Dual vibration configuration interaction (DVCI). An efficient factorization of molecular Hamiltonian for high performance infrared spectrum computation

Abstract

Here is presented an original program based on molecular Schrödinger equations. It is dedicated to target specific states of infrared vibrational spectrum in a very precise way with a minimal usage of memory. An eigensolver combined with a new probing technique accumulates information along the iterations so that desired eigenpairs rapidly tend towards the variational limit. Basis set is augmented from the maximal components of residual vectors that usually require the construction of a big matrix block that here is bypassed with a new factorisation of the Hamiltonian. The latest borrows the mathematical concept of duality and the second quantization formalism of quantum theory.Licensing provisions: GNU General Public License 3.Programming languages: C/C++/Fortran. Supplementary materials: 1. The sources of the code grouped in folder DualVCI.zip also available at https: // github. com/ 4Rom1/ DualVCI 2. The input files of examples treated in section 6.Nature of problem: High computational cost in vibration configuration interaction methods [1,2], coming from the necessity to solve a large eigenvalue problem to acquire a good precision. The dimension of the matrix exponentially increases with the size of the studied molecule.Solution method: The A k decomposition [3] completed by a meaningful error evaluation namely the residue ||HX´EX|| minimised along iterations for specific targets given in input. The approximation space is generated in the same time as the residual vectors computed on the fly thanks to an adapted choice of excitations shaped on the Hamiltonian operator.