Performant implementation of the atomic cluster expansion (PACE) and application to copper and silicon

Lysogorskiy, Yury; Oord, Cas van der; Bochkarev, Anton S.; Menon, Sarath; Rinaldi, M.; Hammerschmidt, Thomas; Mrovec, Matous; Thompson, Aidan P.; Csänyi, Gábor; Ortner, Christoph; Drautz, Ralf

doi:10.1038/s41524-021-00559-9

Cited by 141 publications

(136 citation statements)

References 59 publications

Supporting

Mentioning

134

Contrasting

Order By: Relevance

“…Gaussian Process regression or Neural Network based) or other more recent local-density featurisation approaches (e.g. Atomic Cluster Expansion based [38,43,44] approaches) could enable models which retain the same target accuracy but demand for a lesser number of training structures. Second, co-regionalised [45] or transfer learning [46] approaches may further benefit the making of more efficient and accurate models when tackling the study of LiPS adsorption on a substrates with many and diverse SACs transition metals.…”

Section: Discussionmentioning

confidence: 99%

“…We utilise the elements in the averaged p power specture as the features in kernel ridge regression (KRR). SOAP power spectrum coefficients, [32][33][34][35] and localdensity expansion coefficients more in general, [36][37][38][39] have been largely successful features in kernel-based and linear machine learning models for structure classification and energy regression…”

Section: Soap Representationmentioning

confidence: 99%

See 1 more Smart Citation

Accelerating the theoretical study of Li-polysulphide adsorption on single-atom catalysts via machine learning approaches

Andritsos,

Rossi

2021

Preprint

View full text Add to dashboard Cite

Unlocking the design of Li-S batteries where no shuttle effects appears, and thus their energy storage capacity does not diminish over time, would enable the manufacturing of energy storage devices more performant than the current Li-ion commercial ones. Computational screening of Li-polysulphide (LiPS) adsorption on single-atom catalyst (SAC) substrates is of great aid to the design of lithium-sulphur batteries which are robust against the LiPS shuttling from the cathode to the anode and the electrolyte. To aid this process, we develop a machine learning protocol to accelerate the systematic mapping of dominant local minima found with DFT calculations, and, in turn, fast screen LiPS adsorption properties on SACs. We first validate the approach by probing the potential energy surface for Li-polysulphides adsorbed on graphene decorated with a Fe SAC bound to four nitrogen atoms. We identify minima whose binding energy is better or on par with the one previously reported in the literature. We then move to analyse the adsorption trends on Zn-N4-CSAC and observe similar adsorption strength and behaviour with the Fe-N4-C SAC, highlighting the good predictive power of our protocol.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Soap Representationmentioning

confidence: 99%

Accelerating the theoretical study of Li-polysulphide adsorption on single-atom catalysts via machine learning approaches

Andritsos,

Rossi

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The procedure of generating successively higher correlation order invariants is also at the core of the atomic cluster expansion (ACE) [43,45] scheme which uses linear regression to expand the energy of a collection of atoms in a basis of such invariants. Recent results [58,59] show very good performance both in terms of accuracy and computational efficiency.…”

Section: Rotation Invariantsmentioning

confidence: 91%

Through the eyes of a descriptor: Constructing complete, invertible descriptions of atomic environments

Uhrin

2021

Phys. Rev. B

View full text Add to dashboard Cite

In this work we apply methods for describing three-dimensional images to the problem of encoding atomic environments in a way that is invariant to rotations, translations, and permutations of the atoms and, crucially, can be decoded back into the original environment modulo global orientation without the need for training a model. From the point of view of decoding, the descriptor is optimally complete and can be extended to arbitrary order, allowing for a systematic convergence of the fidelity of the description. In experiments on molecules ranging from 3 to 29 atoms in size, we demonstrate that positions can be decoded with a 97% success rate and positions plus species with a 70% rate of success, rising to 95% if a second fingerprint is used. In all cases, consistent recovery is observed for molecules with 17 or fewer atoms. Additionally, we evaluate the descriptor's performance in predicting the energies and forces of bulk Ni, Cu, Li, Mo, Si, and Ge by means of a neural network model trained on DFT data. When comparing to six machine learning interaction potential methods that use various descriptors and regression schemes, our descriptor is found to be competitive, in several cases outperforming well established methods. The combined ability to both decode and make property predictions from a representation that does not need to be learned lays the foundations for a novel way of building generative models that are tasked with solving the inverse problem of predicting atomic arrangements that are statistically likely to have certain desired properties.

show abstract

“…Both can therefore be parameterised by the ACE model. Here, we closely follow the procedures introduced by Dusson et al 27 , Drautz 35 , Lysogorskiy et al 38 .…”

Section: B Parameterisationmentioning

confidence: 99%

Equivariant analytical mapping of first principles Hamiltonians to accurate and transferable materials models

Zhang¹,

Onat²,

Dusson³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

We propose a data-driven scheme to construct predictive models of Hamiltonian and overlap matrices in atomic orbital representation from ab initio data as a function of local atomic and bond environments. The scheme goes beyond conventional tight binding descriptions as it represents the ab initio model to full order, rather than in two-centre or three-centre approximations. We achieve this by introducing an extension to the Atomic Cluster Expansion (ACE) descriptor that represents intraatomic onsite and interatomic offsite blocks of Hamiltonian and overlap matrices that transform equivariantly with respect to the full rotation group in 3 dimensions. The approach produces equivariant analytical maps from first principles data to linear models for the Hamiltonian and overlap matrices. Through an application to FCC and BCC aluminium, we demonstrate that it is possible to train models from a handful of Hamiltonian and overlap matrices computed with density functional theory using the FHIaims code, and apply them to produce accurate predictions for the band structure and density of states in both phases, as well as along the Bain path that connects them.

show abstract

Performant implementation of the atomic cluster expansion (PACE) and application to copper and silicon

Cited by 141 publications

References 59 publications

Accelerating the theoretical study of Li-polysulphide adsorption on single-atom catalysts via machine learning approaches

Accelerating the theoretical study of Li-polysulphide adsorption on single-atom catalysts via machine learning approaches

Through the eyes of a descriptor: Constructing complete, invertible descriptions of atomic environments

Equivariant analytical mapping of first principles Hamiltonians to accurate and transferable materials models

Contact Info

Product

Resources

About