Skip to content

Research

We develop molecular models and simulation methods that connect quantum chemistry and statistical mechanics to the behavior of water, biomolecules, and materials. Our work combines many-body potentials, molecular simulation, and spectroscopy across six connected research areas.

Many-body potentials

Building molecular models from quantum-mechanical interactions.

We develop potential energy functions that connect electronic structure to molecular behavior. The many-body expansion separates the total energy into contributions from individual molecules, pairs, triples, and higher-order interactions. In MB-nrg, learned short-range terms are combined with electrostatics, dispersion, and many-body polarization.

By organizing large molecules into transferable fragments, we extend this framework from small molecules to peptides and other biomolecular systems. Comparisons of conformational free-energy surfaces test how accurately these models describe molecular structure and flexibility.

Molecular fragments illustrate one-body, two-body, and three-body contributions, followed by classical dispersion, repulsion, electrostatic, and polarization terms.
The many-body description separates molecular contributions while retaining collective polarization.
Gas-phase alanine dipeptide conformational free-energy surfaces from MB-nrg, AMBER FF14SB, and MACE-0, alongside the molecular structure.
Alanine dipeptide free-energy surfaces compare the conformations predicted by MB-nrg, AMBER FF14SB, and MACE-0.

Ion hydration and interfaces

Understanding how water organizes around ions, from solution to interfaces.

Ion hydration depends on a balance of local ion–water interactions and the collective response of the surrounding liquid. We use many-body potentials to study hydration shells in clusters and bulk solutions, including concentrated and mixed-ion environments.

At the air–water interface, we examine how ion–water and water–water interactions determine whether an ion remains in the liquid or approaches the surface. Free-energy calculations let us compare these effects across ions and molecular models.

Water structures around halide ions progress from a single hydration-shell structure to multiple hydration shells, with colored surfaces showing the surrounding water organization.
Many-body models resolve changes in the structure of halide hydration shells.
Free-energy profiles for fluoride, chloride, bromide, and iodide approaching an air–water interface, comparing MB-nrg with TTM-nrg.
Interfacial free-energy profiles reveal how the treatment of many-body interactions changes halide surface propensity.

Molecular separations in MOFs

Connecting pore-scale interactions to water adsorption, carbon capture, and ion transport.

Metal–organic frameworks create confined environments in which pore geometry and molecular interactions control adsorption and transport. We study how water, carbon dioxide, and ions compete for these environments, using molecular simulations to interpret experiments and guide material design.

Our work spans water and carbon dioxide adsorption in CALF-20, ion uptake in MOF-808, and lithium-ion transport in SU-102. We also investigate how tethering organic ligands changes framework structure and properties.

A molecular rendering of a CALF-20 pore surrounds a water molecule and carbon dioxide molecule, shown as competing chess pieces.
Water and carbon dioxide compete for adsorption environments in CALF-20.
SU-102 framework and solvated lithium ions, a table of transitions between coordination environments, and detailed linker-associated and zirconium-associated sites.
Simulations identify distinct lithium-ion coordination environments and transport pathways in SU-102.

Molecular simulation software

Making accurate many-body simulations practical on modern computing hardware.

Our software connects the development of many-body potentials to production molecular simulations. MBX implements MB-pol and MB-nrg energy and force calculations, while MB-Fit supports the construction of many-body potential energy functions.

We develop efficient algorithms and optimize their implementation for CPUs and GPUs. Timing breakdowns identify computational bottlenecks, and hardware benchmarks measure improvements in polynomial evaluation and molecular simulation performance.

Explore MBX and MB-Fit.

Stacked MBX timing bars separate polynomial, dispersion, electrostatic, and overhead costs for different water-system sizes and software versions.
MBX timing breakdowns show how individual energy terms contribute to the cost of a simulation step.
Kernel and tensor algorithm benchmarks compare AMD EPYC CPUs with an NVIDIA A100 GPU as thread counts and the number of polynomial terms change.
CPU and GPU benchmarks compare polynomial-evaluation algorithms across the hardware and problem sizes shown.

Ice and antifreeze proteins

Finding molecular mechanisms that control ice growth and inhibition.

We investigate how antifreeze proteins and small molecules such as phloroglucinol affect ice growth. Molecular simulations connect the structure of an ice-binding surface to the organization and dynamics of nearby water.

We combine these studies with machine learning to design ice-binding proteins and peptides, and develop transferable many-body models for organic ligands with antifreeze properties. The aim is to understand which molecular features control ice binding and to use that knowledge in molecular design.

An antifreeze protein at an ice–water interface, surrounded by water molecules and an ordered molecular surface.
Molecular models resolve the environment around an antifreeze protein at an ice–water interface.
A beta-solenoid antifreeze protein scaffold with highlighted groups arranged along its ice-binding surface.
Deep learning tools enable the design of self-assembling ice-nucleating peptides.

Quantum dynamics and spectroscopy

Using vibrational spectra to connect molecular structure with quantum behavior.

We combine path-integral quantum dynamics with data-driven many-body potentials to study water in the bulk, at interfaces, and across phases. Simulated linear and nonlinear spectra help connect experimental signals to molecular structure and motion.

Current questions include the role of nuclear quantum effects in ice pre-melting, the different environments of free O–H groups at the air–water interface, and the behavior of supercooled water. Polarization-resolved and multidimensional spectroscopy provide complementary views of these molecular environments.

Polarization-resolved experimental and MB-pol/Te-PIGS sum-frequency spectra, with a correlation between free O–H frequency and the local electric field.
Polarization-resolved HD-vSFG connects free O–H spectral signatures to heterogeneous environments at the air–water interface.
A supercooled-water phase diagram labels high- and low-density liquid, a liquid–liquid critical point, and the Widom line, beside distinct HDL and LDL 2D IR–Raman spectra.
Supercooled-water phase behavior and two-dimensional IR–Raman spectra offer complementary views of high- and low-density liquid environments.