The MLIP package: moment tensor potentials with MPI and active learning

Novikov, Ivan S.; Gubaev, Konstantin; Podryabinkin, Evgeny V.; Shapeev, Alexander V.

doi:10.1088/2632-2153/abc9fe

Cited by 300 publications

(304 citation statements)

References 74 publications

Supporting

Mentioning

273

Contrasting

Unclassified

Order By: Relevance

“…MTPs include parameters that are optimized by trying to reproduce the results in training datasets. [ 38 ] We next discuss the creation of the ab initio training set for the passive fitting of MLIPs. To study the mechanical properties of pristine phases, AIMD calculations were performed over the rectangular supercells with only 48 atoms.…”

Section: Resultsmentioning

confidence: 99%

“…MTPs with 329 parameters for pristine graphene and borophene, and 449 parameters for heterostructures were trained using the MLIP package. [ 38 ] Phonon dispersions were obtained by density functional perturbation theory simulations over 6 × 6 × 1 supercells with a 3 × 3 × 1 k‐point grid using the PHONOPY code. [ 53 ] Phonon dispersions were also calculated using the MTPs and PHONOPY as explained in the earlier study.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

First‐Principles Multiscale Modeling of Mechanical Properties in Graphene/Borophene Heterostructures Empowered by Machine‐Learning Interatomic Potentials

et al. 2021

View full text Add to dashboard Cite

Density functional theory calculations are robust tools to explore the mechanical properties of pristine structures at their ground state but become exceedingly expensive for large systems at finite temperatures. Classical molecular dynamics (CMD) simulations offer the possibility to study larger systems at elevated temperatures, but they require accurate interatomic potentials. Herein the authors propose the concept of first‐principles multiscale modeling of mechanical properties, where ab initio level of accuracy is hierarchically bridged to explore the mechanical/failure response of macroscopic systems. It is demonstrated that machine‐learning interatomic potentials (MLIPs) fitted to ab initio datasets play a pivotal role in achieving this goal. To practically illustrate this novel possibility, the mechanical/failure response of graphene/borophene coplanar heterostructures is examined. It is shown that MLIPs conveniently outperform popular CMD models for graphene and borophene and they can evaluate the mechanical properties of pristine and heterostructure phases at room temperature. Based on the information provided by the MLIP‐based CMD, continuum models of heterostructures using the finite element method can be constructed. The study highlights that MLIPs were the missing block for conducting first‐principles multiscale modeling, and their employment empowers a straightforward route to bridge ab initio level accuracy and flexibility to explore the mechanical/failure response of nanostructures at continuum scale.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

First‐Principles Multiscale Modeling of Mechanical Properties in Graphene/Borophene Heterostructures Empowered by Machine‐Learning Interatomic Potentials

et al. 2021

View full text Add to dashboard Cite

show abstract

“…A number of publicly available frameworks for the construction and application of MLPs have been released over the last years, including GAP (Bartók et al, 2010), SNAP (Thompson et al, 2015), aenet (Artrith and Urban, 2016), AMP (Khorshidi and Peterson, 2016), ANI-1 (Smith et al, 2017), N2P2 (Singraber et al, 2019), and MLIP (Novikov et al, 2021).…”

Section: Potentialsmentioning

confidence: 99%

Accelerated Atomistic Modeling of Solid-State Battery Materials With Machine Learning

et al. 2021

View full text Add to dashboard Cite

Materials for solid-state batteries often exhibit complex chemical compositions, defects, and disorder, making both experimental characterization and direct modeling with first principles methods challenging. Machine learning (ML) has proven versatile for accelerating or circumventing first-principles calculations, thereby facilitating the modeling of materials properties that are otherwise hard to access. ML potentials trained on accurate first principles data enable computationally efficient linear-scaling atomistic simulations with an accuracy close to the reference method. ML-based property-prediction and inverse design techniques are powerful for the computational search for new materials. Here, we give an overview of recent methodological advancements of ML techniques for atomic-scale modeling and materials design. We review applications to materials for solid-state batteries, including electrodes, solid electrolytes, coatings, and the complex interfaces involved.

show abstract

“…1. We use a physics-based ML model designed for materials, coded in the MLIP package [19] to build an interatomic potential for the GO system.…”

Section: Moment Tensor Potentials For Gomentioning

confidence: 99%

“…The MLIP code is based on moment tensor potentials (MTP) [19,20]. In this machine learning approach, the quantum mechanical energy of a structure (E QM ) is approximated as a sum of interatomic potentials (V ) dependent on the atomic positions and species of the neighbor atoms (n)…”

Section: Moment Tensor Potentials For Gomentioning

confidence: 99%

Studies of Shape Memory Graphene Nanostructures via Integration of Physics-based Modelling and Machine Learning

León¹,

Melnik²

2021

9th Edition of the International Conference on Computational Methods for Coupled Problems in Science and Engineering

View full text Add to dashboard Cite

In this contribution, we study the properties of new promising graphenebased materials with shape memory effects. While traditional shape memory alloys have been extensively studied, it is a challenge to preserve shape memory properties at the nanoscale. As a result, new materials have been explored, among which graphene oxide (GO) crystals with ordered epoxy groups where a recoverable strain of 14.5% has already been reported. We use such nanoscale GO structures as a benchmark example for our studies here. MBTR and SOAP representations are employed in a general-purpose ML model to analyze the effect of long-range interactions in GO. Finally, a physics-based ML model allows us to build interatomic potentials for 2D and 1D systems. The model predicts quantum mechanical effects due to the electronic confinement in narrow nanoribbons and shows the evolution of the local minimal energies associated with two-phase states.

show abstract

The MLIP package: moment tensor potentials with MPI and active learning

Cited by 300 publications

References 74 publications

First‐Principles Multiscale Modeling of Mechanical Properties in Graphene/Borophene Heterostructures Empowered by Machine‐Learning Interatomic Potentials

First‐Principles Multiscale Modeling of Mechanical Properties in Graphene/Borophene Heterostructures Empowered by Machine‐Learning Interatomic Potentials

Accelerated Atomistic Modeling of Solid-State Battery Materials With Machine Learning

Studies of Shape Memory Graphene Nanostructures via Integration of Physics-based Modelling and Machine Learning

Contact Info

Product

Resources

About