Simone Sturniolo, Theoretical and Computational Physics, SCD
Host: Alin Marin Elena, Computational Chemistry, SCD
Tuesday 15th February 2022 @ 14:00-15:00
A new approach to finding the quantum ground state for atomic nuclei
Quantum nuclear effects are a known issue that affects the results of many experiments and controls the kinetics of many chemical reactions, but they are often ignored in ab initio simulation due to the extreme complexity a full treatment would introduce. Methods based on Path Integral Molecular Dynamics are the most popular approximation used to treat these effects, but they do not represent "real" dynamics, as they evolve in imaginary time, become more computationally expensive for low temperature systems, and can be difficult to set up and interpret for beginners.
In this work I propose a completely new class of methods to treat quantum nuclear effects with multiple classical simulations coupled by an artificial potential. These methods are the offspring of the Many Interacting Worlds (MIW) approach proposed by Hall, Wiseman and Deckert in 2014, which in turn is based on the de Broglie - Bohm interpretation of quantum mechanics. MIW MD methods are the complementary of PIMD: they perform best at low temperature and near the ground state, they naturally approximate real time quantum dynamics, and can be easily set up and their results straightforwardly interpreted. They also allow for mixed simulations in which only some atoms are treated as quantum, and can be used in combination with optimisation algorithms like LBFGS to find the quantum ground state. In particular, I will propose a further simplification of the MIW MD method that could allow to get fair estimates of the delocalisation of atoms within their local potential wells with as little as six or seven classical copies of the system running in parallel, independently of the temperature of the system or mass of the nuclei.
