A catalyst high-throughput screening system and method based on reaction machine learning potential

Abstract

The present application relates to the technical field of catalyst high-throughput screening, and discloses a catalyst high-throughput screening system and method based on reaction machine learning potential, which comprises: a catalyst information acquisition module; a reaction process calculation module; a reaction feature analysis module; a catalytic performance judgment module configured to judge the promotion degree of a candidate catalyst to a target reaction path and a competitive reaction path according to the energy barrier change, the active site change and the reaction path information, and obtain a catalytic performance judgment result; and a screening module configured to screen a plurality of candidate catalysts according to the catalytic performance judgment result, and determine a target catalyst. The present application realizes dynamic modeling and efficient calculation of the whole process of catalytic reaction by introducing the reaction machine learning potential.

Description

Technical Field

[0001] This invention relates to the field of high-throughput screening technology for catalysts, and more specifically, to a high-throughput screening system and method for catalysts based on reaction machine learning potential. Background Technology

[0002] In the development of electrocatalytic carbon dioxide reduction catalysts, it is necessary to screen catalysts with low reaction energy barriers and high selectivity from a large number of metal alloys, single-atom materials, and composite catalytic materials. Existing high-throughput screening systems typically use first-principles calculations to evaluate candidate catalysts one by one. Because surface models need to be built for different catalysts and adsorption processes and reaction pathways need to be calculated, the computational resources consumed are large, making it difficult to meet the needs of large-scale material screening. At the same time, actual catalytic reactions involve dynamic changes in atomic structure, and existing screening methods based on static parameters cannot accurately describe the energy evolution during the reaction process, thus affecting the accuracy of catalyst performance prediction.

[0003] Therefore, it is necessary to design a high-throughput catalyst screening system and method based on reaction machine learning potential to solve the problems existing in the current technology. Summary of the Invention

[0004] In view of this, the present invention proposes a high-throughput catalyst screening system and method based on reaction machine learning potential, aiming to solve the problems of high computational cost, difficulty in capturing dynamic reaction processes and inaccurate selectivity prediction in the prior art.

[0005] In one aspect, the present invention proposes a high-throughput catalyst screening system based on reaction machine learning potential, comprising: The catalyst information acquisition module is configured to acquire the structural information of multiple candidate catalysts and determine the reaction system corresponding to each candidate catalyst. The reaction process calculation module is configured to calculate the reaction process corresponding to the candidate catalyst based on the reaction machine learning potential, and obtain energy change information, atomic structure change information and reaction path information under different reaction states during the reaction process; The reaction characteristic analysis module is configured to analyze the energy barrier changes corresponding to different reaction steps based on the energy change information, and to determine the changes in the active sites of the catalyst based on the atomic structure change information. The catalytic performance judgment module is configured to determine the degree to which the candidate catalyst promotes the target reaction pathway and competing reaction pathway based on the energy barrier change, active site change and reaction pathway information, and obtain the catalytic performance judgment result. The screening module is configured to screen multiple candidate catalysts based on the catalytic performance judgment results to determine the target catalyst.

[0006] Furthermore, when calculating the reaction process corresponding to the candidate catalyst based on the reaction machine learning potential, the calculation includes: Based on the structural information of the candidate catalyst and the reaction system, the corresponding initial reaction state is constructed; Based on the reaction machine learning potential, the initial state of the reaction is calculated to obtain the interatomic interaction information and system energy information corresponding to different reaction stages; Based on the interatomic interaction information corresponding to different reaction stages, determine the changes in the atomic structure of the catalyst and reactants during the reaction process; Based on the energy information and atomic structure changes of the system corresponding to different reaction stages, the reaction state transition relationship is determined, and the energy change information, atomic structure change information, and reaction path information corresponding to the candidate catalyst are obtained.

[0007] Furthermore, when determining the reaction state transition relationships based on the system's energy information and atomic structure changes corresponding to different reaction stages, the following steps are included: Based on the energy information of the system, the energy states corresponding to different reaction stages are correlated and analyzed to determine the energy change relationship between adjacent reaction states. Based on the changes in atomic structure, the atomic arrangement and bonding relationships corresponding to different reaction stages are matched to determine the relationship between atomic migration and chemical bond changes during the reaction process; Based on the energy change relationship and the atomic migration and chemical bond change relationship, the transition process between different reaction states is identified, and the reaction state transition node is determined. Based on the connection relationships between the reaction state transition nodes, the reaction state transition relationships corresponding to the candidate catalysts are constructed.

[0008] Furthermore, when analyzing the energy barrier changes corresponding to different reaction steps based on the energy change information, the analysis includes: Based on the aforementioned reaction state transition relationship, the energy difference between adjacent reaction states is determined, and the energy change corresponding to different reaction steps is obtained. Based on the energy changes corresponding to different reaction steps, determine the peak energy state and the minimum energy state in each reaction step; Calculate the reaction energy barrier corresponding to different reaction steps based on the difference between the energy peak state and the energy minimum state; Correlation analysis is performed based on the reaction energy barriers corresponding to each reaction step to determine the energy barrier changes of the candidate catalyst during the reaction process.

[0009] Furthermore, when determining the changes in the active sites of the catalyst based on the atomic structure change information, the following steps are included: Based on the atomic structure change information corresponding to different reaction states, information on the coordination number change, bond length change, and local structure change of the catalyst surface atoms is extracted. Based on the information on coordination number changes, bond length changes, and local structure changes, the structurally active regions of the catalyst surface atoms during the reaction process are determined; Based on the atomic change characteristics corresponding to the active regions of the structure, the active site activity states during reactant adsorption, activation, and intermediate state formation processes are analyzed. Based on the active site states corresponding to different reaction stages, the changes in the catalyst's active sites during the reaction process are determined.

[0010] Furthermore, when determining the structurally active regions of catalyst surface atoms during the reaction process based on the information regarding coordination number changes, bond length changes, and local structure changes, the following steps are included: The degree of change in the coordination environment around the surface atoms is determined by the change in the coordination number of catalyst surface atoms under different reaction states. Based on the changes in atomic bond lengths on the catalyst surface under different reaction states, the changes in the intensity of interatomic interactions are determined. Based on the degree of change in the coordination environment and the change in the intensity of interatomic interactions, the structural change characteristics corresponding to different atomic regions on the catalyst surface are obtained; Based on the structural change characteristics, the atomic regions on the catalyst surface are divided to determine the structurally active regions on the catalyst surface during the reaction process.

[0011] Furthermore, when dividing the atomic regions on the catalyst surface based on the aforementioned structural change characteristics to determine the structurally active regions on the catalyst surface during the reaction, the process includes: Based on the structural change characteristics, structural change parameters corresponding to atomic regions on the surface of different catalysts are extracted; The degree of structural change in different atomic regions is calculated based on the structural change parameters to obtain the change characteristic values ​​corresponding to each atomic region; Based on the relationship between the change characteristic value and the change threshold corresponding to each atomic region, the atomic regions on the catalyst surface are screened to determine the target atomic regions; Based on the distribution of the target atomic regions on the catalyst surface, the structurally active regions on the catalyst surface during the reaction are determined.

[0012] Furthermore, when determining the degree to which the candidate catalyst promotes the target reaction pathway and competing reaction pathways based on the energy barrier changes, active site changes, and reaction pathway information, and obtaining the catalytic performance assessment result, the following steps are included: Based on the energy barrier changes, determine the reaction energy barrier change characteristics corresponding to the target reaction path and the competing reaction path; Based on the reaction path information, determine the reaction state transition processes corresponding to the target reaction path and competing reaction paths, and obtain the reaction process characteristics corresponding to each reaction path; The degree of participation of active sites in different reaction pathways is determined based on the changes in the active sites. Based on the characteristics of the reaction barrier change, the characteristics of the reaction process, and the degree of participation of active sites, the degree of promotion of the candidate catalyst on the target reaction pathway and the competing reaction pathway is calculated, and the catalytic performance is judged based on the difference between the two.

[0013] Furthermore, based on the catalytic performance assessment results, when screening multiple candidate catalysts to determine the target catalyst, the process includes: Based on the catalytic performance judgment results corresponding to each candidate catalyst, the target reaction pathway promotion characteristics and the competitive reaction pathway inhibition characteristics are extracted. Based on the target reaction pathway promoting characteristics and the competing reaction pathway inhibiting characteristics, reaction selectivity evaluation parameters corresponding to candidate catalysts are constructed. Based on the reaction selectivity evaluation parameters corresponding to multiple candidate catalysts, the selectivity differences of different candidate catalysts for the target reaction pathway are determined; The target catalyst is determined based on the selectivity differences combined with the corresponding energy barrier change characteristics and active site change characteristics.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention, by introducing reaction machine learning potentials, achieves dynamic modeling and efficient computation of the entire catalytic reaction process, significantly reducing computational costs in high-throughput screening while improving the accuracy of catalyst performance prediction. Specifically, the system can not only capture atomic-scale structural evolution information during the reaction process but also comprehensively evaluate the selectivity and activity of candidate catalysts by combining energy changes, dynamic behavior of active sites, and multi-path competition mechanisms. Compared to traditional screening methods that rely on static adsorption energy or single energy barrier indices, this invention fully considers the complexity of reaction pathways and the dynamic response of catalyst surface structures, thus more realistically reflecting material performance under actual catalytic conditions. Furthermore, by constructing a structurally active region identification mechanism and reaction selectivity evaluation parameters, this invention provides a clear structure-performance correlation basis for catalyst design, helping to accelerate the discovery and optimization of high-performance catalysts.

[0015] In another aspect, this invention also proposes a high-throughput catalyst screening method based on reaction machine learning potential, comprising the following steps: Obtain structural information of multiple candidate catalysts and determine the corresponding reaction system for each candidate catalyst; The reaction process corresponding to the candidate catalyst is calculated based on the reaction machine learning potential to obtain energy change information, atomic structure change information and reaction path information under different reaction states during the reaction process; The energy barrier changes for different reaction steps are analyzed based on the energy change information, and the changes in the active sites of the catalyst are determined based on the atomic structure change information. Based on the changes in energy barriers, changes in active sites, and reaction pathway information, the degree to which candidate catalysts promote the target reaction pathway and competing reaction pathways is determined, and the catalytic performance is judged. Based on the catalytic performance assessment results, multiple candidate catalysts were screened to determine the target catalyst.

[0016] It is understandable that the above-mentioned catalyst high-throughput screening system and method based on reaction machine learning potential have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A structural block diagram of a high-throughput catalyst screening system based on reaction machine learning potential provided in an embodiment of the present invention; Figure 2 A flowchart illustrating a high-throughput catalyst screening method based on reaction machine learning potential provided in an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] See Figure 1 As shown in some embodiments of this application, this embodiment provides a high-throughput catalyst screening system based on reaction machine learning potential, including: The catalyst information acquisition module is configured to acquire the structural information of multiple candidate catalysts and determine the reaction system corresponding to each candidate catalyst. The reaction process calculation module is configured to calculate the reaction process corresponding to the candidate catalyst based on the reaction machine learning potential, and obtain energy change information, atomic structure change information and reaction path information under different reaction states during the reaction process; The reaction characteristic analysis module is configured to analyze the energy barrier changes corresponding to different reaction steps based on energy change information, and to determine the changes in the active sites of the catalyst based on atomic structure change information. The catalytic performance assessment module is configured to determine the degree to which candidate catalysts promote the target reaction pathway and competing reaction pathway based on changes in energy barriers, changes in active sites, and reaction pathway information, and to obtain the catalytic performance assessment results. The screening module is configured to screen multiple candidate catalysts based on the catalytic performance judgment results to determine the target catalyst.

[0021] In this embodiment, the structural information includes elemental composition information, crystal structure information, surface structure information, and active site information; the reaction system includes reactant information, reaction condition information, reaction intermediate state information, and target product information.

[0022] It is understood that the high-throughput catalyst screening system based on reaction machine learning potentials provided in this embodiment can effectively achieve high-precision and high-efficiency simulation and evaluation of complex catalytic reaction processes. By embedding reaction machine learning potentials into the high-throughput calculation process, this system avoids the need for repeated, costly first-principles calculations for each candidate catalyst in traditional methods, thereby significantly shortening the screening cycle while ensuring atomic-scale accuracy. In practical applications, users can input the target reaction (such as the electroreduction of CO2 to C2H4) and a library of candidate catalysts, and the system automatically completes the entire process analysis from initial structure construction and reaction pathway exploration to comprehensive performance evaluation.

[0023] Specifically, when calculating the reaction process corresponding to a candidate catalyst based on the reaction machine learning potential, the following are included: Based on the structural information of the candidate catalysts and the reaction system, the corresponding initial reaction states are constructed. Based on the reaction machine learning potential, the state evolution of the initial state of the reaction is calculated to obtain the interatomic interaction information and system energy information corresponding to different reaction stages; Based on the interatomic interaction information corresponding to different reaction stages, determine the changes in the atomic structure of the catalyst and reactants during the reaction process; Based on the system energy information and atomic structure changes corresponding to different reaction stages, the reaction state transition relationships are determined, and the energy change information, atomic structure change information, and reaction path information corresponding to candidate catalysts are obtained.

[0024] Understandably, calculating the state evolution of the initial state of a reaction based on the reaction machine learning potential, and obtaining information on interatomic interactions and system energy at each reaction stage, refers to tracking the atomic configuration of the system over time under the drive of the reaction machine learning potential through molecular dynamics simulation or enhanced sampling methods. It obtains the atomic forces, bonding states, and total energy at each moment in real time, automatically identifies key configurations such as reactant adsorption states, transition states, intermediate states, and product desorption states on the reaction coordinates, and outputs the corresponding energy values ​​and atomic position sequences.

[0025] Understandably, changes in atomic structure include the displacement of atoms on the catalyst surface, the evolution of the adsorption configuration of reactant molecules on the surface, the breaking and formation of chemical bonds, and the resulting local lattice distortion or reconstruction phenomena. It also encompasses changes in the geometric configuration of intermediates during the reaction, dynamic adjustments in the coordination environment, and the cooperative motion characteristics of atomic clusters surrounding the active sites. This structural information is quantified and characterized using high-temporal-resolution trajectory data, providing an atomic-scale structural basis for subsequent identification of reaction state transition nodes and construction of reaction pathways.

[0026] In some embodiments of this application, by combining the interatomic interaction information corresponding to different reaction stages, the changes in the atomic structure of the catalyst and reactants during the reaction can be determined. Specifically, the interatomic interaction information encompasses changes in distance between atoms, changes in interaction strength, changes in coordination relationships, and changes in chemical bonds at each reaction stage. By analyzing the interatomic interaction information during the reaction stage by stage, the structural change patterns of catalyst surface atoms and reactant molecules during adsorption, migration, binding, and dissociation can be clarified, thereby obtaining the changes in atomic positions, atomic arrangement, and local structural changes corresponding to different reaction stages. The changes in atomic structure are used to characterize the dynamic evolution of the catalyst surface structure and reactant structure during the reaction, providing a data foundation for subsequent analysis of changes in catalyst active sites and determination of reaction pathways.

[0027] Specifically, when determining the reaction state transition relationships based on the system's energy information and atomic structure changes corresponding to different reaction stages, the following are included: Based on the system's energy information, correlation analysis is performed on the energy states corresponding to different reaction stages to determine the energy change relationships between adjacent reaction states; By matching the atomic arrangement and bonding relationships corresponding to different reaction stages based on changes in atomic structure, the relationship between atomic migration and chemical bond changes during the reaction process can be determined. Based on the relationships of energy change, atomic migration, and chemical bond change, the transition processes between different reaction states are identified, and the reaction state transition nodes are determined. Based on the connection relationships between reaction state transition nodes, the reaction state transition relationships corresponding to candidate catalysts are constructed.

[0028] Understandably, the specific steps for determining the energy change relationships between adjacent reaction states by performing correlation analysis on the energy states corresponding to different reaction stages based on the system's energy information are as follows: First, the total energy of the system at each sampling time during the reaction process is sorted according to time series or reaction coordinates, and local minimum and maximum points in the energy curve are identified; among them, local minimum points correspond to stable reaction intermediate states or adsorption states, while local maximum points correspond to transition states. Subsequently, by comparing the energy differences between adjacent extreme points, an energy change trend spectrum is established, and cross-validation is performed using abrupt changes in atomic structure (such as sudden bond length changes, coordination number jumps, etc.) to confirm the effective correlation between energy states. On this basis, spurious fluctuations caused by thermal fluctuations or numerical noise are eliminated, and energy transition events with clear structural evolution support are retained, ultimately forming a reliable and physically meaningful energy change relationship between adjacent reaction states.

[0029] Understandably, matching the atomic arrangement and bonding relationships at different reaction stages based on changes in atomic structure to determine the relationship between atomic migration and chemical bond changes during the reaction involves tracing the continuous evolution of atomic coordinates along the reaction trajectory, identifying the spatial migration paths of key atoms (such as active atoms in reactants or coordinating atoms on the catalyst surface), and combining this with bond order analysis, electron density distribution, or bonding criteria (such as Wiberg bond order, ICOHP, etc.) to determine the breaking, formation, or recombination of chemical bonds. Specifically, this involves structurally comparing the atomic configurations between adjacent time steps or reaction coordinate steps, extracting atomic displacement vectors, bond length change rates, and coordination environment transformation matrices to determine which atoms participated in the migration and which chemical bonds underwent substantial changes. Based on this, the spatiotemporal correlation between atomic migration behavior and chemical bond changes is established. For example, determining whether the breaking of a C–O bond is accompanied by the migration of O atoms to specific metal sites on the catalyst surface clarifies the coupling mechanism between atomic rearrangement and bonding evolution during the reaction. This process not only identifies key elementary steps in the reaction pathway but also provides structural-level criteria for subsequently constructing reaction state transition relationships.

[0030] Understandably, based on energy change relationships and relationships between atomic migration and chemical bond changes, the transition processes between different reaction states are identified. The specific steps for determining reaction state transition nodes are as follows: based on the extreme point positions determined in the energy change relationships, combined with the spatiotemporal nodes where abrupt changes occur in atomic migration and chemical bond changes, the state type of each reaction stage is determined: nodes that satisfy local energy minima and correspond to a clearly defined stable atomic configuration are marked as reactant adsorption states, reaction intermediates, or product states; nodes that correspond to local energy maxima and are accompanied by chemical bond breaking / formation are marked as transition states of elementary reactions. Subsequently, according to the order of reaction, all identified state nodes are sequentially numbered, and the state type, energy value, and atomic configuration corresponding to each node are determined, thus completing the determination of all reaction state transition nodes.

[0031] It is understandable that the connection relationships between reaction state transition nodes include the topological connection order and branching structure of the reaction path.

[0032] Specifically, analyzing the energy barrier changes corresponding to different reaction steps based on energy change information includes: Based on the relationship between reaction state transitions, the energy difference between adjacent reaction states is determined, and the energy change corresponding to different reaction steps is obtained. Based on the energy changes corresponding to different reaction steps, determine the peak energy state and the minimum energy state in each reaction step; Calculate the reaction energy barrier corresponding to different reaction steps based on the difference between the energy peak state and the energy minimum state; Correlation analysis is performed based on the reaction energy barriers corresponding to each reaction step to determine the energy barrier changes of the candidate catalyst during the reaction process.

[0033] Understandably, the specific steps for determining the energy barrier changes of candidate catalysts during the reaction process by performing correlation analysis based on the reaction energy barriers corresponding to each reaction step are as follows: Arrange the reaction energy barriers of each elementary step in the order of the reaction path, identify the rate-controlling step (i.e., the step with the highest energy barrier), and analyze the relative strength of this energy barrier compared to corresponding steps in other competing pathways; simultaneously, examine the smoothness and cumulative effect of the energy barrier distribution throughout the target reaction pathway to assess its kinetic feasibility. Based on this, and combining the energy barrier differences of different candidate catalysts in the same reaction step, quantify their influence on the reaction rate trend, providing crucial kinetic basis for subsequent catalytic performance evaluation.

[0034] Specifically, when determining changes in the active sites of a catalyst based on information about changes in atomic structure, this includes: Based on the atomic structure change information corresponding to different reaction states, information on the coordination number change, bond length change and local structure change of the catalyst surface atoms is extracted; Based on information on changes in coordination number, bond length, and local structure, the structurally active regions of atoms on the catalyst surface during the reaction process are determined; Based on the atomic change characteristics corresponding to the structurally active regions, the active site interaction states during reactant adsorption, activation, and intermediate state formation processes are analyzed. Based on the active site status corresponding to different reaction stages, the changes of catalyst active sites during the reaction process are determined.

[0035] Understandably, based on the atomic change characteristics of structurally active regions, the active site functioning state during reactant adsorption, activation, and intermediate state formation is analyzed. This involves identifying surface atomic regions with significantly reduced coordination numbers, bond length stretching, or local geometric reconstruction, and determining them as structurally active regions. The functional role of this region at different reaction stages is analyzed by combining the reactant's adsorption orientation, electron transfer behavior, and intermediate stabilization ability within this region. The active site function types are classified, their dynamic transformation processes are tracked, and the spatiotemporal evolution characteristics of catalyst active sites and their regulatory mechanisms on reaction pathway selectivity are characterized.

[0036] It is understandable that the changes in the active sites of the catalyst during the reaction process include: dynamic increase or decrease in the number of active sites, migration of spatial positions, modulation of electronic structure, and reconstruction of coordination environment.

[0037] Specifically, when determining the structurally active regions of catalyst surface atoms during the reaction process based on information about changes in coordination number, bond length, and local structure, the following are included: The degree of change in the coordination environment around the surface atoms is determined by the change in the coordination number of catalyst surface atoms under different reaction states. Based on the changes in atomic bond lengths on the catalyst surface under different reaction states, the changes in the intensity of interatomic interactions are determined. The structural change characteristics of different atomic regions on the catalyst surface were obtained based on the changes in the degree of coordination environment and the intensity of interatomic interactions. Based on the structural change characteristics, the atomic regions on the catalyst surface are divided to determine the structurally active regions on the catalyst surface during the reaction process.

[0038] Understandably, the specific steps for determining the degree of change in the coordination environment of catalyst surface atoms based on the changes in the coordination number of catalyst surface atoms under different reaction states are as follows: The coordination number of each surface atom in the initial, intermediate, and transition states of the reaction is statistically analyzed, and the change in coordination number relative to the initial state is calculated. If the decrease in the coordination number of a certain atom exceeds a preset threshold (e.g., a decrease of ≥1), its coordination environment is considered significantly weakened, potentially exposing it as a potential active site. Simultaneously, the degree of weakening of interatomic bonding strength is assessed by combining bond length changes; for example, the elongation of metal-metal bonds or metal-adsorbate bonds exceeds a preset proportion. Further, the coordination number change and bond length change are coupled for analysis to identify atomic regions that simultaneously meet the conditions of "significant decrease in coordination number" and "significant bond length stretching," serving as a candidate set of structurally active regions. Based on this, transient fluctuations caused by thermal vibrations or non-chemical interactions are excluded, and regions with persistent structural perturbations that evolve synchronously with the reaction process are retained, ultimately determining the structurally active regions that truly participate in the catalytic process.

[0039] Specifically, when dividing the atomic regions on the catalyst surface based on structural change characteristics to determine the structurally active regions on the catalyst surface during the reaction, this includes: Structural change parameters corresponding to atomic regions on the surface of different catalysts are extracted based on structural change characteristics. The degree of structural change in different atomic regions is calculated based on the structural change parameters to obtain the change characteristic values ​​corresponding to each atomic region; Based on the relationship between the change characteristic value and the change threshold corresponding to each atomic region, the atomic regions on the catalyst surface are screened to determine the target atomic regions; Based on the distribution of target atomic regions on the catalyst surface, the structurally active regions on the catalyst surface during the reaction are determined.

[0040] Understandably, structural change parameters include multidimensional indicators such as coordination number change, relative bond length scaling, local strain tensor modulus, and root mean square atomic displacement. These parameters are integrated into a unified change characteristic value through weighted fusion or principal component analysis to quantify the structural perturbation intensity of each atomic region. The change threshold can be dynamically set according to the characteristics of the reaction system. For example, for strongly exothermic reactions, the threshold can be appropriately increased to filter out non-critical perturbations, while a more sensitive criterion is used for reactions dominated by weak interactions. The selected target atomic regions must continuously exhibit significant structural changes in multiple reaction states and correspond temporally highly with the critical energy barrier steps on the reaction coordinates. The finally determined structurally active regions not only reflect the dynamic reconstruction behavior of the catalyst surface but also provide spatial positioning basis for subsequent catalytic performance evaluation modules, supporting in-depth analysis of the correlation mechanism between the functional state of active sites and reaction pathway selectivity.

[0041] In this embodiment, the preferred value of the change threshold is 0.2 times the standard deviation of the structural change parameter. Alternatively, the optimal threshold can be automatically optimized through cross-validation based on the test dataset to ensure that the threshold can effectively distinguish between substantial structural changes caused by the reaction and minor fluctuations caused by intrinsic thermal vibration. This preserves the true active region while avoiding misjudging non-associated perturbation regions, thereby improving the accuracy and reliability of identifying structurally active regions.

[0042] Specifically, when determining the degree to which candidate catalysts promote the target and competing reaction pathways based on changes in energy barriers, changes in active sites, and reaction pathway information, and obtaining the catalytic performance assessment results, the following are included: Determine the characteristics of reaction energy barrier changes for the target reaction pathway and competing reaction pathways based on the changes in energy barriers. Based on the reaction pathway information, determine the reaction state transition processes corresponding to the target reaction pathway and competing reaction pathways, and obtain the reaction process characteristics corresponding to each reaction pathway; Determine the degree of participation of active sites in different reaction pathways based on changes in active sites; Based on the characteristics of reaction barrier changes, reaction process characteristics, and the degree of participation of active sites, the degree of promotion of candidate catalysts on the target reaction pathway and competing reaction pathways is calculated, and the catalytic performance is judged based on the difference between the two.

[0043] Understandably, the energy barrier change characteristics of the target and competing reaction paths are determined based on the energy barrier changes. Specifically, based on the system energy information under different reaction states calculated by reaction machine learning potential, the energy changes corresponding to each reaction step in the target and competing reaction paths are obtained, and the reaction energy barrier corresponding to each reaction step is determined based on the energy difference between adjacent reaction states. Furthermore, the reaction energy barriers corresponding to different reaction steps in each reaction path are correlated to obtain the energy barrier change trend in the corresponding reaction path. The energy barrier change characteristics include the magnitude of the energy barrier corresponding to the reaction step, the amplitude of the energy barrier change, and the relationship of energy barrier changes between different reaction steps, which are used to characterize the degree of influence of the candidate catalyst on the reaction energy process in different reaction paths.

[0044] Understandably, the reaction state transition processes corresponding to the target reaction path and competing reaction paths are determined based on reaction path information, thereby obtaining the reaction process characteristics of each reaction path. Specifically, based on the reaction state nodes contained in the reaction state transition relationship and the connection relationships between each node, the reaction state sequences corresponding to the target reaction path and competing reaction paths are extracted respectively. Then, based on the sequence, the number of reaction steps, the number of reaction intermediate states, the order of reaction state transitions, and the continuity of the reaction path are determined for each reaction path. Finally, the reaction process characteristics of the corresponding reaction path are obtained, which are used to characterize the process change laws of different reaction paths under the action of candidate catalysts.

[0045] Understandably, the degree of participation of active sites in different reaction pathways can be determined based on changes in active sites. Specifically, by combining information on atomic structure changes at different reaction stages, the positional and structural changes of active sites on the catalyst surface in the target and competing reaction pathways, as well as the changes in their interactions with reactants, are extracted. Then, based on the frequency of occurrence of active sites in different reaction states, the number of reaction steps they participate in, and the degree of structural change in the corresponding regions, the degree of participation of active sites in different reaction pathways is determined. This index is used to characterize the influence of specific regions on the catalyst surface on the advancement of different reaction pathways.

[0046] Understandably, by combining the characteristics of reaction energy barrier changes, reaction process characteristics, and the degree of participation of active sites, the promoting effect of candidate catalysts on the target reaction pathway and competing reaction pathways is calculated separately. The catalytic performance is then judged based on the difference between the two. The steps are as follows: Based on the characteristics of reaction energy barrier changes, the energy barrier change at each step of different pathways is obtained, and the energy barrier promoting parameter is determined according to the degree of energy barrier reduction; the greater the degree of energy barrier reduction, the stronger the promoting effect. Based on the characteristics of the reaction process, the number of reaction state transitions, the continuity of reaction steps, and the formation of intermediate states are obtained for different pathways, and the process promoting parameter is determined. The more continuous the state transition and the more stable the intermediate state, the higher the parameter. Based on the degree of participation of active sites, the number and intensity of participation of active sites are obtained to determine the site promotion parameter. After normalizing the three types of parameters, they are fused according to preset weights to obtain the degree of promotion of the candidate catalyst for different reaction pathways. The calculation formula is: Pi = aEi + bRi + cAi; where Pi represents the degree of promotion of the candidate catalyst for the i-th reaction pathway; Ei is the energy barrier promotion parameter of the pathway; Ri is the process promotion parameter of the pathway; Ai is the site promotion parameter of the pathway; and a, b, and c are the weight coefficients of each parameter.

[0047] By comparing the difference in the degree of promotion between the target reaction pathway and the competing reaction pathway, if the target reaction pathway is promoted more, it indicates that the candidate catalyst is more inclined to promote the target reaction pathway. Finally, the catalytic performance is judged based on the difference.

[0048] Specifically, when selecting a target catalyst by screening multiple candidate catalysts based on catalytic performance evaluation results, the process includes: Based on the catalytic performance evaluation results of each candidate catalyst, the characteristics promoting the target reaction pathway and the characteristics inhibiting the competing reaction pathway are extracted. Based on the promoting characteristics of the target reaction pathway and the inhibiting characteristics of the competing reaction pathway, reaction selectivity evaluation parameters for candidate catalysts are constructed. Based on the reaction selectivity evaluation parameters corresponding to multiple candidate catalysts, the selectivity differences of different candidate catalysts for the target reaction pathway are determined; The target catalyst is determined by combining the selectivity differences with the corresponding energy barrier change characteristics and active site change characteristics.

[0049] It should be noted that the specific steps for determining the target catalyst are as follows: First, based on the catalytic performance assessment results of each candidate catalyst, the promoting characteristics of the target reaction pathway and the inhibiting characteristics of the competing reaction pathway are extracted. Specifically, based on the degree to which the candidate catalysts promote the two types of reaction pathways, the promoting values ​​of the target reaction pathway and the competing reaction pathway are extracted respectively. Among them, the promoting characteristics of the target reaction pathway include the promoting value of the pathway, the degree of reduction of the reaction barrier, and the degree of participation of the active site; the inhibiting characteristics of the competing reaction pathway include the promoting value of the pathway, the change of the reaction barrier, and the effect of the active site.

[0050] Secondly, based on the promoting characteristics of the target reaction pathway and the inhibiting characteristics of the competing reaction pathway, reaction selectivity evaluation parameters for the corresponding candidate catalysts are constructed. Specifically, the selectivity tendency of the candidate catalyst for the target reaction pathway is determined based on the difference in the promoting values ​​of the two pathways. This parameter is used to characterize the overall ability of the candidate catalyst to promote the target reaction and inhibit the competing reaction. In one embodiment, the reaction selectivity evaluation parameter is determined by the difference between the promoting value of the target reaction pathway and the promoting value of the competing reaction pathway; that is, the higher the promoting value of the target reaction pathway and the lower the promoting value of the competing reaction pathway, the larger the corresponding reaction selectivity evaluation parameter.

[0051] Furthermore, based on the reaction selectivity evaluation parameters corresponding to multiple candidate catalysts, the selectivity differences of different candidate catalysts for the target reaction pathway are determined. Specifically, the reaction selectivity evaluation parameters of all candidate catalysts are calculated separately, and the selectivity differences of different candidate catalysts are obtained by comparing the parameter magnitudes; the larger the reaction selectivity evaluation parameter, the stronger the promoting effect of the candidate catalyst on the target reaction pathway, and the stronger its ability to inhibit competing reaction pathways.

[0052] Finally, by combining the selectivity differences with the corresponding energy barrier change characteristics and active site change characteristics, the target catalyst is determined. Specifically, the candidate catalysts are first ranked according to their reaction selectivity evaluation parameters, and the reaction energy barrier change characteristics and active site change characteristics of the top-ranked candidate catalysts are extracted. Then, it is determined whether they meet the energy barrier reduction requirements and active site stable participation requirements corresponding to the target reaction pathway. Finally, the candidate catalyst that meets the above conditions and has the highest reaction selectivity evaluation parameters is determined as the target catalyst.

[0053] See Figure 2 As shown in some embodiments of this application, this embodiment provides a high-throughput screening method for catalysts based on reaction machine learning potential, including the following steps: S100: Obtain structural information of multiple candidate catalysts and determine the reaction system corresponding to each candidate catalyst; S200: Calculate the reaction process corresponding to the candidate catalyst based on the reaction machine learning potential to obtain energy change information, atomic structure change information and reaction path information under different reaction states during the reaction process; S300: Analyze the energy barrier changes corresponding to different reaction steps based on the energy change information, and determine the changes in the active sites of the catalyst based on the atomic structure change information; S400: Based on the changes in energy barriers, changes in active sites, and reaction pathway information, determine the degree to which the candidate catalyst promotes the target reaction pathway and competing reaction pathways, and obtain the catalytic performance judgment result; S500: Based on the catalytic performance judgment results, multiple candidate catalysts are screened to determine the target catalyst.

[0054] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0055] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0056] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0057] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A high-throughput catalyst screening system based on reaction machine learning potential, characterized in that, include: The catalyst information acquisition module is configured to acquire the structural information of multiple candidate catalysts and determine the reaction system corresponding to each candidate catalyst. The reaction process calculation module is configured to calculate the reaction process corresponding to the candidate catalyst based on the reaction machine learning potential, and obtain energy change information, atomic structure change information and reaction path information under different reaction states during the reaction process; The reaction characteristic analysis module is configured to analyze the energy barrier changes corresponding to different reaction steps based on the energy change information, and to determine the changes in the active sites of the catalyst based on the atomic structure change information. The catalytic performance judgment module is configured to determine the degree to which the candidate catalyst promotes the target reaction pathway and competing reaction pathway based on the energy barrier change, active site change and reaction pathway information, and obtain the catalytic performance judgment result. The screening module is configured to screen multiple candidate catalysts based on the catalytic performance judgment results to determine the target catalyst.

2. The high-throughput catalyst screening system based on reaction machine learning potential according to claim 1, characterized in that, When calculating the reaction process corresponding to the candidate catalyst based on the reaction machine learning potential, the calculation includes: Based on the structural information of the candidate catalyst and the reaction system, the corresponding initial reaction state is constructed; Based on the reaction machine learning potential, the initial state of the reaction is calculated to obtain the interatomic interaction information and system energy information corresponding to different reaction stages; Based on the interatomic interaction information corresponding to different reaction stages, determine the changes in the atomic structure of the catalyst and reactants during the reaction process; Based on the energy information and atomic structure changes of the system corresponding to different reaction stages, the reaction state transition relationship is determined, and the energy change information, atomic structure change information, and reaction path information corresponding to the candidate catalyst are obtained.

3. The high-throughput catalyst screening system based on reaction machine learning potential according to claim 2, characterized in that, When determining the reaction state transition relationships based on the system's energy information and atomic structure changes corresponding to different reaction stages, the following are included: Based on the energy information of the system, the energy states corresponding to different reaction stages are correlated and analyzed to determine the energy change relationship between adjacent reaction states. Based on the changes in atomic structure, the atomic arrangement and bonding relationships corresponding to different reaction stages are matched to determine the relationship between atomic migration and chemical bond changes during the reaction process; Based on the energy change relationship and the atomic migration and chemical bond change relationship, the transition process between different reaction states is identified, and the reaction state transition node is determined. Based on the connection relationships between the reaction state transition nodes, the reaction state transition relationships corresponding to the candidate catalysts are constructed.

4. The high-throughput catalyst screening system based on reaction machine learning potential according to claim 3, characterized in that, When analyzing the energy barrier changes corresponding to different reaction steps based on the energy change information, the following is included: Based on the aforementioned reaction state transition relationship, the energy difference between adjacent reaction states is determined, and the energy change corresponding to different reaction steps is obtained. Based on the energy changes corresponding to different reaction steps, determine the peak energy state and the minimum energy state in each reaction step; Calculate the reaction energy barrier corresponding to different reaction steps based on the difference between the energy peak state and the energy minimum state; Correlation analysis is performed based on the reaction energy barriers corresponding to each reaction step to determine the energy barrier changes of the candidate catalyst during the reaction process.

5. The high-throughput catalyst screening system based on reaction machine learning potential according to claim 4, characterized in that, When determining the changes in the active sites of a catalyst based on the atomic structure change information, the following are included: Based on the atomic structure change information corresponding to different reaction states, information on the coordination number change, bond length change, and local structure change of the catalyst surface atoms is extracted. Based on the information on coordination number changes, bond length changes, and local structure changes, the structurally active regions of the catalyst surface atoms during the reaction process are determined; Based on the atomic change characteristics corresponding to the active regions of the structure, the active site activity states during reactant adsorption, activation, and intermediate state formation processes are analyzed. Based on the active site states corresponding to different reaction stages, the changes in the catalyst's active sites during the reaction process are determined.

6. The high-throughput catalyst screening system based on reaction machine learning potential according to claim 5, characterized in that, When determining the structurally active regions of catalyst surface atoms during the reaction process based on the information regarding coordination number changes, bond length changes, and local structure changes, the following are included: The degree of change in the coordination environment around the surface atoms is determined by the change in the coordination number of catalyst surface atoms under different reaction states. Based on the changes in atomic bond lengths on the catalyst surface under different reaction states, the changes in the intensity of interatomic interactions are determined. Based on the degree of change in the coordination environment and the change in the intensity of interatomic interactions, the structural change characteristics corresponding to different atomic regions on the catalyst surface are obtained; Based on the structural change characteristics, the atomic regions on the catalyst surface are divided to determine the structurally active regions on the catalyst surface during the reaction process.

7. The high-throughput catalyst screening system based on reaction machine learning potential according to claim 6, characterized in that, When dividing the atomic regions on the catalyst surface based on the structural change characteristics to determine the structurally active regions on the catalyst surface during the reaction, the following methods are included: Based on the structural change characteristics, structural change parameters corresponding to atomic regions on the surface of different catalysts are extracted; The degree of structural change in different atomic regions is calculated based on the structural change parameters to obtain the change characteristic values ​​corresponding to each atomic region; Based on the relationship between the change characteristic value and the change threshold corresponding to each atomic region, the atomic regions on the catalyst surface are screened to determine the target atomic regions; Based on the distribution of the target atomic regions on the catalyst surface, the structurally active regions on the catalyst surface during the reaction are determined.

8. The high-throughput catalyst screening system based on reaction machine learning potential according to claim 7, characterized in that, When determining the degree to which candidate catalysts promote the target and competing reaction pathways based on the energy barrier changes, active site changes, and reaction pathway information, and obtaining the catalytic performance assessment results, the following steps are included: Based on the energy barrier changes, determine the reaction energy barrier change characteristics corresponding to the target reaction path and the competing reaction path; Based on the reaction path information, determine the reaction state transition processes corresponding to the target reaction path and competing reaction paths, and obtain the reaction process characteristics corresponding to each reaction path; The degree of participation of active sites in different reaction pathways is determined based on the changes in the active sites. Based on the characteristics of the reaction barrier change, the characteristics of the reaction process, and the degree of participation of active sites, the degree of promotion of the candidate catalyst on the target reaction pathway and the competing reaction pathway is calculated, and the catalytic performance is judged based on the difference between the two.

9. The high-throughput catalyst screening system based on reaction machine learning potential according to claim 8, characterized in that, Based on the catalytic performance assessment results, when screening multiple candidate catalysts to determine the target catalyst, the process includes: Based on the catalytic performance judgment results corresponding to each candidate catalyst, the target reaction pathway promotion characteristics and the competitive reaction pathway inhibition characteristics are extracted. Based on the target reaction pathway promoting characteristics and the competing reaction pathway inhibiting characteristics, reaction selectivity evaluation parameters corresponding to candidate catalysts are constructed. Based on the reaction selectivity evaluation parameters corresponding to multiple candidate catalysts, the selectivity differences of different candidate catalysts for the target reaction pathway are determined; The target catalyst is determined based on the selectivity differences combined with the corresponding energy barrier change characteristics and active site change characteristics.

10. A high-throughput screening method for catalysts based on reaction machine learning potential, applied to the high-throughput screening system for catalysts based on reaction machine learning potential as described in any one of claims 1-9, characterized in that, include: Obtain structural information of multiple candidate catalysts and determine the corresponding reaction system for each candidate catalyst; The reaction process corresponding to the candidate catalyst is calculated based on the reaction machine learning potential to obtain energy change information, atomic structure change information and reaction path information under different reaction states during the reaction process; The energy barrier changes for different reaction steps are analyzed based on the energy change information, and the changes in the active sites of the catalyst are determined based on the atomic structure change information. Based on the changes in energy barriers, changes in active sites, and reaction pathway information, the degree to which candidate catalysts promote the target reaction pathway and competing reaction pathways is determined, and the catalytic performance is judged. Based on the catalytic performance assessment results, multiple candidate catalysts were screened to determine the target catalyst.

Images