Atomic Simulation Environment Research Articles

Atomic Simulation Recipes: A Python framework and library for automated workflows

The Atomic Simulation Recipes (ASR) is an open source Python framework for working with atomistic materials simulations in an efficient and sustainable way that is ideally suited for high-throughput projects. Central to ASR is the concept of a Recipe: a high-level Python script that performs a well defined simulation task robustly and accurately while keeping track of the data provenance. The ASR leverages the functionality of the Atomic Simulation Environment (ASE) to interface with external simulation codes and attain a high abstraction level. We provide a library of Recipes for common simulation tasks employing density functional theory and many-body perturbation schemes. These Recipes utilize the GPAW electronic structure code, but may be adapted to other simulation codes with an ASE interface. Being independent objects with automatic data provenance control, Recipes can be freely combined through Python scripting giving maximal freedom for users to build advanced workflows. ASR also implements a command line interface that can be used to run Recipes and inspect results. The ASR Migration module helps users maintain their data while the Database and App modules makes it possible to create local databases and present them as customized web pages.

Machine-learning accelerated geometry optimization in molecular simulation.

Geometry optimization is an important part of both computational materials and surface science because it is the path to finding ground state atomic structures and reaction pathways. These properties are used in the estimation of thermodynamic and kinetic properties of molecular and crystal structures. This process is slow at the quantum level of theory because it involves an iterative calculation of forces using quantum chemical codes such as density functional theory (DFT), which are computationally expensive and which limit the speed of the optimization algorithms. It would be highly advantageous to accelerate this process because then one could do either the same amount of work in less time or more work in the same time. In this work, we provide a neural network (NN) ensemble based active learning method to accelerate the local geometry optimization for multiple configurations simultaneously. We illustrate the acceleration on several case studies including bare metal surfaces, surfaces with adsorbates, and nudged elastic band for two reactions. In all cases, the accelerated method requires fewer DFT calculations than the standard method. In addition, we provide an Atomic Simulation Environment (ASE)-optimizer Python package to make the usage of the NN ensemble active learning for geometry optimization easier.

The abTEM code: transmission electron microscopy from first principles.

Simulation of transmission electron microscopy (TEM) images or diffraction patterns is often required to interpret experimental data. Since nuclear cores dominate electron scattering, the scattering potential is typically described using the independent atom model, which completely neglects valence bonding and its effect on the transmitting electrons. As instrumentation has advanced, new measurements have revealed subtle details of the scattering potential that were previously not accessible to experiment. We have created an open-source simulation code designed to meet these demands by integrating the ability to calculate the potential via density functional theory (DFT) with a flexible modular software design. abTEM can simulate most standard imaging modes and incorporates the latest algorithmic developments. The development of new techniques requires a program that is accessible to domain experts without extensive programming experience. abTEM is written purely in Python and designed for easy modification and extension. The effective use of modern open-source libraries makes the performance of abTEM highly competitive with existing optimized codes on both CPUs and GPUs and allows us to leverage an extensive ecosystem of libraries, such as the Atomic Simulation Environment and the DFT code GPAW. abTEM is designed to work in an interactive Python notebook, creating a seamless and reproducible workflow from defining an atomic structure, calculating molecular dynamics (MD) and electrostatic potentials, to the analysis of results, all in a single, easy-to-read document. This article provides ongoing documentation of abTEM development. In this first version, we show use cases for hexagonal boron nitride, where valence bonding can be detected, a 4D-STEM simulation of molybdenum disulfide including ptychographic phase reconstruction, a comparison of MD and frozen phonon modeling for convergent-beam electron diffraction of a 2.6-million-atom silicon system, and a performance comparison of our fastimplementation of thePRISM algorithm for a decahedral 20000-atom gold nanoparticle.

The abTEM code: transmission electron microscopy from first principles

Simulation of transmission electron microscopy (TEM) images or diffraction patterns is often required to interpret experimental data. Since nuclear cores dominate electron scattering, the scattering potential is typically described using the independent atom model, which completely neglects valence bonding and its effect on the transmitting electrons. As instrumentation has advanced, new measurements have revealed subtle details of the scattering potential that were previously not accessible to experiment. We have created an open-source simulation code designed to meet these demands by integrating the ability to calculate the potential via density functional theory (DFT) with a flexible modular software design. abTEM can simulate most standard imaging modes and incorporates the latest algorithmic developments. The development of new techniques requires a program that is accessible to domain experts without extensive programming experience. abTEM is written purely in Python and designed for easy modification and extension. The effective use of modern open-source libraries makes the performance of abTEM highly competitive with existing optimized codes on both CPUs and GPUs and allows us to leverage an extensive ecosystem of libraries, such as the Atomic Simulation Environment and the DFT code GPAW. abTEM is designed to work in an interactive Python notebook, creating a seamless and reproducible workflow from defining an atomic structure, calculating molecular dynamics (MD) and electrostatic potentials, to the analysis of results, all in a single, easy-to-read document. This article provides ongoing documentation of abTEM development. In this first version, we show use cases for hexagonal boron nitride, where valence bonding can be detected, a 4D-STEM simulation of molybdenum disulfide including ptychographic phase reconstruction, a comparison of MD and frozen phonon modeling for convergent-beam electron diffraction of a 2.6-million-atom silicon system, and a performance comparison of our fast implementation of the PRISM algorithm for a decahedral 20000-atom gold nanoparticle.

PyXtal_FF: a python library for automated force field generation

We present PyXtal_FF—a package based on Python programming language—for developing machine learning potentials (MLPs). The aim of PyXtal_FF is to promote the application of atomistic simulations through providing several choices of atom-centered descriptors and machine learning regressions in one platform. Based on the given choice of descriptors (including the atom-centered symmetry functions, embedded atom density, SO4 bispectrum, and smooth SO3 power spectrum), PyXtal_FF can train MLPs with either generalized linear regression or neural network models, by simultaneously minimizing the errors of energy/forces/stress tensors in comparison with the data from ab-initio simulations. The trained MLP model from PyXtal_FF is interfaced with the Atomic Simulation Environment (ASE) package, which allows different types of light-weight simulations such as geometry optimization, molecular dynamics simulation, and physical properties prediction. Finally, we will illustrate the performance of PyXtal_FF by applying it to investigate several material systems, including the bulk SiO2, high entropy alloy NbMoTaW, and elemental Pt for general purposes. Full documentation of PyXtal_FF is available at https://pyxtal-ff.readthedocs.io.

PyFLOSIC: Python-based Fermi-Löwdin orbital self-interaction correction.

We present pyflosic, an open-source, general-purpose python implementation of the Fermi-Löwdin orbital self-interaction correction (FLO-SIC), which is based on the python simulation of chemistry framework (pyscf) electronic structure and quantum chemistry code. Thanks to pyscf, pyflosic can be used with any kind of Gaussian-type basis set, various kinds of radial and angular quadrature grids, and all exchange-correlation functionals within the local density approximation, generalized-gradient approximation (GGA), and meta-GGA provided in the libxc and xcfun libraries. A central aspect of FLO-SIC is the Fermi-orbital descriptors, which are used to estimate the self-interaction correction. Importantly, they can be initialized automatically within pyflosic; they can also be optimized within pyflosic with an interface to the atomic simulation environment, a python library that provides a variety of powerful gradient-based algorithms for geometry optimization. Although pyflosic has already facilitated applications of FLO-SIC to chemical studies, it offers an excellent starting point for further developments in FLO-SIC approaches, thanks to its use of a high-level programming language and pronounced modularity.

Multiscale electrostatic embedding simulations for modeling structure and dynamics of molecules in solution: A tutorial review

AbstractThe main concepts and important technical details of electrostatic embedding quantum mechanics/molecular mechanics (QM/MM) simulations are explained and illustrated with the intent of assisting newcomers in performing and gauging the accuracy of such simulations, focused on smaller molecules in solution. Beginners are advised on how to increase the reliability and accuracy of the simulations through benchmarking. Central considerations on methodologies for QM/MM Molecular Dynamics (MD) simulations are presented, alongside technical fundamentals regarding the construction and manipulation of simulation systems using the python‐based Atomic Simulation Environment (ASE). A worked example of QM/MM Born–Oppenheimer MD is included, and a flowchart summarizing the most salient decisions and tasks within the methodology is presented.

Improvement of the Evolutionary Algorithm on the Atomic Simulation Environment Though Intuitive Starting Population Creation and Clustering

Improvement of the Evolutionary Algorithm on the Atomic Simulation Environment Though Intuitive Starting Population Creation and Clustering

PiNN: A Python Library for Building Atomic Neural Networks of Molecules and Materials.

Atomic neural networks (ANNs) constitute a class of machine learning methods for predicting potential energy surfaces and physicochemical properties of molecules and materials. Despite many successes, developing interpretable ANN architectures and implementing existing ones efficiently are still challenging. This calls for reliable, general-purpose, and open-source codes. Here, we present a python library named PiNN as a solution toward this goal. In PiNN, we designed a new interpretable and high-performing graph convolutional neural network variant, PiNet, as well as implemented the established Behler-Parrinello neural network. These implementations were tested using datasets of isolated small molecules, crystalline materials, liquid water, and an aqueous alkaline electrolyte. PiNN comes with a visualizer called PiNNBoard to extract chemical insight "learned" by ANNs. It provides analytical stress tensor calculations and interfaces to both the atomic simulation environment and a development version of the Amsterdam Modeling Suite. Moreover, PiNN is highly modularized, which makes it useful not only as a standalone package but also as a chain of tools to develop and to implement novel ANNs. The code is distributed under a permissive BSD license and is freely accessible at https://github.com/Teoroo-CMC/PiNN/ with full documentation and tutorials.

WulffPack: A Python package for Wulff constructions

WulffPack is a Python package that carries out the Wulff construction and its generalizations using an efficient algorithm based on calculation of the convex hull of the vertices of the dual of the Wulff polyhedron. The user provides surface energies and crystal symmetry and WulffPack returns a versatile object that, at its core, contains the coordinates of the Wulff shape. Extraction of symmetry operations is handled internally with spglib. WulffPack includes functionality for visualizing the constructed shapes using Matplotlib. There are also functions for analyzing the constructed shape, most notably in terms of area fraction of symmetrically inequivalent facets. This quantity is important in applications where properties of the material are facet-dependent, such as in catalysis. Finally, using the Atomic Simulation Environment, an atomistic representation of the Wulff shape can also be extracted. An extensive user guide including a documentation of the API is available at http://wulffpack.materialsmodeling.org/.

CLEASE: a versatile and user-friendly implementation of cluster expansion method

Materials exhibiting a substitutional disorder such as multicomponent alloys and mixed metal oxides/oxyfluorides are of great importance in many scientific and technological sectors. Disordered materials constitute an overwhelmingly large configurational space, which makes it practically impossible to be explored manually using first-principles calculations such as density functional theory due to the high computational costs. Consequently, the use of methods such as cluster expansion (CE) is vital in enhancing our understanding of the disordered materials. CE dramatically reduces the computational cost by mapping the first-principles calculation results on to a Hamiltonian which is much faster to evaluate. In this work, we present our implementation of the CE method, which is integrated as a part of the atomic simulation environment (ASE) open-source package. The versatile and user-friendly code automates the complex set up and construction procedure of CE while giving the users the flexibility to tweak the settings and to import their own structures and previous calculation results. Recent advancements such as regularization techniques from machine learning are implemented in the developed code. The code allows the users to construct CE on any bulk lattice structure, which makes it useful for a wide range of applications involving complex materials. We demonstrate the capabilities of our implementation by analyzing the two example materials with varying complexities: a binary metal alloy and a disordered lithium chromium oxyfluoride.

SIMPLE-NN: An efficient package for training and executing neural-network interatomic potentials

The molecular dynamics (MD) simulation is a favored method in materials science for understanding and predicting material properties from atomistic motions. In classical MD simulations, the interaction between atoms is described by an empirical interatomic potential, so the reliability of the simulation hinges on the accuracy of the underlying potential. Recently, machine learning (ML) based interatomic potentials are gaining attention as they can reproduce potential energy surfaces (PES) of abinitio calculations, with a much lower computational cost. Therefore, an efficient code for training ML potentials and inferencing PES in new configurations would widen the application range of MD simulations. Here, we announce an open-source package, SNU Interatomic Machine-learning PotentiaL packagE-version Neural Network (SIMPLE-NN) that generates and utilizes the ML potential based on the artificial neural network with the Behler–Parrinello type symmetry function as descriptors for the chemical environments. SIMPLE-NN uses the Atomic Simulation Environment (ASE) package and Google Tensorflow for high expandability and efficient training, and also supports the in-house code for quasi-Newton method. Notably, the package features a weighting scheme based on the Gaussian density function (GDF), which significantly improves accuracy and reliability of ML potentials by resolving sampling bias that exists in typical training sets. For MD simulations, SIMPLE-NN interfaces with the LAMMPS package. We demonstrate the performance and usage of SIMPLE-NN with examples of SiO2. Program summaryProgram Title: SIMPLE-NNProgram Files doi:http://dx.doi.org/10.17632/pjv2yr7pvr.1Licensing provisions: GPLv3Programming language: Python/C++Nature of problem: Inferencing the potential energy surface for the given system with accuracy comparable to ab initio methods but with much lower computational costs.Solution method: Calculate descriptor vectors that encode local chemical environment. High-dimensional neural network is used to predict the total energy from the descriptor vectors. The trained neural network can be used for molecular dynamics simulations.

R-NEB: Accelerated Nudged Elastic Band Calculations by Use of Reflection Symmetry.

Many activated processes in materials science, physics, and chemistry, e.g. diffusion processes, have initial and final states related by symmetry. Identification of minimum energy paths in such systems with methods such as nudged elastic band (NEB) can gain substantial speed up if the symmetry is exploited. The identification of minimum energy paths and transition states for such processes constitute a large fraction of the CPU-usage within computational materials science; much of which is in essence redundant due to symmetry. Paths with a reflection symmetry can be calculated using about half the computational resources, and the activation energy can, for some transitions, be estimated with high precision with a speed up factor equal to the number of images used in a standard NEB calculation. We present the formal properties required for a system to guarantee a reflection symmetric minimum energy path and an implementation to prepare and effectively speed up nudged elastic band calculations through symmetry considerations. Five examples are given to show the versatility and effectiveness of the method and to validate the implementation. The method is implemented in the open source package Atomic Simulation Environment (ASE) and contains internal methods to identify symmetry relations between the given end point configurations.

SchNetPack: A Deep Learning Toolbox For Atomistic Systems.

SchNetPack is a toolbox for the development and application of deep neural networks that predict potential energy surfaces and other quantum-chemical properties of molecules and materials. It contains basic building blocks of atomistic neural networks, manages their training, and provides simple access to common benchmark datasets. This allows for an easy implementation and evaluation of new models. For now, SchNetPack includes implementations of (weighted) atom-centered symmetry functions and the deep tensor neural network SchNet, as well as ready-to-use scripts that allow one to train these models on molecule and material datasets. Based on the PyTorch deep learning framework, SchNetPack allows one to efficiently apply the neural networks to large datasets with millions of reference calculations, as well as parallelize the model across multiple GPUs. Finally, SchNetPack provides an interface to the Atomic Simulation Environment in order to make trained models easily accessible to researchers that are not yet familiar with neural networks.

Computational Implementation of Nudged Elastic Band, Rigid Rotation, and Corresponding Force Optimization.

The nudged elastic band (NEB) algorithm is the leading method of calculating transition states in chemical systems. However, the current literature lacks adequate guidance for users wishing to implement a key part of NEB, namely, the optimization method. Here, we provide details of this implementation for the following six gradient descent algorithms: steepest descent, quick-min Verlet, FIRE, conjugate gradient, Broyden-Fletcher-Goldfarb-Shanno (BFGS), and limited-memory BFGS (LBFGS). We also construct and implement a new, accelerated backtracking line search method in concert with a partial Procrustes superimposition to improve upon existing methods. Validation is achieved through benchmark calculations of two test cases, the isomerization of CNX and BOX (where X ∈ {H, Li, Na}) and the study of a conformational change within an alanine dipeptide. We also make direct comparisons to the well-established codebase known as the atomic simulation environment.

Theory and Applications of Generalized Pipek-Mezey Wannier Functions.

The theory for the generation of Wannier functions within the generalized Pipek-Mezey approach (Lehtola, S.; Jónsson, H. J. Chem. Theory Comput. 2014, 10, 642) is presented and an implementation thereof is described. Results are shown for systems with periodicity in one, two, and three dimensions as well as isolated molecules. The generalized Pipek-Mezey Wannier functions (PMWF) are highly localized orbitals consistent with chemical intuition where a distinction is maintained between σ- and π-orbitals. The PMWF method is compared with the so-called maximally localized Wannier functions (MLWFs) that are frequently used for the analysis of condensed matter calculations. Whereas PMWFs maximize the localization criterion of Pipek and Mezey, MLWFs maximize that of Foster and Boys and have the disadvantage of mixing σ- and π-orbitals in many cases. The PMWF orbitals turn out to be as localized as the MLWF orbitals as evidenced by cross-comparison of the values of the PMWF and MLWF objective functions for the two types of orbitals. Our implementation in the atomic simulation environment (ASE) is compatible with various representations of the wave function, including real-space grids, plane waves, and linear combinations of atomic orbitals. The projector-augmented wave formalism for the representation of atomic core electrons is also supported. Results of calculations with the GPAW software are described here, but our implementation can also use output from other electronic structure software such as ABINIT, NWChem, and VASP.

Amp: A modular approach to machine learning in atomistic simulations

Electronic structure calculations, such as those employing Kohn–Sham density functional theory or ab initio wavefunction theories, have allowed for atomistic-level understandings of a wide variety of phenomena and properties of matter at small scales. However, the computational cost of electronic structure methods drastically increases with length and time scales, which makes these methods difficult for long time-scale molecular dynamics simulations or large-sized systems. Machine-learning techniques can provide accurate potentials that can match the quality of electronic structure calculations, provided sufficient training data. These potentials can then be used to rapidly simulate large and long time-scale phenomena at similar quality to the parent electronic structure approach. Machine-learning potentials usually take a bias-free mathematical form and can be readily developed for a wide variety of systems. Electronic structure calculations have favorable properties–namely that they are noiseless and targeted training data can be produced on-demand–that make them particularly well-suited for machine learning. This paper discusses our modular approach to atomistic machine learning through the development of the open-source Atomistic Machine-learning Package (Amp), which allows for representations of both the total and atom-centered potential energy surface, in both periodic and non-periodic systems. Potentials developed through the atom-centered approach are simultaneously applicable for systems with various sizes. Interpolation can be enhanced by introducing custom descriptors of the local environment. We demonstrate this in the current work for Gaussian-type, bispectrum, and Zernike-type descriptors. Amp has an intuitive and modular structure with an interface through the python scripting language yet has parallelizable fortran components for demanding tasks; it is designed to integrate closely with the widely used Atomic Simulation Environment (ASE), which makes it compatible with a wide variety of commercial and open-source electronic structure codes. We finally demonstrate that the neural network model inside Amp can accurately interpolate electronic structure energies as well as forces of thousands of multi-species atomic systems. Program summaryProgram title: AmpCatalogue identifier: AFAK_v1_0Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AFAK_v1_0.htmlProgram obtainable from: CPC Program Library, Queen’s University, Belfast, N. IrelandLicensing provisions: yesNo. of lines in distributed program, including test data, etc.: 21239No. of bytes in distributed program, including test data, etc.: 1412975Distribution format: tar.gzProgramming language: Python, Fortran.Computer: PC, Mac.Operating system: Linux, Mac, Windows.Has the code been vectorized or parallelized?: YesRAM: Variable, depending on the number and size of atomic systems.Classification: 16.1, 2.1.External routines: ASE, NumPy, SciPy, f2py, matplotlibNature of problem:Atomic interactions within many-body systems typically have complicated functional forms, difficult to represent in simple pre-decided closed-forms.Solution method:Machine learning provides flexible functional forms that can be improved as new situations are encountered. Typically, interatomic potentials yield from machine learning simultaneously apply to different system sizes.Unusual features:Amp is as modular as possible, providing a framework for the user to create atomic environment descriptor and regression model at will. Moreover, it has Atomic Simulation Environment (ASE) interface, facilitating interactive collaboration with other electronic structure calculators within ASE.Running time:Variable, depending on the number and size of atomic systems.

A universal preconditioner for simulating condensed phase materials.

We introduce a universal sparse preconditioner that accelerates geometry optimisation and saddle point search tasks that are common in the atomic scale simulation of materials. Our preconditioner is based on the neighbourhood structure and we demonstrate the gain in computational efficiency in a wide range of materials that include metals, insulators, and molecular solids. The simple structure of the preconditioner means that the gains can be realised in practice not only when using expensive electronic structure models but also for fast empirical potentials. Even for relatively small systems of a few hundred atoms, we observe speedups of a factor of two or more, and the gain grows with system size. An open source Python implementation within the Atomic Simulation Environment is available, offering interfaces to a wide range of atomistic codes.

An object oriented Python interface for atomistic simulations

Programmable simulation environments allow one to monitor and control calculations efficiently and automatically before, during, and after runtime. Environments directly accessible in a programming environment can be interfaced with powerful external analysis tools and extensions to enhance the functionality of the core program, and by incorporating a flexible object based structure, the environments make building and analysing computational setups intuitive. In this work, we present a classical atomistic force field with an interface written in Python language. The program is an extension for an existing object based atomistic simulation environment. Program summaryProgram title: PysicCatalogue identifier: AEYE_v1_0Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEYE_v1_0.htmlProgram obtainable from: CPC Program Library, Queen’s University, Belfast, N. Ireland.Licensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.htmlNo. of lines in distributed program, including test data, etc.: 74.743.No. of bytes in distributed program, including test data, etc.: 758.903.Distribution format: tar.gzProgramming language: Python, Fortran 90.Computer: Program has been tested on Linux and OS X workstations, and a Cray supercomputer.Operating system: Linux, Unix, OS X, Windows.RAM: Depends on the size of system.Classification: 7.7, 16.9, 4.14.External routines: Atomic Simulation Environment, NumPy necessary. Scipy, Matplotlib, HDF5, h5py recommended. The random number generator, Mersenne Twister, is included from the source: http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/VERSIONS/FORTRAN/mt95.f90Nature of problem: Automated simulation control, interaction tuning and an intuitive interface for running atomistic simulations.Solution method: Object oriented interface to a flexible classical potential.Additional comments:User guide:http://thynnine.github.io/pysic/Running time: Depends on the size of system.

Removing External Degrees of Freedom from Transition-State Search Methods using Quaternions.

In finite systems, such as nanoparticles and gas-phase molecules, calculations of minimum energy paths (MEPs) connecting initial and final states of transitions as well as searches for saddle points are complicated by the presence of external degrees of freedom, such as overall translation and rotation. A method based on quaternion algebra for removing the external degrees of freedom is described here and applied in calculations using two commonly used methods: the nudged elastic band (NEB) method for MEPs and the DIMER method for finding the minimum mode in minimum mode following searches of first-order saddle points. With the quaternion approach, fewer images in the NEB are needed to represent MEPs accurately. In both NEB and DIMER calculations of finite systems, the number of iterations required to reach convergence is significantly reduced. The algorithms have been implemented in the Atomic Simulation Environment (ASE) open source software.