Tree tensor network states have emerged as a powerful framework for the simulation of the quantum dynamics of molecules following photo-excitation. Here, a high-dimensional coefficient tensor representing the quantum state is represented by a hierarchical network of contractions of small-dimensional tensors, generating tree topologies. There exists great flexibility in the tree topology, allowing for a tailoring to the physical process being simulated. However, the computational cost is closely tied to the tree topology, and traditionally this had to be designed by hand on a system-by-system basis: a non-trivial task.

In this presentation, I will discuss an approach to automating the tree topology construction via the combination of ideas from quantum information theory and computer science. Specifically, I shall show how the mutual rate of entanglement of vibrational and electronic degrees of freedom can be combined with hierarchical clustering algorithms to reliably, rapidly, and fully automatically generate good tree topologies that are tailored to a given photo-induced molecular process.