Research

Atomistic Simulations and Artificial Intelligence for Materials Design

Our approach —
atomistic simulations × artificial intelligence

A closed loop between physics and data.

Atomistic SimulationsDFT · MD · microkineticsArtificial IntelligenceMLIPs · GNNs · LLMs · active learningenergies · forces · descriptorscandidates · potentials · predictionsmillions of candidate compositionsa handful of promising materials

Our research starts from atomistic multi-scale simulations. We use density functional theory (DFT) and molecular dynamics to resolve how materials work at the level of electrons and atoms. Adsorption, charge transfer, diffusion, and bond breaking are computed on realistic surfaces, and grand-canonical DFT (GC-DFT) with microkinetic modeling carries them to reaction rates at working potentials.

Artificial intelligence extends what physics alone cannot reach. Machine-learned interatomic potentials bring near-DFT accuracy to large-scale dynamics. Graph neural networks and descriptor models predict properties across wide chemical spaces. Large language models read the literature, gathering experimental data and synthesis conditions at scale. Active learning and Bayesian optimization steer expensive calculations toward the most informative candidates.

Together they form a closed design loop: simulate, learn, propose, and validate. This loop turns millions of candidate compositions into a handful of the most promising materials, before a single experiment is run.

Research map

Where our research connects

Green Hydrogen — 21 papers21Green HydrogenChemical Conversion — 20 papers20Chemical ConversionEnergy Storage — 12 papers12Energy StorageSemiconductors — 20 papers20SemiconductorsMethodology — 8 papers8MethodologySilicon & SiO2 — 9 papers9Silicon & SiO2Metals & alloys — 14 papers14Metals & alloysNanoparticles & clusters — 10 papers10Nanoparticles & clustersCarbon materials — 20 papers20Carbon materialsSingle-atom catalysts — 7 papers7Single-atom catalystsNoble metals — 12 papers12Noble metalsOxides & nitrides — 15 papers15Oxides & nitridesPerovskites — 3 papers3PerovskitesTM hydroxides, phosphides & sulfides — 10 papers10TM hydroxides, phosphides & sulfidesPorous frameworks — 5 papers5Porous frameworksPolymers — 10 papers10PolymersThin-film growth & sputtering — 5 papers5Thin-film growth & sputteringSi oxidation — 8 papers8Si oxidationComposites — 3 papers3CompositesOER — 13 papers13OERHER — 10 papers10HERORR & fuel cells — 4 papers4ORR & fuel cellsH2O2 chemistry — 3 papers3H2O2 chemistryPhotoelectrochemistry — 4 papers4PhotoelectrochemistryCO2 conversion — 4 papers4CO2 conversionThermal catalysis — 7 papers7Thermal catalysisBiomass valorization — 1 paperBiomass valorizationAdsorption & separation — 6 papers6Adsorption & separationZinc batteries — 6 papers6Zinc batteriesLi & Na batteries — 6 papers6Li & Na batteriesPost-CMP cleaning — 4 papers4Post-CMP cleaningCorrosion — 1 paperCorrosionElectronic devices — 6 papers6Electronic devicesInterfaces — 22 papers22InterfacesDefects & vacancies — 12 papers12Defects & vacanciesDoping & promoters — 12 papers12Doping & promotersStrain — 6 papers6StrainElectronic descriptors — 14 papers14Electronic descriptorsSurface reconstruction — 5 papers5Surface reconstructionDiffusion & transport — 9 papers9Diffusion & transportNucleation & dendrites — 5 papers5Nucleation & dendritesClassical MD — 18 papers18Classical MDDFT — 60 papers60DFTAIMD — 4 papers4AIMDFree-energy diagrams — 18 papers18Free-energy diagramsAb initio thermodynamics — 5 papers5Ab initio thermodynamicsReaction barriers — 9 papers9Reaction barriersMultiscale modeling — 3 papers3Multiscale modelingMLIP — 4 papers4MLIPActive learning — 3 papers3Active learningMicrokinetic modeling — 1 paperMicrokinetic modelingGrand-canonical DFT — 1 paperGrand-canonical DFTAutomated synthesis — 2 papersAutomated synthesisMaterials data & LLM — 1 paperMaterials data & LLM
200820102015202020252027

Research · 01

Electrocatalysis for green hydrogen

Catalysts and membranes for water, seawater, and beyond.

Green hydrogen is only as good as the electrocatalyst it runs on. We design HER and OER catalysts from first principles, from high-entropy alloys and single-atom catalysts to layered double hydroxides and metal/metal-oxide hybrid catalysts. Active learning lets us navigate composition spaces far too large for brute-force calculation, optimizing activity, selectivity, and stability at the same time.

Real electrolytes are not ideal. In saline and seawater electrolysis, chlorine evolution competes with oxygen evolution, and durability is decided under operating potentials. We treat these constraints explicitly with grand-canonical DFT and Pourbaix analysis under realistic pH and potential.

We also look beyond the catalyst. Molecular dynamics and machine-learned force fields of ion-solvating polymer membranes show how ions move through alkaline electrolyzers. This links membrane chemistry to cell performance.

Research image: Electrocatalysis for Green Hydrogen

Research · 02

Sustainable chemical conversion

Electrifying chemistry, from biomass and nitrate to value-added products.

Electrochemistry can turn waste streams into resources. We compute the reaction mechanisms behind these conversions: which bonds break, on which sites, and why.

For biomass valorization, we study how lignin-derived molecules adsorb and cleave on catalyst surfaces, aiming at selective C–C and C–O bond scission driven by light and electricity. For the nitrogen cycle, we screen alloy catalysts that reduce nitrate in wastewater to ammonia. For a carbon-neutral chemical industry, we design catalysts for CO2 capture and utilization.

Every thread follows the same pattern. Mechanism first, then descriptors, then machine-guided search.

Research image: Sustainable Chemical Conversion

Research · 03

Advanced energy storage

Materials and interfaces for safer, longer-lived batteries.

Battery performance is decided by its materials. We design electrode materials for higher capacity and faster charging, across chemistries from lithium to aqueous systems.

Battery degradation is an interface problem. We study electrode and electrolyte interfaces at the atomic scale, from dendrite suppression on metal anodes to deposition behavior in aqueous cells. These insights guide materials and operating strategies for safer and longer-lived batteries.

Research image: Advanced Energy Storage

Research · 04

Semiconductor materials & processes

Atomic-scale materials and chemistry for the fab.

Semiconductor manufacturing is materials chemistry pushed to its limits. Across deposition, etching, polishing, and implantation, interactions at the atomic scale decide device performance and yield.

We use DFT and machine learning to discover and design the chemicals used in semiconductor processes. Adsorption and reaction energetics on metal, oxide, and nitride surfaces tell us how a molecule behaves at a critical interface. Property-prediction models trained on large molecular databases let us search chemical space far beyond known formulations.

We also study the materials themselves. From interconnect metals and their oxides to defects formed under processing conditions, we connect atomic-scale structure to the properties that devices depend on.

Research image: Semiconductor Materials & Processes