Electron-Passing Neural Networks for Atomic Charge Prediction in Systems with Arbitrary Molecular Charge

Metcalf, Derek P.; Jiang, Andy; Spronk, Steven A.; Cheney, Daniel L.; Sherrill, C. David

doi:10.1021/acs.jcim.0c01071

Cited by 31 publications

(40 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many other approaches have been proposed to express atomic charges by machine learning [42,[57][58][59][60][61][62]. Often, reference charges obtained in electronic structure calculations are used for training.…”

Section: Beyond Locality-long-ranged Mlpsmentioning

confidence: 99%

“…It is interesting to note that message passing networks can form a bridge between second, third and fourth-generation potentials. For instance, apart from the total energy they have been suggested to predict charges [42,58,59], although to date only rarely explicit Coulomb terms are used to include long-range interactions without truncation [42] as needed for a classification as a third or fourth generation potential. Still, by increasing the number of passing steps in principle they allow to describe more and more interactions irrespective of the physical nature including electrostatics between close atoms, like second-generation MLPs based on Eq.…”

Section: Beyond Locality-long-ranged Mlpsmentioning

confidence: 99%

“…They would become equivalent to fourth-generation potentials for an infinite number of passing steps, or for small molecules in vacuum that can be treated globally. Recently, several message passing network have been proposed which are also capable to describe systems in different total charge states [58,59].…”

Section: Beyond Locality-long-ranged Mlpsmentioning

confidence: 99%

See 2 more Smart Citations

Machine learning potentials for extended systems: a perspective

2021

View full text Add to dashboard Cite

In the past two and a half decades machine learning potentials have evolved from a special purpose solution to a broadly applicable tool for large-scale atomistic simulations. By combining the efficiency of empirical potentials and force fields with an accuracy close to first-principles calculations they now enable computer simulations of a wide range of molecules and materials. In this perspective, we summarize the present status of these new types of models for extended systems, which are increasingly used for materials modelling. There are several approaches, but they all have in common that they exploit the locality of atomic properties in some form. Long-range interactions, most prominently electrostatic interactions, can also be included even for systems in which non-local charge transfer leads to an electronic structure that depends globally on all atomic positions. Remaining challenges and limitations of current approaches are discussed. Graphic Abstract

show abstract

Section: Beyond Locality-long-ranged Mlpsmentioning

confidence: 99%

Section: Beyond Locality-long-ranged Mlpsmentioning

confidence: 99%

See 1 more Smart Citation

Machine learning potentials for extended systems: a perspective

2021

View full text Add to dashboard Cite

show abstract

“…where R (0) ij is the identity. As proposed by Metcalf et al 20 , charge conservation can be enforced elegantly by imposing anti-symmetry on φ M (0) with respect to the hidden features h n i and h n j . Alternatively, an atom-based scheme can be applied where each atomic monopole M (0) i is predicted based on the hidden feature h n i of the respective atom, i.e.…”

Section: Equivariance For Multipolesmentioning

confidence: 99%

“…Previous efforts to use ML for the prediction of electrostatic interactions have mostly focused on partial charges, with models that either solely predict partial charges [19][20][21] , models that combine a local treatment with an explicit treatment of long-range interactions based on partial charges 8,22,23 , or models that predict partial charges as an auxiliary variable for properties such as the molecular dipole 24,25 . Due to their orientational dependence and smaller influence on the ESP, higher-order atomic multipoles have found much less attention.…”

Section: Introductionmentioning

confidence: 99%

Learning Atomic Multipoles: Prediction of the Electrostatic Potential with Equivariant Graph Neural Networks

Thürlemann,

Böselt,

Riniker

2021

Preprint

View full text Add to dashboard Cite

The accurate description of electrostatic interactions remains a challenging problem for fitted potential-energy functions. The commonly used fixed partial-charge approximation fails to reproduce the electrostatic potential at short range due to its insensitivity to conformational changes and anisotropic effects. At the same time, possibly more accurate machine-learned (ML) potentials struggle with the long-range behaviour due to their inherent locality ansatz. Employing a multipole expansion offers in principle an exact treatment of the electrostatic potential such that the long-range and short-range electrostatic interaction can be treated simultaneously with high accuracy. However, such an expansion requires the calculation of the electron density using computationally expensive quantum-mechanical (QM) methods. Here, we introduce an equivariant graph neural network (GNN) to address this issue. The proposed model predicts atomic multipoles up to the quadrupole, circumventing the need of expensive QM computations. By using an equivariant architecture, the model enforces the correct symmetry by design without relying on local reference frames. The GNN reproduces the electrostatic potential of various systems with high fidelity. Possible use cases for such an approach include the separate treatment of long-range interactions in ML potentials, the analysis of electrostatic potential surfaces, and the application in polarizable force fields.

show abstract

Decoding Electrochemical Processes of Lithium‐Ion Batteries by Classical Molecular Dynamics Simulations

Tan,

Chen,

Zhang

et al. 2024

Advanced Energy Materials

View full text Add to dashboard Cite

Lithium‐ion batteries (LIBs) have played an essential role in the energy storage industry and dominated the power sources for consumer electronics and electric vehicles. Understanding the electrochemistry of LIBs at the molecular scale is significant for improving their performance, stability, lifetime, and safety. Classical molecular dynamics (MD) simulations could directly capture the atomic and molecular motions and thus provide dynamic insights into the electrochemical processes and ion transport in LIBs during charging and discharging that are usually challenging to observe experimentally, which is momentous in developing LIBs with superb performance. This review discusses developments in MD approaches using non‐reactive force fields, reactive force fields, and machine learning potential for modeling chemical reactions and transport of reactants in the electrodes, electrolytes, and electrode‐electrolyte interfaces. It also comprehensively discusses how molecular interactions, structures, transport, and reaction processes affect electrode stability, energy capacity, and interfacial properties. Finally, the remaining challenges and envisioned future routes are commented on for high‐fidelity, effective simulation methods to decode the invisible interactions and chemical reactions in LIBs.

show abstract

Electron-Passing Neural Networks for Atomic Charge Prediction in Systems with Arbitrary Molecular Charge

Cited by 31 publications

References 23 publications

Machine learning potentials for extended systems: a perspective

Machine learning potentials for extended systems: a perspective

Learning Atomic Multipoles: Prediction of the Electrostatic Potential with Equivariant Graph Neural Networks

Decoding Electrochemical Processes of Lithium‐Ion Batteries by Classical Molecular Dynamics Simulations

Contact Info

Product

Resources

About