Machine Learning, Density Functional Theory, and Experiments to Understand the Photocatalytic Reduction of CO₂ on CuPt/TiO₂

The Journal of Physical Chemistry C, Journal Year: 2024, Volume and Issue: 128(34), P. 14247 - 14258

Published: Aug. 15, 2024

The photoconversion of CO2 to hydrocarbons is a sustainable route for its transformation into value-added compounds, which crucial mitigating energy and climate crises. CuPt nanoparticles on TiO2 surfaces have been reported show promising efficiencies. For further progress, mechanistic understanding the catalytic properties these CuPt/TiO2 systems vital. Here, we employ ab initio calculations, machine learning, photocatalysis experiments understand photocatalytic reduction CuPt/TiO2. We explore configurational space CO2@CuPt/TiO2 examine their structures energetics. find that interface plays key role in determining activation and, thus, conversion hydrocarbons. stabilizes *CO other intermediates containing CH groups, thus facilitating higher activity selectivity methane. A bias-corrected machine-learning interatomic potential trained density functional theory data enables efficient exploration numerous configurations using basin-hopping Monte Carlo simulations, greatly accelerating study photocatalyst systems. Our simulations preferentially adsorbs at interface, with C atom bonded Pt site one O occupying an O-vacancy site. also promotes formation *CH *CH2 intermediates. confirmation, synthesize samples various compositions, analyze morphologies compositions scanning electron microscopy energy-dispersive X-ray spectroscopy, measure activity. computational experimental findings qualitatively agree highlight importance design selective

Language: Английский

---

Data-efficient modeling of catalytic reactions via enhanced sampling and on-the-fly learning of machine learning potentials

Published: June 6, 2024

Simulating catalytic reactivity under operative conditions poses a significant challenge due to the dynamic nature of catalysts and high computational cost electronic structure calculations. Machine learning potentials offer promising avenue simulate dynamics at fraction cost, but they require datasets containing all relevant configurations, particularly reactive ones. Here we present scheme construct in data-efficient manner. This is achieved by combining enhanced sampling methods first with Gaussian processes discover transition paths then graph neural networks obtain uniformly accurate description. The necessary configurations are extracted via an active procedure based on local environment uncertainty. We validated our approach studying several reactions related decomposition ammonia iron-cobalt alloy catalysts. Our proved efficient, requiring only ~1,000 DFT calculations per reaction, robust, from different accessible pathways. Using this potential, calculated free energy profiles characterized reaction mechanisms, showing ability provide microscopic insights into complex conditions.

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

---

Exploring the Intricacies of Glycerol Hydrodeoxygenation on Copper Surface: A Comprehensive Investigation with the Aid of Machine Learning Force Field

The Journal of Physical Chemistry C, Journal Year: 2025, Volume and Issue: unknown

Published: March 20, 2025

Language: Английский

---

Efficient exploration of reaction pathways using reaction databases and active learning

The Journal of Chemical Physics, Journal Year: 2025, Volume and Issue: 162(11)

Published: March 21, 2025

The fast and accurate simulation of chemical reactions is a major goal computational chemistry. Recently, the pursuit this has been aided by machine learning interatomic potentials (MLIPs), which provide energies forces at quantum mechanical accuracy but fraction cost reference calculations. Assembling training set relevant configurations key to building MLIP. Here, we demonstrate two approaches reactive MLIPs based on reaction pathway information. One approach exploits datasets containing reactant, product, transition state structures. Using an SN2 dataset, accurately locate pathways geometries up 170 unseen reactions. In another approach, does not depend data availability, present efficient active procedure that yields MLIP converged minimum energy path given only end point structures, avoiding mechanics driven search any stage construction. We in gas phase with small number solvating water molecules, predicting barriers within 20 meV chemistry method. then apply more complex involving nucleophilic aromatic substitution proton transfer, comparing results against ReaxFF force field. Our procedure, addition rapidly finding paths for individual reactions, provides large databases transferable potentials.

Language: Английский

---

Automated Pynta-Based Curriculum for ML-Accelerated Calculation of Transition States

The Journal of Physical Chemistry C, Journal Year: 2025, Volume and Issue: 129(16), P. 7751 - 7761

Published: April 13, 2025

Microkinetic models (MKMs) are widely used within the computational heterogeneous catalysis community to investigate complex reaction mechanisms, rationalize experimental trends, and accelerate rational design of novel catalysts. However, constructing these requires computationally expensive manually tedious density functional theory (DFT) calculations for identifying transition states each elementary MKM. To address challenges, we demonstrate a protocol that uses open-source kinetics workflow tool Pynta automate iterative training reactive machine learning potential (rMLP). Specifically, using silver-catalyzed partial oxidation methanol as prototypical example, first our by an rMLP parallel calculation DFT-quality all 53 reactions, achieving 7× speedup compared DFT-only strategy. Detailed analysis curriculum reveals shortcomings adaptive sampling scheme with single model describe reactions MKM simultaneously. We show limitations can be overcome balanced "reaction class" approach multiple models, describing class similar states. Finally, Pynta-based is also compatible large pretrained foundational models. For fine-tuning top-performing graph neural network trained on OC20 dataset, observe impressive 20× 89% success rate in This work highlights synergistic integrating automated tools advance research.

Language: Английский

---

Machine-Learning-Accelerated DFT Conformal Sampling of Catalytic Processes

Journal of Chemical Theory and Computation, Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 30, 2024

Computational modeling of catalytic processes at gas/solid interfaces plays an increasingly important role in chemistry, enabling accelerated materials and process optimization rational design. However, efficiency, accuracy, thoroughness, throughput must be enhanced to maximize its practical impact. By combining interpolation DFT energetics via highly accurate Machine-Learning Potentials with conformal techniques for building the training database, we present here original approach (that name Conformal Sampling Catalytic Processes, CSCP), accelerate achieve thorough sampling novel systems by exporting existing information on a worked-out case. We use methanol decomposition (of interest field hydrogen production storage) as test reaction. Starting from Pt-based systems, show that after only two iterations active-learning CSCP is able provide reaction energy diagrams set 7 diverse (Pd, Ni, Au, Ag, Cu, Co, Fe) leading DFT-accuracy-level predictions. Cases exhibiting change adsorption sites mechanisms are also successfully reproduced tests path modification. The thus offers itself operative tool fully take advantage accumulated high-throughput processes.

Language: Английский

---

Review of External Field Effects on Electrocatalysis: Machine Learning Guided Design

Advanced Functional Materials, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 10, 2024

Abstract External field‐enhanced electrocatalysis is a novel and promising approach for boosting the efficiency of electrocatalytic reactions, potentially achieving significant enhancement without altering composition structure electrocatalysts. In addition, scaling relations typically lead to similar variations initial‐state transition‐state (TS) energy, which minimally impacts reaction energy barrier. A sophisticated design external field effects shall break these relations. This review provides comprehensive overview current research on effect mechanical, electric, magnetic fields electrocatalysis. It meticulously details mechanisms underlying activity based regulations, spanning from synthesis materials their behavior during process modulation electrolyte environment. Additionally, applications emerging machine learning (ML) technologies in design, including interatomic potentials (MLIPs) simulate large‐scale dynamic chemical processes, data‐driven optimization performance, are briefly reviewed. potential ML conjunction with regulation, envisioning them as effective tools optimizing or reverse designing electrocatalysis, considering both thermodynamic kinetic factors well electrocatalyst surfaces under extreme fields, highlighted.

Language: Английский

---

A Molecular View of Methane Activation on Ni(111) through Enhanced Sampling and Machine Learning

The Journal of Physical Chemistry Letters, Journal Year: 2024, Volume and Issue: 15(39), P. 9852 - 9862

Published: Sept. 19, 2024

A combination of machine learned interatomic potentials (MLIPs) and enhanced sampling simulations is used to investigate the activation methane on a Ni(111) surface. The work entails development iterative refinement MLIPs, initially trained data set constructed via ab initio molecular dynamics simulations, supplemented by adaptive biasing forces, enrich catalytically relevant configurations. Our results reveal that upon incorporation collective variables capture behavior reactant molecule, as well additional frames describe dynamic response catalytic surface, it possible enhance considerably accuracy predicted energies forces. By employing schemes in MLIP, we systematically explore potential energy leading refined MLIP capable predicting density functional theory-level forces replicating key geometric characteristics system. resulting free landscapes at several temperatures provide detailed view thermodynamics activation. Specifically, approaches dissociates process involves interplay CH4 Ni catalyst includes both enthalpic entropic contributions. progression toward transition state moiety increasingly restrained its ability rotate or translate, while stage following characterized notable rise atom interacts with cleaved C-H bond. This leads an increase mobility adsorbed species, feature becomes more pronounced higher temperatures.

Language: Английский

---

Application of Machine Learning Interatomic Potentials in Heterogeneous Catalysis

Published: Jan. 1, 2025

Language: Английский

---