Superior performance of the machine-learning GAP force field for fullerene structures

Aghajamali, Alireza; Karton, Amir

doi:10.1007/s11224-021-01864-1

Cited by 8 publications

(1 citation statement)

References 59 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Alternatively, the emerging machine learning or deep learning (DL) techniques have provided a solution for rapid prediction of the energies of isomers for a molecule or cluster with precision comparable to DFT. − For instance, using a geom-C60 database with four symmetric cage isomers and 29 unique C–C bonds, Aghajamali and Karton examined the performance of 12 carbon force fields and found that a machine-learning-based Gaussian approximation potential, namely, GAP-20, outperforms the empirical force fields. In addition to binding energies, Calvo et al created a large database of 753,184 infrared spectra of C n clusters ( n = 24, 33, 42, 52, 60) with different shapes (including fullerene-like cages, graphene-like flakes, pretzel-like and branched structures) using density functional-based tight-binding calculations and developed an interpolation scheme to reproduce the spectral features by encoding the structures using appropriate descriptors and selecting them through principal component analysis and Gaussian regression.…”

Section: Introductionmentioning

confidence: 99%

Predicting Binding Energies and Electronic Properties of Boron Nitride Fullerenes Using a Graph Convolutional Network

Sai,

Fu,

Zhao

2023

J. Chem. Inf. Model.

View full text Add to dashboard Cite

As isoelectronic counterparts of carbon fullerenes, medium-sized boron nitride clusters also prefer cage structures composed of even-sized polygons. As the cluster size increases, the number of cage isomers grows rapidly, and determining the ground state structure requires a tremendous amount of DFT calculations. Herein, we develop a graph convolutional network (GCN) that can describe the energy of a (BN) n cage by its topology connection. We define a vertex feature vector on a dual polyhedron by the permutation of the neighbor vertices' degree and aggregate the information on vertices by two graph convolutional layers to learn the local feature of the dual polyhedron. The GCN is trained on (BN) 28 and subsequently tested on (BN) 23 and (BN) 24 data sets, which satisfactorily reproduce the order of isomer energies from DFT calculations. We further employ the trained GCN to predict the ground state structures within the size range of n = 25−32, which agree well with DFT results. Using the same GCN framework, we also successfully trained the highest-occupied or lowest-unoccupied orbital energies of (BN) 28 isomers. The present graph convolutional network establishes a direct mapping between the topological connection and the energetic or electronic properties of a cage-like cluster or molecule.

show abstract