Electronic Structure of Excited States with Configuration Interaction Methods

Abstract

Computational chemistry is routinely applied to ground state molecular systems to provide chemical insights. Accurate excited state calculations, however, still typically require carefully tailored calculations and sizeable computational resources. This work focuses on the development of methods and strategies that enable the calculation of excited state properties with more accuracy and on larger systems than ever before. The first two Chapters focus on the spin-flip configuration interaction family of methods. Chapter 2 introduces us to the quantities one can obtain with excited state methods, with a challenging example being the electronic structure of a possible intramolecular singlet fission system, a quinoidal bithiophene. The study assigns an experimentally observed long-lived exciton to a long-lived singlet multiexciton state with a combination of energetic and transition dipole moment quantities. The spin-flip methodology is extended in Chapter 3 to provide more insight into the energetic orderings of the multiexciton states of a tetracene dimer, a model singlet fission system, showing that triplet decoupling should occur spontaneously upon population of the intermediate multiexciton state, 1(TT). However, this extension enlarged the configuration spaces to the point that they became a limiting factor in the calculation of larger systems.

Therefore, the latter two Chapters focus on investigating new strategies for identifying and eliminating unneeded configurations. Chapter 4 presents iterative submatrix diagonalization, a procedure for converging the Davidson diagonalization procedure with a reduced set of active orbitals. This is accomplished by generating a systematic series of submatrix approximations to the full configuration space and solving for eigenpairs within the series until convergence of eigenpairs is achieved. The method shows promise, converging eigenvalues with a considerable reduction in orbitals and total computational time. Chapter 5 applies heat-bath configuration interaction towards obtaining exact excitation energies and examines various ways in which convergence is signified. A new convergence metric based on the magnitude of the perturbative correction is developed and converged excitation energies are obtained for systems as large as hexatriene. These results involved treating configuration spaces with as many as 1038 configurations, a full 29 orders of magnitude over what is achievable with conventional configuration interaction methods and 10 orders beyond results reported by other recent state-of-art solvers. While there is still a great deal of work to be done before excited state computational chemistry will be routinely applicable to a wide variety of systems, the various methods investigated and extended here show significant promise, especially those presented in the latter Chapters as these are generally applicable to any configuration interaction method.