Quantifying Chemical Structure and Machine-Learned Atomic Energies in Amorphous and Liquid Silicon (original) (raw)
Related papers
MRS Advances
ABSTRACTWe present an information-based total-energy optimization method to produce nearly defect-free structural models of amorphous silicon. Using geometrical, structural, and topological information from disordered tetrahedral networks, we have shown that it is possible to generate structural configurations of amorphous silicon, which are superior than the models obtained from conventional reverse Monte Carlo and molecular dynamics simulations. The new data-driven hybrid approach presented here is capable of producing atomistic models with structural and electronic properties which are on a par with those obtained from the modified Wooten-Winer-Weaire (WWW) models of amorphous silicon. Structural, electronic, and thermodynamic properties of the hybrid models are compared with the best dynamical models obtained from using machine-intelligence-based algorithms and efficient classical molecular dynamics simulations, reported in the recent literature. We have shown that, together wit...
A machine-learned interatomic potential for silica and its relation to empirical models
npj Computational Materials
Silica (SiO2) is an abundant material with a wide range of applications. Despite much progress, the atomistic modelling of the different forms of silica has remained a challenge. Here we show that by combining density-functional theory at the SCAN functional level with machine-learning-based interatomic potential fitting, a range of condensed phases of silica can be accurately described. We present a Gaussian approximation potential model that achieves high accuracy for the thermodynamic properties of the crystalline phases, and we compare its performance (and performance–cost trade-off) with that of multiple empirically fitted interatomic potentials for silica. We also include amorphous phases, assessing the ability of the potentials to describe structures of melt-quenched glassy silica, their energetic stability, and the high-pressure structural transition to a mainly sixfold-coordinated phase. We suggest that rather than standing on their own, machine-learned potentials for silic...
Scientific Reports
X-ray diffraction, Amorphous silicon, Multi-objective optimization, Monte Carlo methods. This paper addresses a difficult inverse problem that involves the reconstruction of a three-dimensional model of tetrahedral amorphous semiconductors via inversion of diffraction data. By posing the material-structure determination as a multiobjective optimization program, it has been shown that the problem can be solved accurately using a few structural constraints, but no total-energy functionals/forces, which describe the local chemistry of amorphous networks. The approach yields highly realistic models of amorphous silicon, with no or only a few coordination defects (≤1%), a narrow bond-angle distribution of width 9–11.5°, and an electronic gap of 0.8–1.4 eV. These data-driven information-based models have been found to produce electronic and vibrational properties of a-Si that match accurately with experimental data and rival that of the Wooten-Winer-Weaire models. The study confirms the e...
Information-driven inverse approach to disordered solids: Applications to amorphous silicon
Physical Review Materials
Diffraction data play an important role in the structural characterizations of solids. While reverse Monte Carlo (RMC) and similar methods provide an elegant approach to (re)construct a three-dimensional model of non-crystalline solids, a satisfactory solution to the RMC problem is still not available. Following our earlier efforts, we present here an accurate structural solution of the inverse problem by developing a new informationdriven inverse approach (INDIA). The efficacy of the approach is illustrated by choosing amorphous silicon as an example, which is particularly difficult to model using total-energy-based relaxation methods. We demonstrate that, by introducing a subspace optimization technique that sequentially optimizes two objective functions (involving experimental diffraction data, a total-energy functional, and a few geometric constraints), it is possible to produce models of amorphous silicon with very little or no coordination defects and a pristine gap around the Fermi level in the electronic spectrum. The structural, electronic, and vibrational properties of the resulting INDIA models are shown to be fully compliant with experimental data from X-ray diffraction, Raman spectroscopy, differential scanning calorimetry, and inelastic neutron scattering measurements. A direct comparison of the models with those obtained from the Wooten-Winer-Weaire (WWW) approach and from recent high-quality molecular-dynamics simulations is also presented.
Large and realistic models of amorphous silicon
Journal of Non-Crystalline Solids
Amorphous silicon (a-Si) models are analyzed for structural, electronic and vibrational characteristics. Several models of various sizes have been computationally fabricated for this analysis. It is shown that a recently developed structural modeling algorithm known as force-enhanced atomic refinement (FEAR) provides results in agreement with experimental neutron and x-ray diffraction data while producing a total energy below conventional schemes. We also show that a large model (∼ 500 atoms) and a complete basis is necessary to properly describe vibrational and thermal properties. We compute the density for a-Si, and compare with experimental results.
Journal of Applied Physics, 2022
Machine Learning (ML) techniques are revolutionizing the way to perform efficient materials modeling. Nevertheless, not all the ML approaches allow for the understanding of microscopic mechanisms at play in different phenomena. To address the latter aspect, we propose a combinatorial machine-learning approach to obtain physical formulas based on simple and easily-accessible ingredients, such as atomic properties. The latter are used to build materials features that are finally employed, through Linear Regression, to predict the energetic stability of semiconducting binary compounds with respect to zincblende and rocksalt crystal structures. The adopted models are trained using dataset built from firstprinciples calculations. Our results show that already one-dimensional (1D) formulas well describe the energetics; a simple grid-search optimization of the automatically-obtained 1D-formulas enhances the prediction performances at a very small computational cost. In addition, our approach allows to highlight the role of the different atomic properties involved in the formulas. The computed formulas clearly indicate that "spatial" atomic properties (i.e. radii indicating maximum probability densities for s, p, d electronic shells) drive the stabilization of one crystal structure with respect to the other, suggesting the major relevance of the radius associated to the p-shell of the cation species.
Network structure and dynamics of hydrogenated amorphous silicon
Journal of Non-Crystalline Solids, 2008
In this paper we discuss the application of current ab initio computer simulation techniques to hydrogenated amorphous silicon (a-Si:H). We begin by discussing thermal fluctuation in the number of coordination defects in the material, and its temperature dependence. We connect this to the "fluctuating bond center detachment" mechanism for liberating H bonded to Si atoms. Next, from extended thermal MD simulation, we illustrate various mechanisms of H motion. The dynamics of the lattice is then linked to the electrons, and we point out that the squared electronlattice coupling (and the thermally-induced mean square variation in electron energy eigenvalues) is robustly proportional to the localization of the conjugate state, if localization is measured with inverse participation ratio. Finally we discuss the Staebler-Wronski effect using these methods, and argue that a sophisticated local heating picture (based upon reasonable calculations of the electron-lattice coupling and molecular dynamic simulation) explains significant aspects of the phenomenon.
Materials Today Physics, 2019
First-principles based modeling on phonon dynamics and transport using density functional theory and Boltzmann transport equation has proven powerful in predicting thermal conductivity of crystalline materials, but it remains unfeasible for modeling complex crystals and disordered solids due to the prohibitive computational cost to capture the disordered structure, especially when the quasiparticle "phonon" model breaks down. Recently, machine-learning regression algorithms show great promises for building high-accuracy potential fields for atomistic modeling with length and time scales far beyond those achievable by first-principles calculations. In this work, using both crystalline and amorphous silicon as examples, we develop machine learning based potential fields for predicting thermal conductivity. The machine learning based interatomic
Accelerated Prediction of Atomically Precise Cluster Structures Using On-the-fly Machine Learning
2021
The chemical and structural properties of atomically precise nanoclusters are of great interest in numerous applications, but predicting the stable structures of clusters can be computationally expensive. In this work, we present a procedure for rapidly predicting low-energy structures of nanoclusters by combining a genetic algorithm with interatomic potentials actively learned on-the-fly. Applying this approach to aluminum clusters with 21 to 55 atoms, we have identified structures with lower energy than any reported in the literature for 25 out of the 35 sizes. Our benchmarks indicate that the active learning procedure accelerated the average search speed by more than an order of magnitude relative to genetic algorithm searches using only density functional calculations. This work demonstrates a feasible way to systematically discover stable structures for large nanoclusters and provides insights into the transferability of machine-learned interatomic potentials for nanoclusters.
Simple machine-learned interatomic potentials for complex alloys
Physical Review Materials
Developing data-driven machine-learning interatomic potentials for materials containing many elements becomes increasingly challenging due to the vast configuration space that must be sampled by the training data. We study the learning rates and achievable accuracy of machine-learning interatomic potentials for many-element alloys with different combinations of descriptors for the local atomic environments. We show that for a five-element alloy system, potentials using simple low-dimensional descriptors can reach meV/atom-accuracy with modestly sized training datasets, significantly outperforming the high-dimensional SOAP descriptor in data efficiency, accuracy, and speed. In particular, we develop a computationally fast machine-learned and tabulated Gaussian approximation potential (tabGAP) for Mo-Nb-Ta-V-W alloys with a combination of two-body, three-body, and a new simple scalar many-body density descriptor based on the embedded atom method.