A5021241296- Machine Learning in Materials Science
- Computational Drug Discovery Methods
- Spectroscopy and Quantum Chemical Studies
- Quantum, superfluid, helium dynamics
- X-ray Diffraction in Crystallography
- Advanced Chemical Physics Studies
- Protein Structure and Dynamics
- Crystallography and molecular interactions
- Crystallization and Solubility Studies
- Theoretical and Computational Physics
- Advanced NMR Techniques and Applications
- nanoparticles nucleation surface interactions
- Electron and X-Ray Spectroscopy Techniques
- Advanced Materials Characterization Techniques
- Various Chemistry Research Topics
- High-pressure geophysics and materials
- Mass Spectrometry Techniques and Applications
- Semiconductor materials and devices
- Atomic and Subatomic Physics Research
- Solid-state spectroscopy and crystallography
- Phase Equilibria and Thermodynamics
- Microstructure and mechanical properties
- Thermodynamic properties of mixtures
- Molecular spectroscopy and chirality
- Material Dynamics and Properties
École Polytechnique Fédérale de Lausanne
2016-2025
California Institute of Technology
2024
Instituto de Ciencia de Materiales de Sevilla
2016-2024
Collaborative Innovation Center of Chemistry for Energy Materials
2021
Xiamen University
2021
University of Cambridge
2021
Advanced Neural Dynamics (United States)
2021
Centrum Wiskunde & Informatica
2021
University of Applied Sciences and Arts of Southern Switzerland
2019
University College London
2019
Evaluating the (dis)similarity of crystalline, disordered and molecular compounds is a critical step in development algorithms to navigate automatically configuration space complex materials. For instance, structural similarity metric crucial for classifying structures, searching chemical better materials, driving next generation machine-learning techniques predicting stability properties molecules In last few years several strategies have been designed compare atomic coordination...
Determining the stability of molecules and condensed phases is cornerstone atomistic modelling, underpinning our understanding chemical materials properties transformations. Here we show that a machine learning model, based on local description environments Bayesian statistical learning, provides unified framework to predict atomic-scale properties. It captures quantum mechanical effects governing complex surface reconstructions silicon, predicts different classes with accuracy,...
A new scheme, sketch-map, for obtaining a low-dimensional representation of the region phase space explored during an enhanced dynamics simulation is proposed. We show evidence, from examination distribution pairwise distances between frames, that some features free-energy surface are inherently high-dimensional. This makes dimensionality reduction problematic because data does not satisfy assumptions made in conventional manifold learning algorithms therefore propose when performed on...
Machine learning of atomic-scale properties is revolutionizing molecular modelling, making it possible to evaluate inter-atomic potentials with first-principles accuracy, at a fraction the costs. The speed and reliability machine-learning potentials, however, depends strongly on way atomic configurations are represented, i.e. choice descriptors used as input for machine method. raw Cartesian coordinates typically transformed in "fingerprints", or "symmetry functions", that designed encode,...
Thermodynamic properties of liquid water as well hexagonal (Ih) and cubic (Ic) ice are predicted based on density functional theory at the hybrid-functional level, rigorously taking into account quantum nuclear motion, anharmonic fluctuations, proton disorder. This is made possible by combining advanced free-energy methods state-of-the-art machine-learning techniques. The ab initio description leads to structural in excellent agreement with experiments reliable estimates melting points light...
Quantum sieves for hydrogen isotopes One method improving the efficiency of separation from deuterium (D) is to exploit kinetic quantum sieving with nanoporous solids. This requires ultrafine pore apertures (around 3 angstroms), which usually leads low volumes and D 2 adsorption capacities. Liu et al. used organic synthesis tune size internal cavities cage molecules. A hybrid cocrystal contained both a small-pore that imparted high selectivity larger-pore enabled uptake. Science , this issue p. 613
The path integral molecular dynamics (PIMD) method provides a convenient way to compute the quantum mechanical structural and thermodynamic properties of condensed phase systems at expense introducing an additional set high frequency normal modes on top physical vibrations system. Efficiently sampling such wide range frequencies considerable thermostatting challenge. Here we introduce simple stochastic Langevin equation (PILE) thermostat which exploits analytic knowledge free mode...
Statistical learning methods show great promise in providing an accurate prediction of materials and molecular properties, while minimizing the need for computationally demanding electronic structure calculations. The accuracy transferability these models are increased significantly by encoding into procedure fundamental symmetries rotational permutational invariance scalar properties. However, tensorial properties requires that model respects appropriate geometric transformations, rather...
Two of the most successful methods that are presently available for simulating quantum dynamics condensed phase systems centroid molecular (CMD) and ring polymer (RPMD). Despite their conceptual differences, practical implementations these differ in just two respects: choice Parrinello-Rahman mass matrix whether or not a thermostat is applied to internal modes during dynamics. Here we explore method which halfway between approximations: keep path integral bead masses equal physical particle...
Significance There is no doubt about the importance of liquid water for climate and life on Earth. Correctly modeling properties this substance still a formidable challenge, however. Here, we show, using state-of-the-art techniques that allow quantum mechanical effects in motion electrons nuclei, room-temperature not simply molecular liquid; its protons experience wild excursions along hydrogen bond (HB) network driven by fluctuations, which result an unexpectedly large probability transient...
The electronic charge density plays a central role in determining the behavior of matter at atomic scale, but its computational evaluation requires demanding electronic-structure calculations. We introduce an atom-centered, symmetry-adapted framework to machine-learn valence based on small number reference model is highly transferable, meaning it can be trained data molecules and used predict larger compounds with low, linear-scaling cost. Applications are shown for various hydrocarbon...
Polymorphism is common in molecular crystals, whose energy landscapes usually contain many structures with similar stability, but very different physical–chemical properties. Machine-learning techniques can accelerate the evaluation of and properties by side-stepping accurate demanding electronic-structure calculations, provide a data-driven classification most important packing motifs.
The accurate representation of multidimensional potential energy surfaces is a necessary requirement for realistic computer simulations molecular systems. continued increase in power accompanied by advances correlated electronic structure methods nowadays enables routine calculations interaction energies small systems, which can then be used as references the development analytical functions (PEFs) rigorously derived from many-body (MB) expansions. Building on accuracy MB-pol PEF, we...
The molecular dipole polarizability describes the tendency of a molecule to change its moment in response an applied electric field. This quantity governs key intra- and intermolecular interactions, such as induction dispersion; plays vital role determining spectroscopic signatures molecules; is essential ingredient polarizable force fields. Compared with other ground-state properties, accurate prediction considerably more difficult, this quite sensitive underlying electronic structure...
Ions induce changes in the H-bond network of water that extend by >20 nm, vary for H 2 O and D O, lead to surface tension anomalies.
The most successful and popular machine learning models of atomic-scale properties derive their transferability from a locality ansatz. large molecule or bulk material are written as sum over contributions that depend on the configurations within finite atom-centered environments. obvious downside this approach is it cannot capture non-local, non-additive effects such those arising due to long-range electrostatics quantum interference. We propose solution problem by introducing non-local...