-
A dual-cutoff machine-learned potential for condensed organic systems obtained via uncertainty-guided active learning
Authors:
Leonid Kahle,
Benoit Minisini,
Tai Bui,
Jeremy T. First,
Corneliu Buda,
Thomas Goldman,
Erich Wimmer
Abstract:
Machine-learned potentials (MLPs) trained on ab initio data combine the computational efficiency of classical interatomic potentials with the accuracy and generality of the first-principles method used in the creation of the respective training set. In this work, we implement and train a MLP to obtain an accurate description of the potential energy surface and property predictions for organic comp…
▽ More
Machine-learned potentials (MLPs) trained on ab initio data combine the computational efficiency of classical interatomic potentials with the accuracy and generality of the first-principles method used in the creation of the respective training set. In this work, we implement and train a MLP to obtain an accurate description of the potential energy surface and property predictions for organic compounds, as both single molecules and in the condensed phase. We devise a dual descriptor, based on the atomic cluster expansion (ACE), that couples an information-rich short-range description with a coarser long-range description that captures weak intermolecular interactions. We employ uncertainty-guided active learning for the training set generation, creating a dataset that is comparatively small for the breadth of application and consists of alcohols, alkanes, and an adipate. Utilizing that MLP, we calculate densities of those systems of varying chain lengths as a function of temperature, obtaining a discrepancy of less than 4% compared with experiment. Vibrational frequencies calculated with the MLP have a root mean square error of less than 1 THz compared to DFT. The heat capacities of condensed systems are within 11% of experimental findings, which is strong evidence that the dual descriptor provides an accurate framework for the prediction of both short-range intramolecular and long-range intermolecular interactions.
△ Less
Submitted 6 August, 2024;
originally announced August 2024.
-
On the Quality of Uncertainty Estimates from Neural Network Potential Ensembles
Authors:
Leonid Kahle,
Federico Zipoli
Abstract:
Neural network potentials (NNPs) combine the computational efficiency of classical interatomic potentials with the high accuracy and flexibility of the ab initio methods used to create the training set, but can also result in unphysical predictions when employed outside their training set distribution. Estimating the epistemic uncertainty of an NNP is required in active learning or on-the-fly gene…
▽ More
Neural network potentials (NNPs) combine the computational efficiency of classical interatomic potentials with the high accuracy and flexibility of the ab initio methods used to create the training set, but can also result in unphysical predictions when employed outside their training set distribution. Estimating the epistemic uncertainty of an NNP is required in active learning or on-the-fly generation of potentials. Inspired from their use in other machine-learning applications, NNP ensembles have been used for uncertainty prediction in several studies, with the caveat that ensembles do not provide a rigorous Bayesian estimate of the uncertainty. To test whether NNP ensembles provide accurate uncertainty estimates, we train such ensembles in four different case studies, and compare the predicted uncertainty with the errors on out-of-distribution validation sets. Our results indicate that NNP ensembles are often overconfident, underestimating the uncertainty of the model, and require to be calibrated for each system and architecture. We also provide evidence that Bayesian NNPs, obtained by sampling the posterior distribution of the model parameters using Monte-Carlo techniques, can provide better uncertainty estimates.
△ Less
Submitted 21 January, 2022; v1 submitted 12 August, 2021;
originally announced August 2021.
-
Electrolytes for Li-ion all-solid-state batteries: a first-principles comparative study of Li10GeP2O12 and Li10GeP2S12 in the LISICON and LGPS phases
Authors:
Giuliana Materzanini,
Leonid Kahle,
Aris Marcolongo,
Nicola Marzari
Abstract:
In this work we address Li-ion diffusion in thio-LISICON materials and in their oxide counterparts, exploring both the orthorhombic and tetragonal phases of Li10GeP2S12(LGPS) and Li10GeP2O12(LGPO) through extended Car-Parrinello molecular dynamics in the canonical and isobaric-isothermal ensemble. The (quasi-)orthorhombic and tetragonal phases are studied both for the oxide and for the sulfide, wi…
▽ More
In this work we address Li-ion diffusion in thio-LISICON materials and in their oxide counterparts, exploring both the orthorhombic and tetragonal phases of Li10GeP2S12(LGPS) and Li10GeP2O12(LGPO) through extended Car-Parrinello molecular dynamics in the canonical and isobaric-isothermal ensemble. The (quasi-)orthorhombic and tetragonal phases are studied both for the oxide and for the sulfide, with the aim of comparing their conductivity with the same approach; out of these four case studies, tetragonal LGPO has not been reported and, while dynamically stable, it sits (0.04 Ha/formula unit) above orthorhombic LGPO. We calculate activation energies for diffusion of 0.18 eV and 0.23 eV for tetragonal and orthorhombic LGPS, and of 0.22 eV and 0.34eV for tetragonal and orthorhombic LGPO. In line with experiments, we find orthorhombic LGPO orders of magnitude less conductive, at room temperature, than the two sulfide systems. However, this is not the case for tetragonal LGPO, which, although less stable than its orthorhombic allotrope, shows at room temperature a conductivity comparable to orthorhombic and tetragonal LGPS, and, if synthesized, could make a very attractive Li-ion conductor.
△ Less
Submitted 15 October, 2020;
originally announced October 2020.
-
Li4-xGe1-xPxO4 a potential solid-state electrolyte for all-oxide microbatteries
Authors:
E. Gilardi,
G. Materzanini,
L. Kahle,
M. Doebeli,
S. Lacey,
X. Cheng,
N. Marzari,
D. Pergolesi,
A. Hintennach,
T. Lippert
Abstract:
Solid-state electrolytes for Li-ion batteries are attracting growing interest as they allow building safer batteries, also using lithium metal anodes. Here we studied a compound in the lithium superionic conductor (LISICON) family, i.e. Li4-xGe1-xPxO4 (LGPO). Thin films were deposited via pulsed laser deposition and their electrical properties were compared with ceramic pellets. A detailed charact…
▽ More
Solid-state electrolytes for Li-ion batteries are attracting growing interest as they allow building safer batteries, also using lithium metal anodes. Here we studied a compound in the lithium superionic conductor (LISICON) family, i.e. Li4-xGe1-xPxO4 (LGPO). Thin films were deposited via pulsed laser deposition and their electrical properties were compared with ceramic pellets. A detailed characterization of the micro structure shows that thin films can be deposited fully crystalline at higher temperatures but also partially amorphous at room temperature. The conductivity is not strongly influenced by the presence of grain boundaries, exposure to air or lithium deficiencies. First-principles molecular dynamics simulations were employed to calculate the lithium ion diffusion profile and the conductivity at various temperatures of the ideal LGPO crystal. Simulations gives the upper limit of conductivity for a defect free crystal, which is in the range of 10-2 S cm-1 at 300 deg. The ease of thin film fabrication, the room-temperature Li-ion conductivity in the range of a few microS cm-1 make LGPO a very appealing electrolyte material for thin film all-solid-state all-oxide microbatteries.
△ Less
Submitted 6 October, 2020;
originally announced October 2020.
-
AiiDA 1.0, a scalable computational infrastructure for automated reproducible workflows and data provenance
Authors:
Sebastiaan. P. Huber,
Spyros Zoupanos,
Martin Uhrin,
Leopold Talirz,
Leonid Kahle,
Rico Häuselmann,
Dominik Gresch,
Tiziano Müller,
Aliaksandr V. Yakutovich,
Casper W. Andersen,
Francisco F. Ramirez,
Carl S. Adorf,
Fernando Gargiulo,
Snehal Kumbhar,
Elsa Passaro,
Conrad Johnston,
Andrius Merkys,
Andrea Cepellotti,
Nicolas Mounet,
Nicola Marzari,
Boris Kozinsky,
Giovanni Pizzi
Abstract:
The ever-growing availability of computing power and the sustained development of advanced computational methods have contributed much to recent scientific progress. These developments present new challenges driven by the sheer amount of calculations and data to manage. Next-generation exascale supercomputers will harden these challenges, such that automated and scalable solutions become crucial.…
▽ More
The ever-growing availability of computing power and the sustained development of advanced computational methods have contributed much to recent scientific progress. These developments present new challenges driven by the sheer amount of calculations and data to manage. Next-generation exascale supercomputers will harden these challenges, such that automated and scalable solutions become crucial. In recent years, we have been developing AiiDA (http://www.aiida.net), a robust open-source high-throughput infrastructure addressing the challenges arising from the needs of automated workflow management and data provenance recording. Here, we introduce developments and capabilities required to reach sustained performance, with AiiDA supporting throughputs of tens of thousands processes/hour, while automatically preserving and storing the full data provenance in a relational database making it queryable and traversable, thus enabling high-performance data analytics. AiiDA's workflow language provides advanced automation, error handling features and a flexible plugin model to allow interfacing with any simulation software. The associated plugin registry enables seamless sharing of extensions, empowering a vibrant user community dedicated to making simulations more robust, user-friendly and reproducible.
△ Less
Submitted 24 March, 2020;
originally announced March 2020.
-
The solid-state Li-ion conductor Li$_7$TaO$_6$: A combined computational and experimental study
Authors:
Leonid Kahle,
Xi Cheng,
Tobias Binninger,
Steven David Lacey,
Aris Marcolongo,
Federico Zipoli,
Elisa Gilardi,
Claire Villevieille,
Mario El Kazzi,
Nicola Marzari,
Daniele Pergolesi
Abstract:
We study the oxo-hexametallate Li$_7$TaO$_6$ with first-principles and classical molecular dynamics simulations, obtaining a low activation barrier for diffusion of $\sim$0.29 eV and a high ionic conductivity of $5.7 \times 10^{-4}$ S cm$^{-1}$ at room temperature (300 K). We find evidence for a wide electrochemical stability window from both calculations and experiments, suggesting its viable use…
▽ More
We study the oxo-hexametallate Li$_7$TaO$_6$ with first-principles and classical molecular dynamics simulations, obtaining a low activation barrier for diffusion of $\sim$0.29 eV and a high ionic conductivity of $5.7 \times 10^{-4}$ S cm$^{-1}$ at room temperature (300 K). We find evidence for a wide electrochemical stability window from both calculations and experiments, suggesting its viable use as a solid-state electrolyte in next-generation solid-state Li-ion batteries. To assess its applicability in an electrochemical energy storage system, we performed electrochemical impedance spectroscopy measurements on multicrystalline pellets, finding substantial ionic conductivity, if below the values predicted from simulation. We further elucidate the relationship between synthesis conditions and the observed ionic conductivity using X-ray diffraction, inductively coupled plasma optical emission spectrometry, and X-ray photoelectron spectroscopy, and study the effects of Zr and Mo doping.
△ Less
Submitted 24 October, 2019;
originally announced October 2019.
-
High-throughput computational screening for solid-state Li-ion conductors
Authors:
Leonid Kahle,
Aris Marcolongo,
Nicola Marzari
Abstract:
We present a computational screening of experimental structural repositories for fast Li-ion conductors, with the goal of finding new candidate materials for application as solid-state electrolytes in next-generation batteries. We start from ~1400 unique Li-containing materials, of which ~900 are insulators at the level of density-functional theory. For those, we calculate the diffusion coefficien…
▽ More
We present a computational screening of experimental structural repositories for fast Li-ion conductors, with the goal of finding new candidate materials for application as solid-state electrolytes in next-generation batteries. We start from ~1400 unique Li-containing materials, of which ~900 are insulators at the level of density-functional theory. For those, we calculate the diffusion coefficient in a highly automated fashion, using extensive molecular dynamics simulations on a potential energy surface (the recently published pinball model) fitted on first-principles forces. The ~130 most promising candidates are studied with full first-principles molecular dynamics, first at high temperature and then more extensively for the 78 most promising candidates. The results of the first-principles simulations of the candidate solid-state electrolytes found are discussed in detail.
△ Less
Submitted 2 September, 2019;
originally announced September 2019.
-
Unsupervised landmark analysis for jump detection in molecular dynamics simulations
Authors:
Leonid Kahle,
Albert Musaelian,
Nicola Marzari,
Boris Kozinsky
Abstract:
Molecular dynamics is a versatile and powerful method to study diffusion in solid-state ionic conductors, requiring minimal prior knowledge of equilibrium or transition states of the system's free energy surface. However, the analysis of trajectories for relevant but rare events, such as a jump of the diffusing mobile ion, is still rather cumbersome, requiring prior knowledge of the diffusive proc…
▽ More
Molecular dynamics is a versatile and powerful method to study diffusion in solid-state ionic conductors, requiring minimal prior knowledge of equilibrium or transition states of the system's free energy surface. However, the analysis of trajectories for relevant but rare events, such as a jump of the diffusing mobile ion, is still rather cumbersome, requiring prior knowledge of the diffusive process in order to get meaningful results. In this work, we present a novel approach to detect the relevant events in a diffusive system without assuming prior information regarding the underlying process. We start from a projection of the atomic coordinates into a landmark basis to identify the dominant features in a mobile ion's environment. Subsequent clustering in landmark space enables a discretization of any trajectory into a sequence of distinct states. As a final step, the use of the smooth overlap of atomic positions descriptor allows distinguishing between different environments in a straightforward way. We apply this algorithm to ten Li-ionic systems and conduct in-depth analyses of cubic Li$_{7}$La$_{3}$Zr$_{2}$O$_{12}$, tetragonal Li$_{10}$GeP$_{2}$S$_{12}$, and the $β$-eucryptite LiAlSiO$_{4}$. We compare our results to existing methods, underscoring strong points, weaknesses, and insights into the diffusive behavior of the ionic conduction in the materials investigated.
△ Less
Submitted 21 May, 2019; v1 submitted 6 February, 2019;
originally announced February 2019.
-
Modeling lithium-ion solid-state electrolytes with a pinball model
Authors:
Leonid Kahle,
Aris Marcolongo,
Nicola Marzari
Abstract:
We introduce a simple and efficient model to describe the potential energy surface of lithium diffusing in a solid-state ionic conductor. First, we assume that the Li atoms are fully ionized and we neglect the weak dependence of the electronic valence charge density on the instantaneous position of the Li ions. Second, we freeze the atoms of the host lattice in their equilibrium positions; consequ…
▽ More
We introduce a simple and efficient model to describe the potential energy surface of lithium diffusing in a solid-state ionic conductor. First, we assume that the Li atoms are fully ionized and we neglect the weak dependence of the electronic valence charge density on the instantaneous position of the Li ions. Second, we freeze the atoms of the host lattice in their equilibrium positions; consequently, also the valence charge density is frozen. We thus obtain a computational setup (the "pinball model") for which extremely inexpensive molecular dynamics simulation can be performed. To assess the accuracy of the model, we contrast it with full first-principles molecular dynamics simulations performed either with a free or frozen host lattice; in this latter case, the charge density still readjusts itself self-consistently to the actual positions of the diffusing Li ions. We show that the pinball model is able to reproduce accurately the static and dynamic properties of the diffusing Li ions - including forces, power spectra, and diffusion coefficients - when compared to the self-consistent frozen-host lattice simulations. The frozen-lattice approximation itself is often accurate enough, and certainly a good proxy in most materials. These observations unlock efficient ways to simulating the diffusion of lithium in the solid state, and provide additional physical insight into the respective roles of charge-density rearrangements or lattice vibrations in affecting lithium diffusion.
△ Less
Submitted 25 May, 2018;
originally announced May 2018.
