Séminaires à venir

Excited state molecular dynamics (MD) simulations are a powerful computational tool for the study of photoinduced phenomena. A popular method to get the excited state energy and its gradients is linear response time dependent density functional theory (TD-DFT). In this method, the excitation energies and the corresponding perturbations to the electronic density are obtained as eigenvalues and eigenvectors of the molecular response function. The gradients are then used to propagate the equations of motion of the molecule. At each step of an MD simulation, regardless of the method, the new molecular geometry is relatively close to the previous ones. This observation suggests that some extrapolation strategy can be applied, such that the results of the previous calculations (which are available for free) can be used to predict the result of the upcoming calculation. The prediction can then be used as a guess for the iterative solver to lower the number of iterations and thus the cost. In the case of ground state MD simulations, several strategies extrapolation strategies have been proposed. Among these, the most successful are the extended Lagrangian from Niklasson[1], the GExt scheme from my coworkers[2,3] and Car-Parrinello MD strategies[4]. In the case of linear response TD-DFT MD, only a strategy based on the extended Lagrangian has been proposed so far[5]. In this talk, I first introduce the linear response TD-DFT method for excited states, highlighting its connection with differential geometry. Then, I present an extension of the GExt scheme presented in [3] to linear response TD-DFT, in which the manifold structure to which the solutions belong is used to make the extrapolation more accurate. Lastly, I show how the new extrapolation strategy can be applied to excited state MD simulations resulting in a significant acceleration. [1] Niklasson, J. Chem. Phys. 152 (2020). [2] Polack et al., J. Chem. Theory Comput. 17 (2021). [3] Pes et al., J. Phys. Chem. Lett. 14 (2023). [4] Kühne, WIREs Comput. Mol. Sci 4 (2014). [5] Niklasson et al., J. Chem. Theory Comput. 14 (2018).