Catalysis in the digital age: Unlocking the power of data with machine learning

Wiley Interdisciplinary Reviews Computational Molecular Science, Journal Year: 2024, Volume and Issue: 14(5)

Published: Sept. 1, 2024

Abstract The design and discovery of new improved catalysts are driving forces for accelerating scientific technological innovations in the fields energy conversion, environmental remediation, chemical industry. Recently, use machine learning (ML) combination with experimental and/or theoretical data has emerged as a powerful tool identifying optimal various applications. This review focuses on how ML algorithms can be used computational catalysis materials science to gain deeper understanding relationships between properties their stability, activity, selectivity. development repositories, mining techniques, tools that navigate structural optimization problems highlighted, leading highly efficient sustainable future. Several data‐driven models commonly research diverse applications reaction prediction discussed. key challenges limitations using presented, which arise from catalyst's intrinsic complex nature. Finally, we conclude by summarizing potential future directions area ML‐guided catalyst development. article is categorized under: Structure Mechanism > Reaction Mechanisms Catalysis Data Science Artificial Intelligence/Machine Learning Electronic Theory Density Functional

Language: Английский

---

Navigating the Evolution of Carbon Nitride Research: Integrating Machine Learning into Conventional Approaches

Physical Chemistry Chemical Physics, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 1, 2025

Carbon nitride research has reached a promising point in today's endeavours with diverse applications including photocatalysis, energy storage, and sensing due to their unique electronic structural properties. Recent advances machine learning (ML) have opened new avenues for exploring optimizing the potential of these materials. This study presents comprehensive review integration ML techniques carbon an introduction CN classifications recent advancements. We discuss methodologies employed, such as supervised learning, unsupervised reinforcement predicting material properties, synthesis conditions, enhancing performance metrics. Key findings indicate that algorithms can significantly reduce experimental trial-and-error, accelerate discovery processes, provide deeper insights into structure-property relationships nitride. The synergistic effect combining traditional approaches is highlighted, showcasing studies where driven models successfully predicted novel compositions enhanced functional Future directions this field are also proposed, emphasizing need high-quality datasets, advanced models, interdisciplinary collaborations fully realize materials next-generation technologies.

Language: Английский

---

Recent developments in the use of machine learning in catalysis: A broad perspective with applications in kinetics

Chemical Engineering Journal, Journal Year: 2025, Volume and Issue: unknown, P. 160872 - 160872

Published: Feb. 1, 2025

Language: Английский

---

Neural Network Potential for Uranium-Niobium Alloy and Molecular Dynamics Study of Its Low-Temperature Aging Behaviors

Acta Physica Sinica, Journal Year: 2025, Volume and Issue: 74(5), P. 056102 - 056102

Published: Jan. 1, 2025

<sec>Uranium-niobium alloys exhibit complex crystal phases and unique mechanical behaviors under various thermodynamic states external loads. However, due to the lack of accurate interatomic potentials, atomic-scale phase dynamical processes in this important alloy are still unclear. In recent years, development machine-learning-based force fields has provided a systematic way generate potentials on large first-principle-based datasets. crucial nuclear material received limited attention from researchers field machine-learning potentials.</sec><sec>In work, based our previous researches neural-network potential training evaluation framework, which we called NNAP (neural-network atomic potential), new neural network is constructed for uranium-niobium system. A combination random structure search active learning algorithms utilized enhance coverage chemical structural space Testing generated demonstrates high generalization performance accuracy. On testing set, mean absolute error energy 5.6 meV/atom 0.095 eV/Å, respectively. Further calculation results parameters, equation state, phonon dispersions coincide well with first-principle or experimental references.</sec><sec>The evolution spinodal decomposition process U-Nb investigated newly trained potential. It shown that atom-swapping hybrid Monte Carlo can be powerful tool understand systems. By using method, decrease segregation observed within 5000 steps, while no significant reduction found after 3-ns MD simulation. Finally, stress-strain curves shear load different initial obtained. Nb precipitation generates strengthened deformation behavior significantly changed, where disorder band emerges path <inline-formula><tex-math id="M1">\begin{document}$ {\mathrm{\gamma }} $\end{document}</tex-math></inline-formula>-phase alloys. Our work lays foundation understanding system.</sec>

Language: Английский

---

Machine Learning Potentials for Heterogeneous Catalysis

ACS Catalysis, Journal Year: 2025, Volume and Issue: 15(3), P. 1616 - 1634

Published: Jan. 15, 2025

The production of many bulk chemicals relies on heterogeneous catalysis. rational design or improvement the required catalysts critically depends insights into underlying mechanisms atomic scale. In recent years, substantial progress has been made in applying advanced experimental techniques to complex catalytic reactions operando, but order achieve a comprehensive understanding, additional information from computer simulations is indispensable cases. particular, ab initio molecular dynamics (AIMD) become an important tool explicitly address atomistic level structure, dynamics, and reactivity interfacial systems, high computational costs limit applications systems consisting at most few hundred atoms for simulation times up tens picoseconds. Rapid advances development modern machine learning potentials (MLP) now offer promising approach bridge this gap, enabling with accuracy small fraction costs. Perspective, we provide overview current state art MLPs relevant catalysis along discussion prospects use science years come.

Language: Английский

---

Computational Approaches for Designing Heterostructured Electrocatalysts

Small Science, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 11, 2025

Electrocatalysts for oxidation and reduction reactions are crucial sustainable energy production carbon reduction. While precious metal catalysts exhibit superior activity, reducing reliance on them is necessary large‐scale applications. To address this, transition metal‐based studied with strategies to enhance catalytic performance. One promising strategy heterostructures, which integrate multiple materials harness synergistic effects. Developing efficient heterostructured electrocatalysts requires understanding their intricate characteristics, poses challenges. in situ operando spectroscopy provides insights, computational science essential capturing reaction mechanisms, analyzing the origins at atomic scale, efficiently exploring innovative heterostructures. Despite growing recognition of science, standardized criteria these systems remain lacking. This review consolidates case studies propose approaches modeling It categorizes heterostructure types into vertical, semivertical, lateral, defines insights minimizing or exploiting strain effects from lattice mismatches. Furthermore, it summarizes analyses stability activity across reactions, including oxygen evolution, hydrogen reduction, dioxide nitrogen urea oxidation. an overview refine designs establish a framework systematic analysis develop electrocatalysts.

Language: Английский

---

Application of Machine Learning Interatomic Potentials in Heterogeneous Catalysis

Published: Jan. 1, 2025

Language: Английский

---

Fine-Tuning Graph Neural Networks via Active Learning: Unlocking the Potential of Graph Neural Networks Trained on Nonaqueous Systems for Aqueous CO₂ Reduction

Journal of Chemical Theory and Computation, Journal Year: 2025, Volume and Issue: unknown

Published: March 14, 2025

Graph neural networks (GNNs) have revolutionized catalysis research with their efficiency and accuracy in modeling complex chemical interactions. However, adapting GNNs trained on nonaqueous data sets to aqueous systems poses notable challenges due intricate water In this study, we proposed an active learning-based fine-tuning approach extend the applicability of environments. The geometry optimization transition state search workflows are designed reduce computational costs while maintaining DFT-level accuracy. Applied CO2 reduction reaction, workflow delivers a 2-3-fold acceleration through relaxed force threshold combined DFT refinement. versatility algorithm was demonstrated key C-C coupling pathways, pinpointing *CO-*COH as most energetically favorable pathway Cu Cu-based Ag, Au, Zn alloys. Brønsted-Evans-Polanyi relationship remains robust under water-induced fluctuations, alloyed metals such Al, Ga, Pd, along Zn, exhibiting comparable that Cu. Additionally, perturbation-based training forces energies extends application ab initio molecular dynamics simulations, enabling efficient dynamical trajectories. This work presents novel approaches models for systems, highlighting GNNs' potential solvated environments laying foundation accelerating predictions catalytic mechanisms realistic conditions.

Language: Английский

---

Prediction and optimization of key factors for catalytic O3 degradation of antibiotics based on Catboost model coupled Bayesian optimisation algorithm

Journal of Water Process Engineering, Journal Year: 2025, Volume and Issue: 72, P. 107481 - 107481

Published: March 15, 2025

Language: Английский

---

The Future of Catalysis: Applying Graph Neural Networks for Intelligent Catalyst Design

Wiley Interdisciplinary Reviews Computational Molecular Science, Journal Year: 2025, Volume and Issue: 15(2)

Published: March 1, 2025

ABSTRACT With the increasing global demand for energy transition and environmental sustainability, catalysts play a vital role in mitigating climate change, as they facilitate over 90% of chemical material conversions. It is important to investigate complex structures properties enhanced performance, which artificial intelligence (AI) methods, especially graph neural networks (GNNs) could be useful. In this article, we explore cutting‐edge applications future potential GNNs intelligent catalyst design. The fundamental theories their practical catalytic simulation inverse design are first reviewed. We analyze critical roles accelerating screening, performance prediction, reaction pathway analysis, mechanism modeling. By leveraging convolution techniques accurately represent molecular structures, integrating symmetry constraints ensure physical consistency, applying generative models efficiently space, these approaches work synergistically enhance efficiency accuracy Furthermore, highlight high‐quality databases crucial catalysis research innovative application thermocatalysis, electrocatalysis, photocatalysis, biocatalysis. end, key directions advancing catalysis: dynamic frameworks real‐time conditions, hierarchical linking atomic details features, multi‐task interpretability mechanisms reveal pathways. believe advancements will significantly broaden science, paving way more efficient, accurate, sustainable methodologies.

Language: Английский

---