Our recent article in npj Computational Materials presents an efficient ML protocol for accelerating trajectory surface hopping dynamics, while tackling many key issues making machine learning of excited states difficult. The protocol introduces a new machine learning interatomic potential based …
Are universal machine learning potentials for excited states possible? Such a potential would be a major breakthrough — enabling key applications like the design of advanced photomaterials. We’ve already seen successful universal potentials for ground states — ANI-1ccx, MACE-OFF, our …
Meet OMNI-P2x — the First Universal ML Potential for Excited States! Read more »
Another year flew by like a comet. And what a year it was! This year was packed full of updates from my group and, of course, MLatom, which is at the center of our research. MLatom provides us with a …
Highlights of the Year 2024! Last Weekly Update? Read more »
Theoretical IR (infrared) spectroscopy is a powerful tool for assisting chemical structure identification. However, approaches based on quantum chemical calculations suffer from either high computational cost (e.g., density functional theory, DFT) or insufficient accuracy (semi-empirical methods). Hence, we introduce a new …
ML-enhanced Fast and Interpretable Simulation of IR Spectra Read more »
Density functional theory (DFT) methods are by far the most popular approaches for electronic structure calculations. However, the “best” functional remains elusive despite the increasing variety of functionals and continuous efforts to improve their computational accuracy. In our work published in Advanced …
Adv. Sci.: The Best DFT Functional Is the Ensemble of Functionals Read more »
Recently, we published a paper in JOC about the surprising dynamics phenomena in the Diels–Alder reaction of fullerene C60. The AI-accelerated molecular dynamics uncovers that in a small fraction (10%) of reactive trajectories, the diene molecule (2,3-dimethyl-1,3-butadiene) is roaming around …
JOC: Surprising dynamics phenomena in the Diels–Alder reaction of C60 uncovered with AI Read more »
Recently, we published a paper in JCTC about the end-to-end physics-informed active learning with data-efficient construction of machine learning potentials. It shortens molecular simulation time to a couple of days which could have taken weeks of pure quantum chemical calculations.
A year ago, we released MLatom 3, making MLatom the fully-fledged Python package. This shift endowed MLatom with much of its current flexibility that also allowed us to rapidly do method development and new implementations: in one year we had …
XACS team in collaboration with Mario Barbatti and groups in Warsaw University and Zhejiang lab has recently published a paper in JCTC about the versatile Python implementation of surface-hopping dynamics. This implementation is based on a powerful MLatom ecosystem for …
JCTC: Surface hopping dynamics with QM and ML methods Read more »
A machine learning potential with low error in the potential energies does not guarantee good performance for the simulations. One of the reasons is that it is hard to train machine learning potentials with balanced descriptions of different PES regions, …