Automated Algorithms for Quantum-Level Accuracy in Atomistic Simulations: LDRD Final Report.

Thompson, Aidan P.; Schultz, Peter A.; Moore, Stan Gerald; Swiler, Laura Painton; Stephens, J. A.; Trott, Christian Robert; Foiles, Stephen M.; Tucker, Garritt J.

doi:10.2172/1158668

Cited by 5 publications

(4 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…features) that are inherently rotation and permutation invariant. 7,8 Such descriptors may be employed in, e.g., linear regression 9 or neural network models. [10][11][12][13][14] Recently, sophisticated descriptors have been proposed based on an expansion of invariant polynomials 15 and, inspired by convolutional neural networks, on a cascade of multiscale wavelet transformations.…”

Section: Introductionmentioning

confidence: 99%

Learning molecular energies using localized graph kernels

Ferré

Haut

Barros

2017

The Journal of Chemical Physics

View full text Add to dashboard Cite

Recent machine learning methods make it possible to model potential energy of atomic configurations with chemical-level accuracy (as calculated from ab initio calculations) and at speeds suitable for molecular dynamics simulation. Best performance is achieved when the known physical constraints are encoded in the machine learning models. For example, the atomic energy is invariant under global translations and rotations; it is also invariant to permutations of same-species atoms. Although simple to state, these symmetries are complicated to encode into machine learning algorithms. In this paper, we present a machine learning approach based on graph theory that naturally incorporates translation, rotation, and permutation symmetries. Specifically, we use a random walk graph kernel to measure the similarity of two adjacency matrices, each of which represents a local atomic environment. This Graph Approximated Energy (GRAPE) approach is flexible and admits many possible extensions. We benchmark a simple version of GRAPE by predicting atomization energies on a standard dataset of organic molecules.

show abstract

Section: Introductionmentioning

confidence: 99%

Learning molecular energies using localized graph kernels

Ferré

Haut

Barros

2017

The Journal of Chemical Physics

View full text Add to dashboard Cite

show abstract

“…For each set of candidate hyperparameters or group weights proposed by DAKOTA, linear regression was used to solve for the SNAP coefficients using FitSNAP. 40 With each new candidate potential, LAMMPS is used 38,41 to relax a full set of InP defect configurations as well as the bulk InP zinc blende structure. For each relaxation, the configuration was first annealed at 10 K for 0.1 ps before performing the minimization wherein the volume of the cell was also allowed to relax.…”

Section: In the Multicomponent Version Of The Atomic Cluster Expansio...mentioning

confidence: 99%

“…Referring back to the workflow shown in Figure , one pass through the optimization loop proceeds as follows. For each set of candidate hyperparameters or group weights proposed by DAKOTA, linear regression was used to solve for the SNAP coefficients using FitSNAP . With each new candidate potential, LAMMPS is used , to relax a full set of InP defect configurations as well as the bulk InP zinc blende structure.…”

Section: Computational Detailsmentioning

confidence: 99%

Explicit Multielement Extension of the Spectral Neighbor Analysis Potential for Chemically Complex Systems

2020

View full text Add to dashboard Cite

A natural extension of the descriptors used in the Spectral Neighbor Analysis Potential (SNAP) method is derived to treat atomic interactions in chemically complex systems. Atomic environment descriptors within SNAP are obtained from a basis function expansion of the weighted density of neighboring atoms. This new formulation instead partitions the neighbor density into partial densities for each chemical element, thus leading to explicit multielement descriptors. For N elem chemical elements, the number of descriptors increases as N ( ) elem 3 , while the computational cost of the force calculation as implemented in LAMMPS is limited to N ( ) elem 2 and the favorable linear scaling in the number of atoms is retained. We demonstrate these chemically aware descriptors by producing an interatomic potential for indium phosphide capable of capturing high-energy defects that result from radiation damage cascades. This new explicit multielement SNAP method reproduces the relaxed defect formation energies with substantially greater accuracy than weighted-density SNAP, while retaining accurate representation of the bulk indium phosphide properties.

show abstract

“…For each set of candidate hyperparameters or group weights proposed by DAKOTA, linear regression was used to solve for the SNAP coefficients using FitSNAP. 40 With each new candidate potential, LAMMPS is used 38,41 to relax a full set of InP defect configurations as well as the bulk InP zincblende structure. For each relaxation, the configuration was first annealed at 10 K for 0.1 ps before performing the minimization wherein the volume of the cell was also allowed to relax.…”

Section: Fitting Proceduresmentioning

confidence: 99%

Explicit Multi-element Extension of the Spectral Neighbor Analysis Potential for Chemically Complex Systems

Cusentino,

Wood,

Thompson

2020

Preprint

View full text Add to dashboard Cite

A natural extension of the descriptors used in the Spectral Neighbor Analysis Potential (SNAP) method is derived to treat atomic interactions in chemically complex systems. Atomic environment descriptors within SNAP are obtained from a basis function expansion of the weighted density of neighboring atoms. This new formulation instead partitions the neighbor density into partial densities for each chemical element, thus leading to explicit multi-element descriptors. For N elem chemical elements, the number of descriptors increases as O(N 3 elem ), while the computational cost of the force calculation as implemented in LAMMPS is limited to O(N 2 elem ) and the favorable linear scaling in the number of atoms is retained. We demonstrate these chemically aware descriptors by producing an interatomic potential for indium phosphide capable of capturing high-energy defects that result from radiation damage cascades. This new explicit multi-element SNAP method reproduces the relaxed defect formation energies with substantially greater accuracy than weighted-density SNAP, while retaining accurate representation of the bulk indium phosphide properties.

show abstract

Automated Algorithms for Quantum-Level Accuracy in Atomistic Simulations: LDRD Final Report.

Cited by 5 publications

References 24 publications

Learning molecular energies using localized graph kernels

Learning molecular energies using localized graph kernels

Explicit Multielement Extension of the Spectral Neighbor Analysis Potential for Chemically Complex Systems

Explicit Multi-element Extension of the Spectral Neighbor Analysis Potential for Chemically Complex Systems

Contact Info

Product

Resources

About