Dual partitioning for effective Hamiltonians to avoid intruders

TL;DR

The paper introduces a dual partitioning method for effective Hamiltonians that converges multiple states simultaneously, avoiding intruder states and improving efficiency for excited state calculations.

Contribution

A novel dual partitioning technique that prevents intruder states and enhances convergence in effective Hamiltonian methods for excited states.

Findings

Successfully applied to model systems showing improved convergence.

Avoids intruder state problems in Hamiltonian partitioning.

Applicable to excited states with conical intersections.

Abstract

We present a new Hamiltonian partitioning which converges an arbitrary number of states of interest in the effective Hamiltonian to the full configuration interaction limits simultaneously. This feature is quite useful for the recently developed model space quantum Monte Carlo. A dual partitioning (DP) technique is introduced to avoid the intruder state problem present in the previous eigenvalue independent partitioning of Coope. The new approach is computationally efficient and applicable to general excited states involving conical intersections. We present a preliminary application of the method to model systems to investigate the performance.