“…In terms of the basis expansion, we separate the basis functions into the inner (r 1 and r 2 ) and outer (θ 1 , θ 2 , ϕ, and r 0 ) groups, and the expression of Hamiltonian matrix elements is given in eq S2 (Supporting Information), resulting in a large, sparse, and well-structured matrix. Herein, the PIST method combined with an optimal separable basis plus Wyatt (OSBW) preconditioner 31,35,36 is employed to solve its eigenvalue problem. To make large-scale computations possible, message passing interface (MPI) is utilized to parallelize the most timeconsuming parts of this scheme, which include the evaluation of Hamiltonian matrix elements, block Jacobi diagonalization and quasi-minimal residual iterations, and very good scalability is achieved.…”