Atoms, Planets and the Cosmos

How do scientists predict chemical reactions before anyone gets hurt? Traditional chemistry faces a dilemma: lab experiments are slow and dangerous, while quantum calculations are accurate but take months to simulate tiny proteins. Machine learned interatomic potentials (MLIPs) are AI models that predict atomic interactions with quantum accuracy at a fraction of the cost. By training AI on both positions and atomic charges we create models that better understand atoms. This lets us predict reactions faster without losing accuracy, answering questions previously unthinkable.

We learn that matter is solid, liquid, or gas. But under extreme pressure and temperature, atoms can behave in ways that blur these boundaries. In some metals like calcium and tantalum, scientists have observed cooperative diffusion, where atoms move collectively while the material still appears solid. Could iron, the main ingredient of Earth-like planetary cores, behave this way too? Using advanced simulations, I explore whether iron enters this exotic state and what that means for exoplanetary interiors.

About 85% of the matter in the Universe is completely invisible to our telescopes. We call this mysterious substance dark matter. Although we cannot see it directly, its gravity shapes how galaxies form and how the Universe is structured. One of the best ways to study it is by observing how its gravity bends light from distant galaxies. In this talk, I’ll show how we use this subtle effect to uncover the hidden side of the Universe.