Multiscale phenomena are the starting point of our research. Engineering behavior emerges from coupled mechanisms that span electronic structure, atoms, molecular and mesoscale organization, evolving features, and continuum transport. We develop models at each relevant scale and connect them with physics-based and data-driven scale-bridging methods, allowing information to move from fundamental mechanisms to engineering prediction while retaining physical meaning. This framework is shared across semiconductor processing, functional materials, and equipment-scale systems.
We begin by identifying the phenomena, state variables, and interfaces that control an engineering problem across length and time. Quantum chemistry, molecular organization, transport, mechanics, and field effects may each require a different description. We formulate them as a connected system in which each model operates at the scale where its physics remains valid.
DFT, ab-initio molecular dynamics, and classical or reactive molecular dynamics resolve electronic energetics, chemical reactions, defects, interfaces, impact, and deformation. Applications span plasma–surface reactions, adsorption and deposition chemistry, wafer-bonding interfaces, and the molecular mechanisms of functional materials.
Coarse-grained molecular dynamics, dissipative particle dynamics, and kinetic Monte Carlo extend simulation to collective behavior and longer time scales. We use them to study molecular organization, phase separation, viscoelasticity, microstructure evolution, and thin-film growth that cannot be represented efficiently at full atomistic resolution.
Feature-scale and continuum models translate microscopic reaction and transport data into engineering observables. DSMC and level-set methods predict transport and profile evolution inside patterned features, while CFD and finite-element models describe reactor behavior, equipment conditions, and continuum material response.
AI connects models whose accuracy and computational cost differ by orders of magnitude. Machine-learning force fields extend first-principles information to larger atomistic systems, while surrogate and reduced-order models transfer microscopic results into feature- and equipment-scale calculations. Active learning and uncertainty quantification identify where new data or higher-fidelity physics is required.
DFT
VASP
Ab-initio MD
LAMMPS
ReaxFF
OVITO
Structure-based coarse-graining
IBI
CG MD
DPD
kMC
DSMC
Level-set profile evolution
Ansys Fluent
Chemkin
FEM
Homogenization
MLFF
Graph neural-network potentials
Reduced-order modeling
Active learning
Uncertainty quantification
Structural characterization and bonding energy analysis for plasma-activated bonding of SiCN films: A reactive molecular dynamics study
Deep neural network-based reduced-order modeling of ion–surface interactions combined with molecular dynamics simulation
Multiscale modeling of load transfer characteristics in crosslinked epoxy nanocomposites
Atomistic kinetic Monte Carlo simulation on atomic layer deposition of TiN thin film
Multi-scale coarse-grained molecular dynamics simulation to investigate the thermo-mechanical behavior of shape-memory polyurethane copolymers