A pseudospectral algorithm for the computation of transitional-mode eigenfunctions in loose transition states. II. Optimized primary and grid representations
Aj. Rasmussen et al., A pseudospectral algorithm for the computation of transitional-mode eigenfunctions in loose transition states. II. Optimized primary and grid representations, J CHEM PHYS, 110(3), 1999, pp. 1354-1364
A highly optimized pseudospectral algorithm is presented for effecting the
exact action of a transitional-mode Hamiltonian on a state vector within th
e context of iterative quantum dynamical calculations (propagation,diagonal
ization, etc.). The method is implemented for the benchmark case of singlet
dissociation of ketene. Following our earlier work [Chem. Phys. Lett. 243,
359 (1995)] the action of the kinetic energy operator is performed in a ba
sis consisting of a direct product of Wigner functions. We show how one can
compute an optimized (k, Omega) resolved spectral basis by diagonalizing a
reference Hamiltonian (adapted from the potential surface at the given cen
ter-of-mass separation) in a basis of Wigner functions. This optimized spec
tral basis then forms the working basis for all iterative computations. Two
independent transformations from the working basis are implemented: the fi
rst to the Wigner representation which facilitates the action of the kineti
c energy operator and the second to an angular discrete variable representa
tion (DVR) which facilitates the action of the potential energy operator. T
he angular DVR is optimized in relation to the reference Hamiltonian by sta
ndard procedures. In addition, a scheme which exploits the full sparsity of
the kinetic energy operator in the Wigner representation has been devised
which avoids having to construct full-length vectors in the Wigner represen
tation. As a demonstration of the power and efficiency of this algorithm, a
ll transitional mode eigenstates lying between the potential minimum and 10
0 cm(-1) above threshold have been computed for a center-of-mass separation
of 3 Angstrom in the ketene system. The performance attributes of the earl
ier primitive algorithm and the new optimized algorithm are compared. (C) 1
999 American Institute of Physics. [S0021-9606(99)00502-4].