Things to make and do with data-derived potentials

Our paper demonstrating the wider applications of so-called Ephemeral Data-Derived Potentials (EDDP) has just appeared in JCP!

Developments and further applications of ephemeral data derived potentials
Pascal T. Salzbrenner, Se Hun Joo , Lewis J. Conway , Peter I. C. Cooke , Bonan Zhu , Milosz P. Matraszek , William C. Witt, and Chris J. Pickard
The Journal of Chemical Physics, vol. 159, no. 14, p. 144801, Oct. 2023

We show that while EDDPs are great for structure prediction, they are also not too bad for molecular dynamics and phonon calculations.

We benchmark EDDPs against a set of other Carbon potentials; turns out they are pretty comparable.

With a decent potential you can see things which would be difficult/impossible with ab initio methods.

For some perspective, a calculation with an EDDP for 8 atoms of Lead, accounting for spin-orbit coupling, is 10,000 times faster than with DFT.

To demonstrate this, we show a series of case studies which show:

• ⚛️ The phase diagram of heavy elements with relativistic effects.
• 💎 Hydrogen diffusion in hydrogen-rich hydrides at high-pressure.
• 🏗️ Thermal expansion of metal organic frameworks.