysis scheme for MOFs catalyst design, and the method further comprises the following method steps: s1, preparing data; s2, calculating metal charges; s3, calculating the specific surface area and the pore volume; S4, selecting a machine learning algorithm; s5, evaluating an algorithm; s6, analyzing model data; In the step S2, further: the metal charges are calculated by a PACMOF program obtained from Github, an algorithm model is calculated by adopting a DDEC model based on a random forest algorithm provided by the program, no additional model training is carried out, and all MOFs structure files are subjected to pretreatment to remove solvents and guest molecules in the structure before calculation; in the step S3, further: the specific surface area and the pore volume are calculated by carrying out Monte Carlo calculation for 50000 times of megarule system circulation times through RASPA codes, helium is used as a probe molecule to calculate the specific surface area, the pore volume is obtained by multiplying the porosity of helium by specific volume, the force field is selected as a UFF general force field, and the cut-off value is set to be 12.8.
2. The method for high throughput screening of MOFs catalyzed carbon dioxide cycloaddition catalysts of claim 1 wherein S1 is further: the data set extracted from the paper is used as raw data for machine learning, structural features and reaction condition information are selected as feature descriptors, atomic partial charges of metal centers are calculated using a machine learning model PACMOF based on Density Functional Theory (DFT) calculation, lewis acidity for describing metal open sites are calculated using monte carlo (GCMC) simulation of RASPA to calculate specific surface area and porosity of the structure, three-dimensional characteristics for describing MOF as feature inputs.
3. The method for high throughput screening of MOFs catalyzed carbon dioxide cycloaddition catalysts of claim 1 wherein in S4, further: As a common expression of catalytic activity, turnover frequency (TOF) was used as the target for model training, 16 mainstream machine learning classification models were used, using the SCIKIT-Learn, catboost, xgboost and LGBM packages of Python, specifically, the scheme used KNN、LR、DT、RF、SVM、SGD、QDA、NN、BernoulliNB(BNB)、MultinomialNB(MNB)、CatBoost(Cat)、LightGBM(LGBM)、XGBoost(XGB)、GBDT、ET、AdaBoost(Ada) models, all model training was divided into training and test sets at a ratio of 80/20%, random seeds were set to 1, super parameters were searched by grid, and triple cross validation was performed on the training sets.
4. The method for high throughput screening of MOFs catalyzed carbon dioxide cycloaddition catalysts of claim 1 wherein in S5, further: The performance of each machine learning algorithm is evaluated by calculating the accuracy, precision, recall and f1 score of the model, resulting in a preferred machine learning model with accuracy up to 97%.
5. The method for high throughput screening of MOFs catalyzed carbon dioxide cycloaddition catalysts of claim 4 wherein the computational model is validated using Topological Based Crystal Constructor (ToBaCCo3.0) codes to construct MOFs that include, but are not limited to, 12415 virtual MOF structures and 100 actual MOF structures obtained from Cambridge university database, and the model successfully screens MOFs materials with excellent catalytic activity.
6. The method for high throughput screening of MOFs catalyzed carbon dioxide cycloaddition catalyst of claim 1 wherein in S6, further data analysis is performed on a machine learning algorithm using SHAPLEY ADDITIVE exPlanations and PARTIAL DEPENDENCE Plot to obtain empirical guidance with chemical significance.

Description

Method for high-throughput screening MOFs (metal-oxide-semiconductor field effect transistors) catalytic carbon dioxide cycloaddition catalyst Technical Field The invention relates to the technical field of chemical MOFs catalysis, in particular to a method for screening MOFs catalytic carbon dioxide cycloaddition catalysts in a high throughput manner. Background MOFs, known as Metal-organic framework materials (metals-OrganicFrameworks), are a class of materials consisting of Metal ions and organic ligands. MOFs material has highly ordered pore canal structure and surface area, and is characterized by adjustable pore size and pore structure, so that the MOFs material has wide application in the field of catalysis. Due to structural characteristics, MOFs can implement catalytic processes of a variety of different mechanisms of action including electrocatalytic, photocatalytic and traditional thermocatalytic fields where carbon dioxide cycloaddition reactions are located. Among them, MOFs have made their high throughput screening a major problem in the development of this material due to the theoretically infinite possibilities and microscopic high tunability. Other applications of MOFs, such as adsorption separation, can also be modeled by low cost monte carlo (GCMC) calculations [ CN 115458073A ], but because thermocatalysis involves multiple factors of heat transfer, mass transfer, transmission, etc., it is generally only possible to screen using laboratory experimental methods. However, it usually takes several months to study a material, and since MOFs are composed of metal and organic ligands according to a certain topology, there are theoretically infinite possibilities, and the micro-scale regulation of MOFs is complex and has a large probability. The traditional experimental method is difficult to realize screening and developing of materials in a short time. It is less likely to achieve high throughput screening of materials. Theoretical calculation has been used for development of chemical materials as a calculation means of chemical engineering, but the calculation cost is high due to the complex mechanism of the catalytic process due to the large number of atoms in the unit cells of MOFs materials. In addition, by the Density Functional Theory (DFT), the theoretical calculation methods such as Molecular Dynamics (MD) and the like are difficult to obtain data directly related to the performance of the material. In view of the above, the prior art is currently unable to achieve high throughput screening of catalysts for MOFs materials. For this purpose, we take the catalytic reaction of carbon dioxide cycloaddition as an example, and propose a method for high throughput screening of MOFs-catalyzed carbon dioxide cycloaddition catalysts. Disclosure of Invention The invention mainly solves the technical problems in the prior art and provides a method for high-throughput screening of MOFs-catalyzed carbon dioxide cycloaddition catalysts. In order to achieve the above object, the present invention adopts the following technical scheme, and is a method for high throughput screening of MOFs catalyzed carbon dioxide cycloaddition catalysts, a machine learning model is established by using a machine learning program based on DFT calculation and a low-cost GCMC simulation to obtain MOFs structure and electronic properties and a descriptor combined with reaction conditions, a TOF value of the catalyst is selected as a classification standard, a high throughput screening model is established, and in addition, the application of the machine learning technology also provides a data analysis scheme for MOFs catalyst design, and the method further comprises the following steps: s1, preparing data; s2, calculating metal charges; s3, calculating the specific surface area and the pore volume; S4, selecting a machine learning algorithm; s5, evaluating an algorithm; S6, analyzing model data. Preferably, in S1, further: The data set extracted from the recent paper is used as raw data for machine learning, and structural features and reaction condition information are selected as feature descriptors. The reasonable feature descriptors reduce the calculation cost, so that the calculation accuracy is ensured while large-scale screening can be realized. Preferably, in S2, further: the metal charge was calculated using the PACMOF program obtained from Github, the algorithm model was calculated using the DDEC model based on random forest algorithm provided by the program, without additional model training, and all CIF files were treated to remove solvent and guest molecules prior to calculation. Preferably, in S3, further: the specific surface area and the pore volume are calculated by carrying out 50000 times of giant regularization system synthesis Monte Carlo calculation through RASPA codes, helium is used as a probe molecule to calculate the specific surface area, the pore volume is obtained by multiplying the porosity