An implementation of artificial neural-network potentials for atomistic materials simulations: Performance for TiO2

Artrith, Nongnuch; Urban, Alexander

doi:10.1016/j.commatsci.2015.11.047

Cited by 476 publications

(370 citation statements)

References 86 publications

Supporting

Mentioning

364

Contrasting

Unclassified

Order By: Relevance

“…If training is successful, this makes atomistic simulations close to quantummechanical accuracy accessible but requires less computational effort by many orders of magnitude. Recent implementations use high-dimensional artificial neural networks, [30][31][32] compressed sensing, 33 or Gaussian process regression. 34 Interatomic ML-based potentials have been developed for several prototypical solids [30][31][32][33][34][35][36][37][38][39] and applied, e.g., in studies of phase transitions.…”

Section: -6mentioning

confidence: 99%

“…Recent implementations use high-dimensional artificial neural networks, [30][31][32] compressed sensing, 33 or Gaussian process regression. 34 Interatomic ML-based potentials have been developed for several prototypical solids [30][31][32][33][34][35][36][37][38][39] and applied, e.g., in studies of phase transitions. 40 We mention in passing that ML schemes are currently being developed to estimate other fundamental properties of molecules and solids, including atomization energies, 41 multipolar polarization, 42 band gaps, 43 or NMR parameters.…”

Section: -6mentioning

confidence: 99%

See 1 more Smart Citation

Machine learning based interatomic potential for amorphous carbon

2017

View full text Add to dashboard Cite

We introduce a Gaussian approximation potential (GAP) for atomistic simulations of liquid and amorphous elemental carbon. Based on a machine-learning representation of the density-functional theory (DFT) potential-energy surface, such interatomic potentials enable materials simulations with close-to DFT accuracy but at much lower computational cost. We first determine the maximum accuracy that any finite-range potential can achieve in carbon structures; then, using a novel hierarchical set of two-, three-, and many-body structural descriptors, we construct a GAP model that can indeed reach the target accuracy. The potential yields accurate energetic and structural properties over a wide range of densities; it also correctly captures the structure of the liquid phases, at variance with state-of-the-art empirical potentials. Exemplary applications of the GAP model to surfaces of "diamond-like" tetrahedral amorphous carbon (ta-C) are presented, including an estimate of the amorphous material's surface energy, and simulations of high-temperature surface reconstructions ("graphitization"). The new interatomic potential appears to be promising for realistic and accurate simulations of nanoscale amorphous carbon structures.

show abstract

Section: -6mentioning

confidence: 99%

Section: -6mentioning

confidence: 99%

Machine learning based interatomic potential for amorphous carbon

2017

View full text Add to dashboard Cite

show abstract

“…In fact, while it already provides an excellent platform not only for the development of potentials using "classic" functional forms (EAM, ABOP, MEAM etc. ), it can be extended to include e.g., artificial neural network (ANN) potentials [30,31,6], tight binding models [32], or Gaussian approximation potentials [33]. In this context, we provide a full documentation of the application programming interface (available as part of the Git repository) to enable other researchers to contribute to the development e.g., via new models (potentials) or optimization schemes.…”

Section: Discussionmentioning

confidence: 99%

“…Since the development of interatomic potentials is frequently a tedious and time consuming process, it requires tools that are both efficient and flexible. While various potential development tools have been developed for internal use by research groups, relatively few have been made widely available including e.g., potfit [2], GARFfield [3], MEAMfit [4], the "EAM Alloy Potential Generator" [5], and the aenet package for artificial neural network (ANN) potentials [6]. Several of these codes target specific potential types and/or functional forms; also they can be difficult to extend and/or integrate with other processing pipelines, in particular the popular Python scripting language.…”

Section: Introductionmentioning

confidence: 99%

Atomicrex—a general purpose tool for the construction of atomic interaction models

Stukowski

Fransson

Mock

et al. 2017

Modelling Simul. Mater. Sci. Eng.

View full text Add to dashboard Cite

Abstract. We introduce atomicrex, an open-source code for constructing interatomic potentials as well as more general types of atomic-scale models. Such effective models are required to simulate extended materials structures comprising many thousands of atoms or more, because electronic structure methods become computationally too expensive at this scale. atomicrex covers a wide range of interatomic potential types and fulfills many needs in atomistic model development. As inputs, it supports experimental property values as well as ab initio energies and forces, to which models can be fitted using various optimization algorithms. The open architecture of atomicrex allows it to be used in custom model development scenarios beyond classical interatomic potentials while thanks to its Python interface it can be readily integrated e.g., with electronic structure calculations or machine learning algorithms.arXiv:1701.08864v1 [cond-mat.mtrl-sci] 30 Jan 2017A general purpose tool for the construction of atomic interaction models 2

show abstract

“…Artrith and Alexander Urbanb [5] have worked about Machine learning interpolation of atomic potential energy surfaces enables the nearly automatic construction of highly accurate atomic interaction potentials. Mansouri Iman and Ozbakkaloglu Togay [6] studies the ability of artificial neural network (ANN), adaptive neuro fuzzy inference system (ANFIS), multivariate adaptive regression splines (MARS) and M5 Model Tree (M5Tree) techniques to predict ultimate conditions FRP-confined concrete.…”

Section: Introductionmentioning

confidence: 99%