Gaussian Process Regression for Materials and Molecules

Chemical Reviews, Journal Year: 2021, Volume and Issue: 121(16), P. 10073 - 10141

Published: Aug. 16, 2021

We provide an introduction to Gaussian process regression (GPR) machine-learning methods in computational materials science and chemistry. The focus of the present review is on atomistic properties: particular, construction interatomic potentials, or force fields, Approximation Potential (GAP) framework; beyond this, we also discuss fitting arbitrary scalar, vectorial, tensorial quantities. Methodological aspects reference data generation, representation, regression, as well question how a data-driven model may be validated, are reviewed critically discussed. A survey applications variety research questions chemistry illustrates rapid growth field. vision outlined for development methodology years come.

Language: Английский

---

Machine Learning Interatomic Potentials as Emerging Tools for Materials Science

Advanced Materials, Journal Year: 2019, Volume and Issue: 31(46)

Published: Sept. 5, 2019

Abstract Atomic‐scale modeling and understanding of materials have made remarkable progress, but they are still fundamentally limited by the large computational cost explicit electronic‐structure methods such as density‐functional theory. This Progress Report shows how machine learning (ML) is currently enabling a new degree realism in modeling: “learning” data, ML‐based interatomic potentials give access to atomistic simulations that reach similar accuracy levels orders magnitude faster. A brief introduction tools given, then, applications some select problems science highlighted: phase‐change for memory devices; nanoparticle catalysts; carbon‐based electrodes chemical sensing, supercapacitors, batteries. It hoped present work will inspire development wider use diverse areas research.

Language: Английский

---

Physics-Inspired Structural Representations for Molecules and Materials

Chemical Reviews, Journal Year: 2021, Volume and Issue: 121(16), P. 9759 - 9815

Published: July 26, 2021

The first step in the construction of a regression model or data-driven analysis, aiming to predict elucidate relationship between atomic-scale structure matter and its properties, involves transforming Cartesian coordinates atoms into suitable representation. development representations has played, continues play, central role success machine-learning methods for chemistry materials science. This review summarizes current understanding nature characteristics most commonly used structural chemical descriptions atomistic structures, highlighting deep underlying connections different frameworks ideas that lead computationally efficient universally applicable models. It emphasizes link their physical chemistry, mathematical description, provides examples recent applications diverse set science problems, outlines open questions promising research directions field.

Language: Английский

---

Quantum Chemistry in the Age of Machine Learning

The Journal of Physical Chemistry Letters, Journal Year: 2020, Volume and Issue: 11(6), P. 2336 - 2347

Published: March 3, 2020

As the quantum chemistry (QC) community embraces machine learning (ML), number of new methods and applications based on combination QC ML is surging. In this Perspective, a view current state affairs in exciting research field offered, challenges using are described, potential future developments outlined. Specifically, examples how used to improve accuracy accelerate chemical shown. Generalization classification existing techniques provided ease navigation sea literature guide researchers entering field. The emphasis Perspective supervised learning.

Language: Английский

---

Machine learning for interatomic potential models

The Journal of Chemical Physics, Journal Year: 2020, Volume and Issue: 152(5)

Published: Feb. 5, 2020

The use of supervised machine learning to develop fast and accurate interatomic potential models is transforming molecular materials research by greatly accelerating atomic-scale simulations with little loss accuracy. Three years ago, Jörg Behler published a perspective in this journal providing an overview some the leading methods field. In perspective, we provide updated discussion recent developments, emerging trends, promising areas for future We include three approaches developing machine-learned that have not been extensively discussed existing reviews: moment tensor potentials, message-passing networks, symbolic regression.

Language: Английский

---

Applying Classical, Ab Initio, and Machine-Learning Molecular Dynamics Simulations to the Liquid Electrolyte for Rechargeable Batteries

Chemical Reviews, Journal Year: 2022, Volume and Issue: 122(12), P. 10970 - 11021

Published: May 16, 2022

Rechargeable batteries have become indispensable implements in our daily life and are considered a promising technology to construct sustainable energy systems the future. The liquid electrolyte is one of most important parts battery extremely critical stabilizing electrode–electrolyte interfaces constructing safe long-life-span batteries. Tremendous efforts been devoted developing new solvents, salts, additives, recipes, where molecular dynamics (MD) simulations play an increasingly role exploring structures, physicochemical properties such as ionic conductivity, interfacial reaction mechanisms. This review affords overview applying MD study electrolytes for rechargeable First, fundamentals recent theoretical progress three-class summarized, including classical, ab initio, machine-learning (section 2). Next, application exploration electrolytes, probing bulk structures 3), deriving macroscopic conductivity dielectric constant 4), revealing mechanisms 5), sequentially presented. Finally, general conclusion insightful perspective on current challenges future directions provided. Machine-learning technologies highlighted figure out these challenging issues facing research promote rational design advanced next-generation

Language: Английский

---

FCHL revisited: Faster and more accurate quantum machine learning

The Journal of Chemical Physics, Journal Year: 2020, Volume and Issue: 152(4)

Published: Jan. 27, 2020

We introduce the FCHL19 representation for atomic environments in molecules or condensed-phase systems. Machine learning models based on are able to yield predictions of forces and energies query compounds with chemical accuracy scale milliseconds. is a revision our previous work [F. A. Faber et al., J. Chem. Phys. 148, 241717 (2018)] where discretized individual features rigorously optimized using Monte Carlo optimization. Combined Gaussian kernel function that incorporates elemental screening, reached energy QM7b QM9 datasets after training minutes hours, respectively. The model also shows good performance non-bonded interactions condensed phase set water clusters mean absolute error (MAE) binding less than 0.1 kcal/mol/molecule 3200 samples. For force MD17 dataset, similarly displays state-of-the-art regressor process regression. When revised combined operator quantum machine regressor, can be predicted only few milliseconds per atom. presented herein fast lightweight enough use general chemistry problems as well molecular dynamics simulations.

Language: Английский

---

Neuroevolution machine learning potentials: Combining high accuracy and low cost in atomistic simulations and application to heat transport

Physical review. B./Physical review. B, Journal Year: 2021, Volume and Issue: 104(10)

Published: Sept. 20, 2021

We develop a neuroevolution-potential (NEP) framework for generating neural network-based machine-learning potentials. They are trained using an evolutionary strategy performing large-scale molecular dynamics (MD) simulations. A descriptor of the atomic environment is constructed based on Chebyshev and Legendre polynomials. The method implemented in graphic processing units within open-source gpumd package, which can attain computational speed over ${10}^{7}$ atom-step per second one Nvidia Tesla V100. Furthermore, per-atom heat current available NEP, paves way efficient accurate MD simulations transport materials with strong phonon anharmonicity or spatial disorder, usually cannot be accurately treated either traditional empirical potentials perturbative methods.

Language: Английский

---

GPUMD: A package for constructing accurate machine-learned potentials and performing highly efficient atomistic simulations

The Journal of Chemical Physics, Journal Year: 2022, Volume and Issue: 157(11)

Published: Aug. 24, 2022

We present our latest advancements of machine-learned potentials (MLPs) based on the neuroevolution potential (NEP) framework introduced in [Fan et al., Phys. Rev. B 104, 104309 (2021)] and their implementation open-source package GPUMD. increase accuracy NEP models both by improving radial functions atomic-environment descriptor using a linear combination Chebyshev basis extending angular with some four-body five-body contributions as atomic cluster expansion approach. also detail efficient approach graphics processing units well workflow for construction models, we demonstrate application large-scale atomistic simulations. By comparing to state-of-the-art MLPs, show that not only achieves above-average but is far more computationally efficient. These results GPUMD promising tool solving challenging problems requiring highly accurate, To enable MLPs minimal training set, propose an active-learning scheme latent space pre-trained model. Finally, introduce three separate Python packages, GPYUMD, CALORINE, PYNEP, which integration into workflows.

Language: Английский

---

Modeling the formation and growth of atmospheric molecular clusters: A review

Journal of Aerosol Science, Journal Year: 2020, Volume and Issue: 149, P. 105621 - 105621

Published: July 3, 2020

Language: Английский

---