A5079309376- Machine Learning in Materials Science
- Computational Drug Discovery Methods
- X-ray Diffraction in Crystallography
- Magnetic confinement fusion research
- Protein Structure and Dynamics
- Advanced NMR Techniques and Applications
- Spectroscopy and Quantum Chemical Studies
- Ionosphere and magnetosphere dynamics
- Quantum, superfluid, helium dynamics
- Metabolomics and Mass Spectrometry Studies
- Fusion materials and technologies
- Plant Surface Properties and Treatments
- History and advancements in chemistry
- Various Chemistry Research Topics
- Solar and Space Plasma Dynamics
- Block Copolymer Self-Assembly
- Perovskite Materials and Applications
- Crystallography and molecular interactions
- Corrosion Behavior and Inhibition
- Molecular spectroscopy and chirality
- Quantum Computing Algorithms and Architecture
- ZnO doping and properties
- Electronic and Structural Properties of Oxides
- Neural Networks and Applications
- Solid-state spectroscopy and crystallography
Freie Universität Berlin
2022-2025
École Polytechnique Fédérale de Lausanne
2016-2021
University of Lausanne
2018
Instituto de Ciencia de Materiales de Sevilla
2018
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.
We present a scheme to obtain an inexpensive and reliable estimate of the uncertainty associated with predictions machine-learning model atomic molecular properties. The is based on resampling, multiple models being generated subsampling same training data. accuracy prediction can be benchmarked by maximum likelihood estimation, which also used correct for correlations between resampled improve performance estimation cross-validation procedure. In case sparse Gaussian Process Regression...
The calculation of chemical shifts in solids has enabled methods to determine crystal structures powders. dependence on local atomic environments sets them among the most powerful tools for structure elucidation powdered or amorphous materials. Unfortunately, this dependency comes with cost high accuracy first-principle calculations qualitatively predict solids. Machine learning have recently emerged as a way overcome need explicit calculations. However, vast and combinatorial space spanned...
The applications of machine learning techniques to chemistry and materials science become more numerous by the day. main challenge is devise representations atomic systems that are at same time complete concise, so as reduce number reference calculations needed predict properties different types reliably. This has led a proliferation alternative ways convert an structure into input for machine-learning model. We introduce abstract definition chemical environments based on smoothed density,...
Machine-learning of atomic-scale properties amounts to extracting correlations between structure, composition and the quantity that one wants predict. Representing input structure in a way best reflects such makes it possible improve accuracy model for given amount reference data. When using description structures is transparent well-principled, optimizing representation might reveal insights into chemistry data set. Here we show how can generalize SOAP kernel introduce distance-dependent...
Physically motivated and mathematically robust atom-centered representations of molecular structures are key to the success modern atomistic machine learning. They lie at foundation a wide range methods predict properties both materials molecules explore visualize their chemical compositions. Recently, it has become clear that many most effective share fundamental formal connection. can all be expressed as discretization n-body correlation functions local atom density, suggesting opportunity...
Magnesium exhibits a high potential for variety of applications in areas such as transport, energy and medicine. However, untreated magnesium alloys are prone to corrosion, restricting their practical application. Therefore, it is necessary develop new approaches that can prevent or control corrosion degradation processes order adapt the specific needs One solution using inhibitors which capable drastically reducing rate result interactions with metal surface components corrosive medium. As...
The vibrational spectra of condensed and gas-phase systems are influenced by thequantum-mechanical behavior light nuclei. Full-dimensional simulations approximate quantum dynamics possible thanks to the imaginary time path-integral (PI) formulation statistical mechanics, albeit at a high computational cost which increases sharply with decreasing temperature. By leveraging advances in machine-learned coarse-graining, we develop PI method reduced classical simulation. We also propose simple...
High-throughput computational materials design promises to greatly accelerate the process of discovering new and compounds, optimizing their properties. The large databases structures properties that result from searches, as well agglomeration data heterogeneous provenance leads considerable challenges when it comes navigating database, representing its structure at a glance, understanding structure-property relations, eliminating duplicates identifying inconsistencies. Here we present case...
The input of almost every machine learning algorithm targeting the properties matter at atomic scale involves a transformation list Cartesian coordinates into more symmetric representation. Many most popular representations can be seen as an expansion symmetrized correlations atom density and differ mainly by choice basis. Considerable effort has been dedicated to optimization basis set, typically driven heuristic considerations on behavior regression target. Here, we take different,...
To investigate blob properties in the tokamak scrape-off layer (SOL), we perform dedicated numerical nonlinear simulations of plasma turbulence SOL a TCV discharge using Global Braginskii Solver code. A detection technique is used for first time three-dimensional (3D) full-turbulence simulation to track motion filaments SOL. The specific size, density amplitude and radial velocity blobs are computed, with typical values being , respectively. analysis structure parallel direction shows that...
Metal halide perovskite-based semi-conducting hetero-structures have emerged as promising electronics for solar cells, light-emitting diodes, detectors, and photo-catalysts. Perovskites' efficiency, electronic properties their long-term stability directly depend on morphology [1-24]. Therefore, to manufacture stable higher efficiency perovskite cells electronics, it is now crucial understand micro-structure evolution. In this study, we perform molecular dynamics simulations investigate the...
Statistical learning algorithms are finding more and applications in science technology. Atomic-scale modeling is no exception, with machine becoming commonplace as a tool to predict energy, forces properties of molecules condensed-phase systems. This short review summarizes recent progress the field, focusing particular on problem representing an atomic configuration mathematically robust computationally efficient way. We also discuss some regression that have been used construct surrogate...
Abstract The most popular and universally predictive protein simulation models employ all-atom molecular dynamics (MD), but they come at extreme computational cost. development of a universal, computationally efficient coarse-grained (CG) model with similar prediction performance has been long-standing challenge. By combining recent deep learning methods large diverse training set simulations, we here develop bottom-up CG force field chemical transferability, which can be used for...
The contribution of nuclear quantum effects (NQEs) to the properties various hydrogen-bound systems, including biomolecules, is increasingly recognized. Despite development many acceleration techniques, computational overhead incorporating NQEs in complex systems sizable, particularly at low temperatures. In this work, we leverage deep learning and multiscale coarse-graining techniques mitigate burden path integral molecular dynamics (PIMD). particular, employ a machine-learned potential...
The most popular and universally predictive protein simulation models employ all-atom molecular dynamics (MD), but they come at extreme computational cost. development of a universal, computationally efficient coarse-grained (CG) model with similar prediction performance has been long-standing challenge. By combining recent deep learning methods large diverse training set simulations, we here develop bottom-up CG force field chemical transferability, which can be used for extrapolative on...
The contribution of nuclear quantum effects (NQEs) to the properties various hydrogen-bound systems, including biomolecules, is increasingly recognized. Despite development many acceleration techniques, computational overhead incorporating NQEs in complex systems sizable, particularly at low temperatures. In this work, we leverage deep learning and multiscale coarse-graining techniques mitigate burden path integral molecular dynamics (PIMD). Specifically, employ a machine-learned potential...
High-throughput computational materials design promises to greatly accelerate the process of discovering new and compounds, optimizing their properties. The large databases structures properties that result from searches, as well agglomeration data heterogeneous provenance leads considerable challenges when it comes navigating database, representing its structure at a glance, understanding structure-property relations, eliminating duplicates identifying inconsistencies. Here we present case...