Next Generation of Selected Configuration Interaction and Exact Diagonalization Techniques: The Adaptive Sampling Configuration Interaction (ASCI) Method

Selected configuration interaction techniques (SCI) have been shown to produce especially accurate results for treating strongly correlated Hamiltonians, with applications in lattice models, quantum chemistry, embedding techniques, and excited states. We have recently introduced a new SCI technique, the adaptive sampling CI (ASCI) approach, and demonstrated that this algorithm, together with a heat bath extension of it, are among the most efficient algorithms for treatment of strongly correlated systems. In this work, we show how these adapted CI techniques lead to a new level of accuracy and efficiency for selected CI calculations, when tailored to modern computer architectures. These computational advances allow for simulations with the ASCI algorithm (and other SCI methods) that achieve accuracy beyond what is possible with most other simulation techniques.