Speaker
Description
Accurately computing the electronic energy of molecular reactants and during chemical reactions is the central challenge of quantum chemistry. Unlike mean-field methods such as Hartree-Fock or Kohn-Sham density functional theory, configuration interaction (CI) methods exactly calculate the repulsive interaction between electrons based on a many-particle wave function rather than relying on approximate functionals. This leads to a systematically improvable accuracy, but also requires far greater computational resources and becomes increasingly intractable for large molecules. Complete Active Space (CAS) methods mitigate this computational cost by restricting the full configuration interaction (FCI) calculation to a subset of orbitals (the active space) while keeping the occupation of the remaining (nonactive) ones fixed. This simplification inevitably introduces an error depending on the quality of the involved orbitals, but it was shown that this error can be minimised by an algorithm based on quantum information: minimizing the entanglement of the nonactive orbitals also leads to a minimization of the resulting CAS ground-state energy. This method is called Quantum Information-Assisted Complete Active Space Optimization (QICAS).
QICAS has previously only been formulated and tested for calculations on electronic ground states. Due to the high relevance of excited states in most subfields of chemistry including catalysis, photonic materials, and spectroscopy, and their more complex electronic structures, there is a significant demand for their reliable and efficient computational modeling. Motivated by this, we extended QICAS to target excited states (and ensembles thereof) as well. We present a newly developed quantum information cost function based on the ensemble variational principle and discuss that without its invocation, the QICAS cost function loses its validity as a proxy for active-space quality.
