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

Zhang, Liwei; Onat, Berk; Dusson, Geneviève; McSloy, Adam; Anand, G.; Maurer, Reinhard J.; Ortner, Christoph; Kermode, James R.

doi:10.1038/s41524-022-00843-2

Cited by 27 publications

(19 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To go beyond the simple models investigated here, in the future, it may be possible to parameterize high-dimensional models more closely to ab initio data 90 and make dynamics simulations feasible using machine learning techniques. 91 It is hoped that this work can be used as a foundation for further tests of mixed quantum-classical methods for dynamics at surfaces. To this end, the models introduced may be used as test systems for methods that emerge in the future to comprehensively compare their performance or as starting points to explore other effects and parameter regimes.…”

Section: Discussionmentioning

confidence: 98%

“…On the topic of decoherence corrections in IESH, it will be worthwhile to investigate the relative performance of different decoherence corrections for a collection of benchmark problems. To go beyond the simple models investigated here, in the future, it may be possible to parameterize high-dimensional models more closely to ab initio data and make dynamics simulations feasible using machine learning techniques …”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Assessing Mixed Quantum-Classical Molecular Dynamics Methods for Nonadiabatic Dynamics of Molecules on Metal Surfaces

2023

Self Cite

View full text Add to dashboard Cite

Mixed quantum-classical (MQC) methods for simulating the dynamics of molecules at metal surfaces have the potential to accurately and efficiently provide mechanistic insight into reactive processes. Here, we introduce simple two-dimensional models for the scattering of diatomic molecules at metal surfaces based on recently published electronic structure data. We apply several MQC methods to investigate their ability to capture how nonadiabatic effects influence molecule−metal energy transfer during the scattering process. Specifically, we compare molecular dynamics with electronic friction, Ehrenfest dynamics, independent electron surface hopping, and the broadened classical master equation approach. In the case of independent electron surface hopping, we implement a simple decoherence correction approach and assess its impact on vibrationally inelastic scattering. Our results show that simple, low-dimensional models can be used to qualitatively capture experimentally observed vibrational energy transfer and provide insight into the relative performance of different MQC schemes. We observe that all approaches predict similar kinetic energy dependence but return different vibrational energy distributions. Finally, by varying the molecule−metal coupling, we can assess the coupling regime in which some MQC methods become unsuitable.

show abstract

Section: Discussionmentioning

confidence: 98%

Section: Discussionmentioning

confidence: 99%

Assessing Mixed Quantum-Classical Molecular Dynamics Methods for Nonadiabatic Dynamics of Molecules on Metal Surfaces

2023

Self Cite

View full text Add to dashboard Cite

show abstract

“…The electron Hamiltonian of molecules and materials is vital for understanding the systems and calculating electron-related physical quantities. In recent years, the Hamiltonian derived from the neural network has been developed a lot for nonmagnetic systems [44][45][46][47][48][49]. In a recent work, we have implemented a transferable E(3) equivariant model for predicting the electron Hamiltonian of non-magnetic systems [49].…”

Section: Main Textmentioning

confidence: 99%

Time-reversal equivariant neural network potential and Hamiltonian for magnetic materials

Zhong

et al. 2022

Preprint

View full text Add to dashboard Cite

This work presents Time-reversal Equivariant Neural Network (TENN) framework. With TENN, the time-reversal symmetry is considered in the equivariant neural network (ENN), which generalizes the ENN to consider physical quantities related to time-reversal symmetry such as spin and velocity of atoms. TENN-e3, as the time-reversal-extension of E(3) equivariant neural network, is developed to keep the Time-reversal E(3) equivariant with consideration of whether to include the spin-orbit effect for both collinear and non-collinear magnetic moments situations for magnetic material. TENN-e3 can construct spin neural network potential and the Hamiltonian of magnetic material from ab-initio calculations. Time-reversal-E(3)-equivariant convolutions for interactions of spinor and geometric tensors are employed in TENN-e3. Compared to the popular ENN, TENN-e3 can describe the complex spin-lattice coupling with high accuracy and keep time-reversal symmetry which is not preserved in the existing E(3)-equivariant model. Also, the Hamiltonian of magnetic material with time-reversal symmetry can be built with TENN-e3. TENN paves a new way to spin-lattice dynamics simulations over long-time scales and electronic structure calculations of large-scale magnetic materials.

show abstract

“…25−30 The electronic structure is constrained by physical symmetries, which can be exploited by constructing equivariant ML schemes that ensure that the data-driven model conforms to these constraints. 24,31,32 Once the Hamiltonian matrix is obtained, all sorts of ground-state properties, such as the electron density, can be obtained with simple, inexpensive manipulations. Furthermore, excited states can also be predicted, at least approximately, by postprocessing the ground-state single-particle Hamiltonian.…”

Section: ■ Introductionmentioning

confidence: 99%

Electronic Excited States from Physically Constrained Machine Learning

Cignoni,

Suman,

Nigam

et al. 2024

ACS Cent. Sci.

View full text Add to dashboard Cite

Data-driven techniques are increasingly used to replace electronic-structure calculations of matter. In this context, a relevant question is whether machine learning (ML) should be applied directly to predict the desired properties or combined explicitly with physically grounded operations. We present an example of an integrated modeling approach in which a symmetry-adapted ML model of an effective Hamiltonian is trained to reproduce electronic excitations from a quantum-mechanical calculation. The resulting model can make predictions for molecules that are much larger and more complex than those on which it is trained and allows for dramatic computational savings by indirectly targeting the outputs of well-converged calculations while using a parametrization corresponding to a minimal atomcentered basis. These results emphasize the merits of intertwining data-driven techniques with physical approximations, improving the transferability and interpretability of ML models without affecting their accuracy and computational efficiency and providing a blueprint for developing ML-augmented electronic-structure methods.

show abstract

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

Cited by 27 publications

References 50 publications

Assessing Mixed Quantum-Classical Molecular Dynamics Methods for Nonadiabatic Dynamics of Molecules on Metal Surfaces

Assessing Mixed Quantum-Classical Molecular Dynamics Methods for Nonadiabatic Dynamics of Molecules on Metal Surfaces

Time-reversal equivariant neural network potential and Hamiltonian for magnetic materials

Electronic Excited States from Physically Constrained Machine Learning

Contact Info

Product

Resources

About