Funding & Support

Machine Learning Molecular Dynamics

We develop machine‑learning molecular dynamics (ML‑MD) simulations that enables near‑DFT accuracy with near-MD speed of reactive atomistic simulation. By training neural network force fields on high‑quality ab initio datasets, we overcome the limitations of classical MD and accelerate the design of next‑generation avanced devices, including batteries, fuel cells, brain-inspired computing, and molecular separations.

ML-MD Simulation for Energy Applications
Multi‑scale atomistic simulation using machine‑learning force fields to enable energy and molecular biomimetic applications.

Sub‑projects

  • ML‑Driven Molecular Dynamics: Reactive simulations trained on high‑quality DFT datasets for energy materials.
  • Multiscale Energy Applications: Batteries, fuel cells, energy‑harvesting, and solar conversion systems.
  • Innovative Biomimetic devices: Ionic neuromorphic computing, Bioinspired sensing/conversion/separation.
Ionic Neuromorphic Computing for Energy‑Efficient AI

We develop ionic neuromorphic computing systems that mimic the human brain’s ion‑based signaling to achieve ultra‑low‑power artificial intelligence. This research focuses on subnanometer 2D ionic memristors that enable synaptic plasticity, memory retention without continuous energy input, and energy‑efficient signal processing.

Ionic Neuromorphic Computing
Schematic of biological neuron, software‑based artificial neuron, and physical neuron realized in ionic neuromorphic computing.

Sub‑projects

  • Å‑Scale Ionic Memristors: Synthetic ion channels that switch and retain memory through precisely controlled ion flow.
  • Physical Neural Networks: Building energy‑efficient ionic circuits as AI hardware.
  • Multiscale Simulation & Theory: Molecular simulations to reveal sub‑1 nm ion transport mechanisms and guide device design.
Nanopore-Based Molecular Separation for Hydrogen & Rare-Earth Elements

We design bio-inspired nanopore separation systems to selectively extract target molecules, such as hydrogen and rare-earth elements (REEs). Inspired by biological ion channels, we leverage confined electrostatics, molecular binding, entropic barriers, hydration energy, and nanostructured pore geometries to achieve highly selective and energy-efficient molecular separations.

Nanopore-Based Molecular Separation
Bio-inspired nanopore separation for hydrogen and rare-earth element extraction, mimicking the ultra-selective transport of biological ion channels.

Sub‑projects

  • Hydrogen Sieving: Quantum chemisty and MD simulations to design Å‑scale pores for selective H2 separation from gas mixtures.
  • Rare‑Earth Ion Extraction: Free‑energy landscape modeling of rare-earth ion transport through nanostructured membranes.
  • Multi‑Component Competition: Simulation of ion competition and selectivity hierarchies in complex extraction environments.
Generalized Stochastic Transport Theory

We are developing a stochastic transport theory to model and predict the behavior of ion transport through nanopores and membranes. In subatomically localized regions, Boltzmann statistics fail to describe the ensemble-averaged local properties of molecules. By employing Fermi–Dirac statistics, we are creating a generalized stochastic transport framework that accurately reproduces ensemble-averaged molecular dynamics.

Fermi-Dirac Transport Theory
Schematic comparison between continuum electrokinetic transport theory and stochastic electrochemical transport theory.

Sub‑projects

  • Fermi–Dirac Transport Theory: Demonstration of the foundational principles of Fermi–Dirac-type transport.
  • Generalized Field Theory of Fermi–Dirac Transport: A continuum-based framework that describes ensemble-averaged molecular dynamics.