“…The size of the direct-product basis and the direct-product grid grows exponentially with the number of active vibrational degrees of freedom. The corresponding computational cost can be mitigated by efficient algorithmic and implementation techniques, most importantly by (a) evaluating nested sums, e.g., eqn (52), sequentially; 43,[45][46][47][48][49][50] (b) using an iterative (Lanczos) eigensolver, 51,52 which requires only multiplication of a trial vector with the Hamiltonian matrix without storage or even explicit construction of the full matrix. 21,52,53 Nevertheless, even in the most efficient implementation, a few vectors of the total size of the basis (and the grid) must be stored, which grows exponentially with the vibrational dimensionality.…”