Theory and practice of modeling van der Waals interactions in electronic-structure calculations

Stöhr, Martin; Voorhis, Troy Van; Tkatchenko, Alexandre

doi:10.1039/c9cs00060g

Cited by 148 publications

(121 citation statements)

References 351 publications

(576 reference statements)

Supporting

Mentioning

116

Contrasting

Order By: Relevance

“…In addition to electrostatics, van der Waals (vdW) dispersion interactions can also have a substantial long-range character, i.e. they can extend to separations of tens of nanometers or more in large molecular and nanoscale systems [133,134,135]. Developing ML models that correctly treat the quantum-mechanical many-body nature of vdW interactions remains a difficult challenge to overcome [15].…”

Section: Long-ranged Interactionsmentioning

confidence: 99%

Machine Learning for Molecular Simulation

Noé

Tkatchenko

Müller

et al. 2020

Annu. Rev. Phys. Chem.

Self Cite

672

476

View full text Add to dashboard Cite

Machine learning (ML) is transforming all areas of science. The complex and time-consuming calculations in molecular simulations are particularly suitable for a machine learning revolution and have already been profoundly impacted by the application of existing ML methods. Here we review recent ML methods for molecular simulation, with particular focus on (deep) neural networks for the prediction of quantum-mechanical energies and forces, coarse-grained molecular dynamics, the extraction of free energy surfaces and kinetics and generative network approaches to sample molecular equilibrium structures and compute thermodynamics. To explain these methods and illustrate open methodological problems, we review some important principles of molecular physics and describe how they can be incorporated into machine learning structures. Finally, we identify and describe a list of open challenges for the interface between ML and molecular simulation. 1 arXiv:1911.02792v1 [physics.chem-ph] 7 Nov 2019 complex relationship between the input (the pixels) and the output (the labels) that is unknown in its explicit form but can be inferred by a suitable algorithm. Clearly, such an operating principle can be very useful in the description of atomic and molecular systems as well. We know that atomistic configurations dictate the chemical properties, and the machine can learn to associate the latter to the former without solving first principle equations, if presented with enough examples. Although different machine learning tools are available and have been applied to molecular simulation (e.g., kernel methods [2]), here we mostly focus on the use of neural networks, now often synonymously used with the term "deep learning". We assume the reader has basic knowledge of machine learning and we refer to the literature for an introduction to statistical learning theory [3,4] and deep learning [5,6].One of the first applications of machine learning in Chemistry has been to extract classical potential energy surfaces from quantum mechanical (QM) calculations, in order to efficiently perform molecular dynamics (MD) simulations that can incorporate quantum effects. The seminal work of Behler and Parrinello in this direction [7] has opened the way to a now rapidly advancing area of research [8,9,10,11,12,13,14,15]. In addition to atomistic force fields, it has been recently shown that, in the same spirit, effective molecular models at resolution coarser than atomistic can be designed by ML [16,17,18]. Analysis and simulation of MD trajectories has also been affected by ML, for instance for the definition of optimal reaction coordinates [19,20,21,22,23,24], the estimate of free energy surfaces [25,26,27,22], the construction of Markov State Models [21,23,28], and for enhancing MD sampling by learning bias potentials [29,30,31,32,33] or selecting starting configurations by active learning [34,35,36]. Finally, ML can be used to generate samples from the equilibrium distribution of a molecular system without performing MD altogether, as proposed...

show abstract

Section: Long-ranged Interactionsmentioning

confidence: 99%

Machine Learning for Molecular Simulation

Noé

Tkatchenko

Müller

et al. 2020

Annu. Rev. Phys. Chem.

Self Cite

672

476

View full text Add to dashboard Cite

show abstract

“…The lattice energy can be obtained from experimental heats of sublimation or Born-Haber cycles; see [9] for a comparison of the two approaches. The challenges and approaches to calculating the interaction energies of molecular crystals have recently been reviewed in [10][11][12][13][14][15][16]. In particular, they originate from the weak intermolecular interactions due to hydrogen bonding and van-der-Waals forces.…”

Section: = − (1)mentioning

confidence: 99%

Intermolecular Interactions in Molecular Organic Crystals upon Relaxation of Lattice Parameters

Stein

Heimsaat

2019

Crystals

View full text Add to dashboard Cite

Crystal structure prediction is based on the assumption that the most thermodynamically stable structure will crystallize first. The existence of other structures such as polymorphs or from counterenantiomers requires an accurate calculation of the electronic energy. Using atom-centered Gaussian basis functions in periodic Density Functional Theory (DFT) calculations in Turbomole, the performance of two dispersion-corrected functionals, PBE-D3 and B97-D, is assessed for molecular organic crystals of the X23 benchmark set. B97-D shows a MAE (mean absolute error) of 4 kJ/mol, compared to 9 kJ/mol for PBE-D3. A strategy for the convergence of lattice energies towards the basis set limit is outlined. A simultaneous minimization of molecular structures and lattice parameters shows that both methods are able to reproduce experimental unit cell parameters to within 4–5%. Calculated lattice energies, however, deviate slightly more from the experiment, i.e., by 0.4 kJ/mol after unit cell optimization for PBE-D3 and 0.5 kJ/mol for B97-D. The accuracy of the calculated lattice energies compared to the experimental values demonstrates the ability of current DFT methods to assist in the quest for possible polymorphs and enantioselective crystallization processes.

show abstract

“…Noncovalent interactions are not only ubiquitous, but they are also absolutely necessary to describe many physical, chemical, and biochemical phenomena, ranging from thermodynamics of nonideal gases to spectroscopy and scattering cross sections of interstellar complexes to the performance and selectivity (including enantioselectivity) of molecular catalysts to the secondary, tertiary, and quaternary structure of proteins, the double‐helix structure of DNA, and the interaction between enzymes and their substrates (or inhibitors). Thus, theoretical and experimental investigations of noncovalent interactions are published in very large quantities, and various aspects of these interactions have been reviewed in this journal and elsewhere …”

Section: Introductionmentioning

confidence: 99%

Recent developments in symmetry‐adapted perturbation theory

Patkowski

2019

WIREs Comput Mol Sci

159

144

View full text Add to dashboard Cite

Symmetry-adapted perturbation theory (SAPT) is a well-established method to compute accurate intermolecular interaction energies in terms of physical effects such as electrostatics, induction (polarization), dispersion, and exchange. With many theory levels and variants, and several computer implementations available, closed-shell SAPT has been applied to produce numerous intermolecular potential energy surfaces for complexes of experimental interest, and to elucidate the interactions in various complexes relevant to catalysis, organic synthesis, and biochemistry. In contrast, the development of SAPT for general open-shell complexes is still a work in progress. In the last decade, new developments from several research groups, including the author's, have greatly enhanced the capabilities of SAPT. The new and emerging approaches are designed to make SAPT more widely applicable (including interactions involving multireference systems, complexes in arbitrary spin states, and intramolecular noncovalent interactions), more accurate (enhanced description of intramolecular correlation, a better account of exchange effects, relativistic SAPT, and explicitly correlated SAPT), and more efficient (enhanced density-fitted implementations, linearscaling variants, empirical dispersion, and an implementation on graphics processing units).The new developments open up avenues for SAPT applications to an unprecedented variety of weakly interacting complexes.

show abstract

Theory and practice of modeling van der Waals interactions in electronic-structure calculations

Abstract: Opening the black box of van der Waals-inclusive electronic structure calculations: a tutorial-style introduction to van der Waals dispersion interactions, state-of-the-art methods in computational modeling and complementary experimental techniques.

Cited by 148 publications

References 351 publications

Machine Learning for Molecular Simulation

Machine Learning for Molecular Simulation

Intermolecular Interactions in Molecular Organic Crystals upon Relaxation of Lattice Parameters

Recent developments in symmetry‐adapted perturbation theory

Contact Info

Product

Resources

About