Communication: Active-space decomposition for molecular dimers

Abstract

We have developed an active-space decomposition strategy for molecular dimers that allows for the efficient computation of the dimer's complete-active-space wavefunction while only constructing the monomers’ active-space wavefunctions. Dimer states are formed from linear combinations of direct products of localized orthogonal monomer states and Hamiltonian matrix elements are computed directly without explicitly constructing the product space. This decomposition is potentially exact in the limit where a full set of monomer states is included. The adiabatic states are then found by diagonalizing the dimer Hamiltonian matrix. We demonstrate the convergence of our method to a complete-active-space calculation of the full dimer with two test cases: the benzene and naphthalene dimers.

Orbital Optimization in the Active Space Decomposition Model

“…(1) is exact when each I and J entirely span the corresponding monomer space, and converges rapidly with respect to the number of monomer states included in the summation. 12 An extension to more than two active subspaces has also been reported by the authors. 15 Furthermore, the dimer basis states have well-defined charge, spin, and spin-projection quantum numbers on each monomer, allowing us to extract model Hamiltonians for electron and exciton dynamics through diagonalization of diabatic subblocks of a dimer Hamiltonian matrix.…”

“…The cc-pVDZ basis set 20 was used, and singlet (S), triplet (T), and chargetransfer (CT) monomer states are included. The CAS (12,12) active space for a dimer consists of 12 π electrons distributed in the π orbitals, which is decomposed to two 6-orbital ASD subspaces. First, the ground state energies and the first few excitation energies computed by ASD-CASCI and ASD-CASSCF as a function of the number of states in each charge and spin sector in the ASD expansion (N) are presented in Fig.…”

“…The errors are calculated relative to the conventional CASCI (−461.547119 E h ) and CASSCF (−461.584702 E h ) energies. ASD-CASSCF and ASD-CASCI converge to the conventional CASSCF and CASCI with full CAS (12,12) at large N. The Hartree-Fock orbitals were used in the CASCI calculations. For the ground state, the convergence of ASD-CASSCF is slightly slower than that with ASD-CASCI, although they are both almost exponential.…”

“…The details of the algorithm are given elsewhere. 12 We introduce a stationarity condition with respect to variations of MO coefficients to define orbital-optimized ASD models:…”

“…[4][5][6][7][8][9][10][11] We have recently introduced the active space decomposition (ASD) method to provide accurate model Hamiltonians. The ASD model takes advantage of molecular geometries to compress active-space wave functions of molecular dimers, [12][13][14][15] …”