Research

From atoms to better materials

I use AI, machine learning and multiscale simulation, from electrons to membranes, to explain why a material works and to predict which one will work better.

Animation: an atomic structure is fed through a neural network, which predicts the force on each atom

Teaching machines physics and chemistry

Quantum-mechanical calculations tell us how atoms push and pull on each other, but they are far too slow for the system sizes and timescales that decide whether a material survives in a real device. I train machine-learned interatomic potentials on DFT data, so that molecular dynamics can run at close to first-principles accuracy for a fraction of the cost.

For CALF-20, a benchmark MOF for industrial-scale CO₂ capture, this approach revealed negative thermal expansion, negative area compression, a two-step elastic response and a fracture strain of up to 27%, all properties that standard force fields could not have predicted. I release the potentials and training data openly, and I use the same data-driven mindset elsewhere: screening candidate frameworks, and clustering large simulation datasets to find the descriptors that matter.

Animation: an atomistic model becomes a coarse-grained bead chain and then a continuum field

From electrons to membranes

No single simulation method covers everything that matters in a real material. Electrons decide how a molecule binds, atoms decide how it moves through a pore, and the arrangement of grains, binders and polymers decides whether a pellet or a membrane works at all. I link these scales: DFT for chemistry, molecular dynamics and grand-canonical Monte Carlo for adsorption and transport, and coarse-grained models where the interesting physics only shows up at larger sizes.

That chain of methods produced the first multiscale screen of MOF/binder composites for CO₂ capture, showed how MOF surface morphology shapes the pores that form at a MOF/polymer interface, and identified a 2D-MOF monolayer that lets water through while rejecting salt completely. With experimental partners, it also helped explain why putting the MOF only at the membrane surface raises permeance five-fold.

Animation: a uniform mixture separates into droplets that grow and merge

Matter that organises itself

Soft matter builds itself. Proteins condense into liquid-like droplets inside cells, short DNA duplexes stack into liquid crystals, and DNA nanostars link up into gels as the temperature drops. None of this is designed atom by atom. It emerges from weak interactions acting on many molecules at once, which makes it a natural problem for simulation across scales.

I study how the SUMO proteins tune the phase separation and aggregation of TDP-43, a protein linked to neurodegenerative disease, together with experimental colleagues. During my PhD I built the first atomistic models of RNA nanotubes, measured how salt stiffens DNA nanotubes, and compared coarse-grained DNA models against all-atom simulation to find out which ones can be trusted.