-
Machine learning assisted canonical sampling (MLACS)
Authors:
Aloïs Castellano,
Romuald Béjaud,
Pauline Richard,
Olivier Nadeau,
Clément Duval,
Grégory Geneste,
Gabriel Antonius,
Johann Bouchet,
Antoine Levitt,
Gabriel Stoltz,
François Bottin
Abstract:
The acceleration of material property calculations while maintaining ab initio accuracy (1 meV/atom) is one of the major challenges in computational physics. In this paper, we introduce a Python package enhancing the computation of (finite temperature) material properties at the ab initio level using machine learning interatomic potentials (MLIP). The Machine-Learning Assisted Canonical Sampling (…
▽ More
The acceleration of material property calculations while maintaining ab initio accuracy (1 meV/atom) is one of the major challenges in computational physics. In this paper, we introduce a Python package enhancing the computation of (finite temperature) material properties at the ab initio level using machine learning interatomic potentials (MLIP). The Machine-Learning Assisted Canonical Sampling (MLACS) method, grounded in a self-consistent variational approach, iteratively trains a MLIP using an active learning strategy in order to significantly reduce the computational cost of ab initio simulations.
MLACS offers a modular and user-friendly interface that seamlessly integrates Density Functional Theory (DFT) codes, MLIP potentials, and molecular dynamics packages, enabling a wide range of applications, while maintaining a near-DFT accuracy. These include sampling the canonical ensemble of a system, performing free energy calculations, transition path sampling, and geometry optimization, all by utilizing surrogate MLIP potentials, in place of ab initio calculations.
This paper provides a comprehensive overview of the theoretical foundations and implementation of the MLACS method. We also demonstrate its accuracy and efficiency through various examples, showcasing the capabilities of the MLACS package.
△ Less
Submitted 19 December, 2024;
originally announced December 2024.
-
Ab initio Canonical Sampling based on Variational Inference
Authors:
Alois Castellano,
Francois Bottin,
Johann Bouchet,
Antoine Levitt,
Gabriel Stoltz
Abstract:
Finite temperature calculations, based on ab initio molecular dynamics (AIMD) simulations, are a powerful tool able to predict material properties that cannot be deduced from ground state calculations. However, the high computational cost of AIMD limits its applicability for large or complex systems. To circumvent this limitation we introduce a new method named Machine Learning Assisted Canonical…
▽ More
Finite temperature calculations, based on ab initio molecular dynamics (AIMD) simulations, are a powerful tool able to predict material properties that cannot be deduced from ground state calculations. However, the high computational cost of AIMD limits its applicability for large or complex systems. To circumvent this limitation we introduce a new method named Machine Learning Assisted Canonical Sampling (MLACS), which accelerates the sampling of the Born--Oppenheimer potential surface in the canonical ensemble. Based on a self-consistent variational procedure, the method iteratively trains a Machine Learning Interatomic Potential to generate configurations that approximate the canonical distribution of positions associated with the ab initio potential energy. By proving the reliability of the method on anharmonic systems, we show that the method is able to reproduce the results of AIMD with an ab initio accuracy at a fraction of its computational cost.
△ Less
Submitted 20 October, 2022;
originally announced October 2022.
-
Tin (Sn) at high pressure: review, X-ray diffraction, DFT calculations, and Gibbs energy modeling
Authors:
Guillaume Deffrennes,
Philippe Faure,
François Bottin,
Jean-Marc Joubert,
Benoit Oudot
Abstract:
An assessment of the Sn unary system is presented. First, the literature on phase equilibria, the thermodynamic properties, the volume and related properties, and shock compression of tin is thoroughly reviewed. Second, the Sn system is investigated by means of synchrotron X-ray diffraction in a diamond-anvil cell up to pressures and temperatures of 57 GPa and 730 K. New information is obtained on…
▽ More
An assessment of the Sn unary system is presented. First, the literature on phase equilibria, the thermodynamic properties, the volume and related properties, and shock compression of tin is thoroughly reviewed. Second, the Sn system is investigated by means of synchrotron X-ray diffraction in a diamond-anvil cell up to pressures and temperatures of 57 GPa and 730 K. New information is obtained on the thermal stability and thermal expansion coefficient of the γ (I4/mmm) and γ" (Im-3 m) phases. Third, density functional theory calculations are conducted on the six allotropic phases of tin observed in experiments using both a local density approximation (LDA) and a generalized gradient approximation (GGA) functional. This combined experimental and theoretical investigation provides further insights on the pronounced metastable nature of Sn in the 30 - 70 GPa range. Last, a Gibbs energy modeling is conducted using the recently proposed Joubert-Lu-Grover model which is compatible with the CALPHAD method. Special emphasis is placed on discussing extrapolations to high pressures and temperatures of the volume and of the thermodynamic properties. While the description of the heat capacity is approximate at moderate pressure, all available data are closely reproduced up to 2500 K, which is 5 times higher than the atmospheric pressure melting point of tin, and 150 GPa, which is almost 3 times the standard bulk modulus of \b{eta}-Sn.
△ Less
Submitted 23 June, 2022; v1 submitted 30 March, 2022;
originally announced March 2022.
-
Tuneable Correlated Disorder in Alloys
Authors:
D. Chaney,
A. Castellano,
A. Bosak,
J. Bouchet,
F. Bottin,
B. Dorado,
L. Paolasini,
S. Rennie,
C. Bell,
R. Springell,
G. H. Lander
Abstract:
Understanding the role of disorder and the correlations that exist within it, is one of the defining challenges in contemporary materials science. However, there are few material systems, devoid of other complex interactions, which can be used to systematically study the effects of crystallographic conflict on correlated disorder. Here, we report extensive diffuse x-ray scattering studies on the e…
▽ More
Understanding the role of disorder and the correlations that exist within it, is one of the defining challenges in contemporary materials science. However, there are few material systems, devoid of other complex interactions, which can be used to systematically study the effects of crystallographic conflict on correlated disorder. Here, we report extensive diffuse x-ray scattering studies on the epitaxially stabilised alloy $\mbox{U}_{1-x}\mbox{Mo}_x$, showing that a new form of intrinsically tuneable correlated disorder arises from a mismatch between the preferred symmetry of a crystallographic basis and the lattice upon which it is arranged. Furthermore, combining grazing incidence inelastic x-ray scattering and state-of-the-art ab initio molecular dynamics simulations we discover strong disorder-phonon coupling. This breaks global symmetry and dramatically suppresses phonon-lifetimes compared to alloying alone, providing an additional design strategy for phonon engineering. These findings have implications wherever crystallographic conflict can be accommodated and may be exploited in the development of future functional materials.
△ Less
Submitted 12 November, 2020; v1 submitted 7 September, 2020;
originally announced September 2020.
-
Hydrogen diffusion in the proton conductor Gd-doped barium cerate
Authors:
Jessica Hermet,
Marc Torrent,
François Bottin,
Guilhem Dezanneau,
Gregory Geneste
Abstract:
The energy landscape and diffusion barriers of protonic defects in Gd-doped BaCeO3, a compound candidate as electrolyte for protonic ceramic fuel cells, have been investigated by density functional theory calculations, starting from a previously computed energy landscape consisting of 16 kinds of stable sites (8 close to dopants and 8 far from them). The simplified string method has been used to d…
▽ More
The energy landscape and diffusion barriers of protonic defects in Gd-doped BaCeO3, a compound candidate as electrolyte for protonic ceramic fuel cells, have been investigated by density functional theory calculations, starting from a previously computed energy landscape consisting of 16 kinds of stable sites (8 close to dopants and 8 far from them). The simplified string method has been used to determine accurately the Minimum Energy Paths between those sites, that might imply either proton reorientations, intra-octahedral or inter-octahedral hopping mechanisms. At contrast with simple cubic perovskites such as barium stannate or barium zirconate, very different values for energy barriers (from 0.02 eV to 0.58 eV) are found in this highly distorted orthorhombic perovskite, and no specific process appears to be clearly rate-limiting. Some inter-octahedral hoppings (when possible) are found to be more favourable than the intra-octahedral ones, while reorientations exhibit a wide range of energy barriers.
△ Less
Submitted 1 January, 2013;
originally announced January 2013.
-
Strong Isotopic Effect in Phase II of Dense Solid Hydrogen and Deuterium
Authors:
Grégory Geneste,
Marc Torrent,
François Bottin,
Paul Loubeyre
Abstract:
Quantum nuclear zero-point motions in solid H$_2$ and D$_2$ under pressure are investigated at 80 K up to 160 GPa by first-principles path-integral molecular dynamics calculations. Molecular orientations are well-defined in phase II of D$_2$, while solid H$_2$ exhibits large and very asymmetric angular quantum fluctuations in this phase, with possible rotation in the (bc) plane, making it difficul…
▽ More
Quantum nuclear zero-point motions in solid H$_2$ and D$_2$ under pressure are investigated at 80 K up to 160 GPa by first-principles path-integral molecular dynamics calculations. Molecular orientations are well-defined in phase II of D$_2$, while solid H$_2$ exhibits large and very asymmetric angular quantum fluctuations in this phase, with possible rotation in the (bc) plane, making it difficult to associate a well-identified single classical structure. The mechanism for the transition to phase III is also described. Existing structural data support this microscopic interpretation.
△ Less
Submitted 17 July, 2012; v1 submitted 10 May, 2012;
originally announced May 2012.
-
Large scale ab initio calculations based on three levels of parallelization
Authors:
François Bottin,
Stéphane Leroux,
Andrew Knyazev,
Gilles Zérah
Abstract:
We suggest and implement a parallelization scheme based on an efficient multiband eigenvalue solver, called the locally optimal block preconditioned conjugate gradient LOBPCG method, and using an optimized three-dimensional (3D) fast Fourier transform (FFT) in the ab initio}plane-wave code ABINIT. In addition to the standard data partitioning over processors corresponding to different k-points,…
▽ More
We suggest and implement a parallelization scheme based on an efficient multiband eigenvalue solver, called the locally optimal block preconditioned conjugate gradient LOBPCG method, and using an optimized three-dimensional (3D) fast Fourier transform (FFT) in the ab initio}plane-wave code ABINIT. In addition to the standard data partitioning over processors corresponding to different k-points, we introduce data partitioning with respect to blocks of bands as well as spatial partitioning in the Fourier space of coefficients over the plane waves basis set used in ABINIT. This k-points-multiband-FFT parallelization avoids any collective communications on the whole set of processors relying instead on one-dimensional communications only. For a single k-point, super-linear scaling is achieved for up to 100 processors due to an extensive use of hardware optimized BLAS, LAPACK, and SCALAPACK routines, mainly in the LOBPCG routine. We observe good performance up to 200 processors. With 10 k-points our three-way data partitioning results in linear scaling up to 1000 processors for a practical system used for testing.
△ Less
Submitted 23 July, 2007;
originally announced July 2007.
-
Stability and electronic structure of the $(1\times 1)$ SrTiO$_3$(110) polar surfaces by first principles calculations
Authors:
Francois Bottin,
Claudine Noguera,
Fabio Finocchi
Abstract:
The electronic and atomic structure of several $(1\times 1)$ terminations of the (110) polar orientation of SrTiO$_3$ surface are systematically studied by first-principles calculations. The electronic structure of the two stoichiometric SrTiO- and O$_2$-terminations are characterized by marked differences with respect to the bulk, as a consequence of the polarity compensation. In the former, th…
▽ More
The electronic and atomic structure of several $(1\times 1)$ terminations of the (110) polar orientation of SrTiO$_3$ surface are systematically studied by first-principles calculations. The electronic structure of the two stoichiometric SrTiO- and O$_2$-terminations are characterized by marked differences with respect to the bulk, as a consequence of the polarity compensation. In the former, the Fermi level is located at the bottom of the conduction band, while in the latter the formation of a peroxo bond between the two surface oxygens results in a small-gap insulating surface with states in the gap of the bulk projected band structure. We also consider three non stoichiometric terminations with TiO, Sr and O compositions, respectively, in the outermost atomic layer, which automatically allows the surface to be free from any macroscopic polarization. They are all insulating. The cleavage and surface energies of the five terminations are computed and compared, taking into account the influence of the chemical environment as a function of the relative richness in O and Sr. From our calculations it appears that some (110) faces can even compete with the TiO$_2$ and SrO terminations of the (100) cleavage surface: in particular, the (110)-TiO termination is stable in Sr-poor conditions, the (110)-Sr one in simultaneously O- and Sr-rich environments. The available experimental data are compared to the outcomes of our calculations and discussed.
△ Less
Submitted 10 April, 2003;
originally announced April 2003.
