arXiv subjects

Gábor Csányi

Publications and source records attributed to Gábor Csányi.

At least 73 records · Page 4Linked to original sources

arXiv2021

Physics-inspired structural representations for molecules and materials

The first step in the construction of a regression model or a data-driven analysis, aiming to predict or elucidate the relationship between the atomic scale structure of matter and its properties, involves transforming the Cartesian coordinates of the atoms into a suitable representation. The development of atomic-scale representations has played, and continues to play, a central role in the success of machine-learning methods for chemistry and materials science. This review summarizes the current understanding of the nature and characteristics of the most commonly used structural and chemical descriptions of atomistic structures, highlighting the deep underlying connections between different frameworks, and the ideas that lead to computationally efficient and universally applicable models. It emphasizes the link between properties, structures, their physical chemistry and their mathematical description, provides examples of recent applications to a diverse set of chemical and materials science problems, and outlines the open questions and the most promising research directions in the field.

physics.chem-ph↗

arXiv2021

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

The atomic cluster expansion is a general polynomial expansion of the atomic energy in multi-atom basis functions. Here we implement the atomic cluster expansion in the performant C++ code \verb+PACE+ that is suitable for use in large scale atomistic simulations. We briefly review the atomic cluster expansion and give detailed expressions for energies and forces as well as efficient algorithms for their evaluation. We demonstrate that the atomic cluster expansion as implemented in \verb+PACE+ shifts a previously established Pareto front for machine learning interatomic potentials towards faster and more accurate calculations. Moreover, general purpose parameterizations are presented for copper and silicon and evaluated in detail. We show that the new Cu and Si potentials significantly improve on the best available potentials for highly accurate large-scale atomistic simulations.

cond-mat.mtrl-sci↗

arXiv2020

Symmetry-Aware Actor-Critic for 3D Molecular Design

Automating molecular design using deep reinforcement learning (RL) has the potential to greatly accelerate the search for novel materials. Despite recent progress on leveraging graph representations to design molecules, such methods are fundamentally limited by the lack of three-dimensional (3D) information. In light of this, we propose a novel actor-critic architecture for 3D molecular design that can generate molecular structures unattainable with previous approaches. This is achieved by exploiting the symmetries of the design process through a rotationally covariant state-action representation based on a spherical harmonics series expansion. We demonstrate the benefits of our approach on several 3D molecular design tasks, where we find that building in such symmetries significantly improves generalization and the quality of generated molecules.

stat.ML↗

arXiv2020

Learning the electronic density of states in condensed matter

The electronic density of states (DOS) quantifies the distribution of the energy levels that can be occupied by electrons in a quasiparticle picture, and is central to modern electronic structure theory. It also underpins the computation and interpretation of experimentally observable material properties such as optical absorption and electrical conductivity. We discuss the challenges inherent in the construction of a machine-learning (ML) framework aimed at predicting the DOS as a combination of local contributions that depend in turn on the geometric configuration of neighbours around each atom, using quasiparticle energy levels from density functional theory as training data. We present a challenging case study that includes configurations of silicon spanning a broad set of thermodynamic conditions, ranging from bulk structures to clusters, and from semiconducting to metallic behavior. We compare different approaches to represent the DOS, and the accuracy of predicting quantities such as the Fermi level, the DOS at the Fermi level, or the band energy, either directly or as a side-product of the evaluation of the DOS. The performance of the model depends crucially on the smoothening of the DOS, and there is a tradeoff to be made between the systematic error associated with the smoothening and the error in the ML model for a specific structure. We demonstrate the usefulness of this approach by computing the density of states of a large amorphous silicon sample, for which it would be prohibitively expensive to compute the DOS by direct electronic structure calculations, and show how the atom-centred decomposition of the DOS that is obtained through our model can be used to extract physical insights into the connections between structural and electronic features.

cond-mat.mtrl-sci↗

arXiv2020

Machine learning driven simulated deposition of carbon films: from low-density to diamondlike amorphous carbon

Amorphous carbon (a-C) materials have diverse interesting and useful properties, but the understanding of their atomic-scale structures is still incomplete. Here, we report on extensive atomistic simulations of the deposition and growth of a-C films, describing interatomic interactions using a machine learning (ML) based Gaussian Approximation Potential (GAP) model. We expand widely on our initial work [Phys. Rev. Lett. 120, 166101 (2018)] by now considering a broad range of incident ion energies, thus modeling samples that span the entire range from low-density ($sp^{2}$-rich) to high-density ($sp^{3}$-rich, "diamond-like") amorphous forms of carbon. Two different mechanisms are observed in these simulations, depending on the impact energy: low-energy impacts induce $sp$- and $sp^{2}$-dominated growth directly around the impact site, whereas high-energy impacts induce peening. Furthermore, we propose and apply a scheme for computing the anisotropic elastic properties of the a-C films. Our work provides fundamental insight into this intriguing class of disordered solids, as well as a conceptual and methodological blueprint for simulating the atomic-scale deposition of other materials with ML-driven molecular dynamics.

cond-mat.mtrl-sci↗

arXiv2020

Atomic Permutationally Invariant Polynomials for Fitting Molecular Force Fields

We introduce and explore an approach for constructing force fields for small molecules, which combines intuitive low body order empirical force field terms with the concepts of data driven statistical fits of recent machine learned potentials. We bring these two key ideas together to bridge the gap between established empirical force fields that have a high degree of transferability on the one hand, and the machine learned potentials that are systematically improvable and can converge to very high accuracy, on the other. Our framework extends the atomic Permutationally Invariant Polynomials (aPIP) developed for elemental materials in [Mach. Learn.: Sci. Technol. 2019 1 015004] to molecular systems. The body order decomposition allows us to keep the dimensionality of each term low, while the use of an iterative fitting scheme as well as regularisation procedures improve the extrapolation outside the training set. We investigate aPIP force fields with up to generalised 4-body terms, and examine the performance on a set of small organic molecules. We achieve a high level of accuracy when fitting individual molecules, comparable to those of the many-body machine learned force fields. Fitted to a combined training set of short linear alkanes, the accuracy of the aPIP force field still significantly exceeds what can be expected from classical empirical force fields, while retaining reasonable transferability to both configurations far from the training set and to new molecules.

physics.chem-ph↗

arXiv2020

An Accurate and Transferable Machine Learning Potential for Carbon

We present an accurate machine learning (ML) model for atomistic simulations of carbon, constructed using the Gaussian approximation potential (GAP) methodology. The potential, named GAP-20, describes the properties of the bulk crystalline and amorphous phases, crystal surfaces and defect structures with an accuracy approaching that of direct ab initio simulation, but at a significantly reduced cost. We combine structural databases for amorphous carbon and graphene, which we extend substantially by adding suitable configurations, for example, for defects in graphene and other nanostructures. The final potential is fitted to reference data computed using the optB88-vdW density functional theory (DFT) functional. Dispersion interactions, which are crucial to describe multilayer carbonaceous materials, are therefore implicitly included. We additionally account for long-range dispersion interactions using a semianalytical two-body term and show that an improved model can be obtained through an optimisation of the many-body smooth overlap of atomic positions (SOAP) descriptor. We rigorously test the potential on lattice parameters, bond lengths, formation energies and phonon dispersions of numerous carbon allotropes. We compare the formation energies of an extensive set of defect structures, surfaces and surface reconstructions to DFT reference calculations. The present work demonstrates the ability to combine, in the same ML model, the previously attained flexibility required for amorphous carbon [Phys. Rev. B, 95, 094203, (2017)] with the high numerical accuracy necessary for crystalline graphene [Phys. Rev. B, 97, 054303, (2018)], thereby providing an interatomic potential that will be applicable to a wide range of applications concerning diverse forms of bulk and nanostructured carbon.

physics.comp-ph↗

arXiv2020

On the Completeness of Atomic Structure Representations

Many-body descriptors are widely used to represent atomic environments in the construction of machine learned interatomic potentials and more broadly for fitting, classification and embedding tasks on atomic structures. It was generally believed that 3-body descriptors uniquely specify the environment of an atom, up to a rotation and permutation of like atoms. We produce several counterexamples to this belief, with the consequence that any classifier, regression or embedding model for atom-centred properties that uses 3 (or 4)-body features will incorrectly give identical results for different configurations. Writing global properties (such as total energies) as a sum of many atom-centred contributions mitigates, but does not eliminate, the impact of this fundamental deficiency -- explaining the success of current "machine-learning" force fields. We anticipate the issues that will arise as the desired accuracy increases, and suggest potential solutions.

physics.chem-ph↗

arXiv2020

Combining phonon accuracy with high transferability in Gaussian approximation potential models

Machine learning driven interatomic potentials, including Gaussian approximation potential (GAP) models, are emerging tools for atomistic simulations. Here, we address the methodological question of how one can fit GAP models that accurately predict vibrational properties in specific regions of configuration space, whilst retaining flexibility and transferability to others. We use an adaptive regularization of the GAP fit that scales with the absolute force magnitude on any given atom, thereby exploring the Bayesian interpretation of GAP regularization as an "expected error", and its impact on the prediction of physical properties for a material of interest. The approach enables excellent predictions of phonon modes (to within 0.1-0.2 THz) for structurally diverse silicon allotropes, and it can be coupled with existing fitting databases for high transferability. These findings and workflows are expected to be useful for GAP-driven materials modeling more generally.

cond-mat.mtrl-sci↗

arXiv2020

Gaussian Approximation Potentials: a brief tutorial introduction

We present a swift walk-through of our recent work that uses machine learning to fit interatomic potentials based on quantum mechanical data. We describe our Gaussian Approximation Potentials (GAP) framework, discussing a variety of descriptors, how to train the model on total energies and derivatives and the simultaneous use of multiple models. We also show a small example using QUIP, the software sandbox implementation of GAP that is available for non-commercial use.

cond-mat.mtrl-sci↗

arXiv2020

Comparing molecules and solids across structural and alchemical space

Evaluating the (dis)similarity of crystalline, disordered and molecular compounds is a critical step in the development of algorithms to navigate automatically the configuration space of complex materials. For instance, a structural similarity metric is crucial for classifying structures, searching chemical space for better compounds and materials, and driving the next generation of machine-learning techniques for predicting the stability and properties of molecules and materials. In the last few years several strategies have been designed to compare atomic coordination environments. In particular, the Smooth Overlap of Atomic Positions (SOAP) has emerged as an elegant framework to obtain translation, rotation and permutation-invariant descriptors of groups of atoms, driven by the design of various classes of machine-learned inter-atomic potentials. Here we discuss how one can combine such local descriptors using a Regularized Entropy Match (REMatch) approach to describe the similarity of both whole molecular and bulk periodic structures, introducing powerful metrics that enable the navigation of alchemical and structural complexity within a unified framework. Furthermore, using this kernel and a ridge regression method we can predict atomization energies for a database of small organic molecules with a mean absolute error below 1kcal/mol, reaching an important milestone in the application of machine-learning techniques to the evaluation of molecular properties.

cond-mat.mtrl-sci↗

arXiv2019

Structural transitions in dense disordered silicon from quantum-accurate ultra-large-scale simulations

Structurally disordered materials continue to pose fundamental questions, including that of how different disordered phases ("polyamorphs") can coexist and transform from one to another. As a widely studied case, amorphous silicon (a-Si) forms a fourfold-coordinated, covalent random network at ambient conditions, but much higher-coordinated, metallic-like phases under pressure. However, a detailed mechanistic understanding of the liquid-amorphous and amorphous-amorphous transitions in silicon has been lacking, due to intrinsic limitations of even the most advanced experimental and computational techniques. Here, we show how machine-learning (ML)-driven simulations can break through this long-standing barrier, affording a comprehensive, quantum-accurate, and fully atomistic description of all relevant liquid and amorphous phases of silicon. Combining a model system size of 100,000 atoms (ten-nanometre length scale) with a prediction accuracy of a few meV per atom, our simulations reveal a remarkable, three-step transformation sequence for a-Si under increasing external pressure. First, up to 10-11 GPa, polyamorphic low- and high-density amorphous (LDA and HDA) regions are found to coexist, rather than appearing sequentially. Then, we observe a structural collapse into a distinct, very-high-density amorphous (VHDA) phase at 12-13 GPa, reminiscent of the dense liquid but being formed at a much lower temperature. Finally, our simulations indicate the transient nature of this VHDA phase: it rapidly nucleates crystallites at 13-16 GPa, ultimately leading to the formation of a poly-crystalline, simple-hexagonal structure, consistent with experiments but not seen in earlier simulations.

cond-mat.mtrl-sci↗

arXiv2019

A Performance and Cost Assessment of Machine Learning Interatomic Potentials

Machine learning of the quantitative relationship between local environment descriptors and the potential energy surface of a system of atoms has emerged as a new frontier in the development of interatomic potentials (IAPs). Here, we present a comprehensive evaluation of ML-IAPs based on four local environment descriptors --- Behler-Parrinello symmetry functions, smooth overlap of atomic positions (SOAP), the Spectral Neighbor Analysis Potential (SNAP) bispectrum components, and moment tensors --- using a diverse data set generated using high-throughput density functional theory (DFT) calculations. The data set comprising bcc (Li, Mo) and fcc (Cu, Ni) metals and diamond group IV semiconductors (Si, Ge) is chosen to span a range of crystal structures and bonding. All descriptors studied show excellent performance in predicting energies and forces far surpassing that of classical IAPs, as well as predicting properties such as elastic constants and phonon dispersion curves. We observe a general trade-off between accuracy and the degrees of freedom of each model, and consequently computational cost. We will discuss these trade-offs in the context of model selection for molecular dynamics and other applications.

physics.comp-ph↗

arXiv2019

Machine-learned Interatomic Potentials for Alloys and Alloy Phase Diagrams

We introduce machine-learned potentials for Ag-Pd to describe the energy of alloy configurations over a wide range of compositions. We compare two different approaches. Moment tensor potentials (MTP) are polynomial-like functions of interatomic distances and angles. The Gaussian Approximation Potential (GAP) framework uses kernel regression, and we use the Smooth Overlap of Atomic Positions (SOAP) representation of atomic neighbourhoods that consists of a complete set of rotational and permutational invariants provided by the power spectrum of the spherical Fourier transform of the neighbour density. Both types of potentials give excellent accuracy for a wide range of compositions and rival the accuracy of cluster expansion, a benchmark for this system. While both models are able to describe small deformations away from the lattice positions, SOAP-GAP excels at transferability as shown by sensible transformation paths between configurations, and MTP allows, due to its lower computational cost, the calculation of compositional phase diagrams. Given the fact that both methods perform as well as cluster expansion would but yield off-lattice models, we expect them to open new avenues in computational materials modeling for alloys.

cond-mat.mtrl-sci↗

arXiv2019

De novo exploration and self-guided learning of potential-energy surfaces

Interatomic potential models based on machine learning (ML) are rapidly developing as tools for materials simulations. However, because of their flexibility, they require large fitting databases that are normally created with substantial manual selection and tuning of reference configurations. Here, we show that ML potentials can be built in a largely automated fashion, exploring and fitting potential-energy surfaces from the beginning (de novo) within one and the same protocol. The key enabling step is the use of a configuration-averaged kernel metric that allows one to select the few most relevant structures at each step. The resulting potentials are accurate and robust for the wide range of configurations that occur during structure searching, despite only requiring a relatively small number of single-point DFT calculations on small unit cells. We apply the method to materials with diverse chemical nature and coordination environments, marking a milestone toward the more routine application of ML potentials in physics, chemistry, and materials science.

cond-mat.mtrl-sci↗

arXiv2019

Machine-learned multi-system surrogate models for materials prediction

Surrogate machine-learning models are transforming computational materials science by predicting properties of materials with the accuracy of ab initio methods at a fraction of the computational cost. We demonstrate surrogate models that simultaneously interpolate energies of different materials on a dataset of 10 binary alloys (AgCu, AlFe, AlMg, AlNi, AlTi, CoNi, CuFe, CuNi, FeV, NbNi) with 10 different species and all possible fcc, bcc and hcp structures up to 8 atoms in the unit cell, 15\,950 structures in total. We find that the deviation of prediction errors when increasing the number of simultaneously modeled alloys is less than 1\,meV/atom. Several state-of-the-art materials representations and learning algorithms were found to qualitatively agree on the prediction errors of formation enthalpy with relative errors of $<$2.5\% for all systems.

cond-mat.mtrl-sci↗

arXiv2019

Machine-learning of atomic-scale properties based on physical principles

We briefly summarize the kernel regression approach, as used recently in materials modelling, to fitting functions, particularly potential energy surfaces, and highlight how the linear algebra framework can be used to both predict and train from linear functionals of the potential energy, such as the total energy and atomic forces. We then give a detailed account of the Smooth Overlap of Atomic Positions (SOAP) representation and kernel, showing how it arises from an abstract representation of smooth atomic densities, and how it is related to several popular density-based representations of atomic structure. We also discuss recent generalisations that allow fine control of correlations between different atomic species, prediction and fitting of tensorial properties, and also how to construct structural kernels---applicable to comparing entire molecules or periodic systems---that go beyond an additive combination of local environments.

physics.comp-ph↗

arXiv2018

Quantifying Chemical Structure and Atomic Energies in Amorphous Silicon Networks

Amorphous materials are coming within reach of realistic computer simulations, but new approaches are needed to fully understand their intricate atomic structures. Here, we show how machine-learning (ML)-based techniques can give new, quantitative chemical insight into the atomic-scale structure of amorphous silicon (a-Si). Based on a similarity function ("kernel"), we define a structural metric that unifies the description of nearest- and next-nearest-neighbor environments in the amorphous state. We apply this to an ensemble of a-Si networks, generated in melt-quench simulations with an ML-based interatomic potential, in which we tailor the degree of ordering by varying the quench rates down to $10^{10}$ K/s (leading to a structural model that is lower in energy than the established WWW network). We then show how "machine-learned" atomic energies permit a chemical interpretation, associating coordination defects in a-Si with distinct energetic stability regions. The approach is straightforward and inexpensive to apply to arbitrary structural models, and it is therefore expected to have more general significance for developing a quantitative understanding of the amorphous state.

cond-mat.mtrl-sci↗