Quantum mechanical and quasiclassical trajectory surface hopping studies of the electronically nonadiabatic predissociation of the (A)over-tilde state of NaH2
Md. Hack et al., Quantum mechanical and quasiclassical trajectory surface hopping studies of the electronically nonadiabatic predissociation of the (A)over-tilde state of NaH2, J PHYS CH A, 103(32), 1999, pp. 6309-6326
Fully coupled quantum mechanical scattering calculations and adiabatic unco
upled bound-state calculations are used to identify Feshbach funnel resonan
ces that correspond to long-lived exciplexes in the (A) over tilde state of
NaH2, and the scattering calculations are used to determine their partial
and total widths. The total widths determine the lifetimes, and the partial
widths determine the branching probabilities for competing decay mechanism
s. We compare the quantum mechanical calculations of the resonance lifetime
s and the average final vibrational and rotational quantum numbers of the d
ecay product, H-2(nu', j'), to trajectory surface hopping calculations carr
ied out by various prescriptions for the hopping event. Tully's fewest swit
ches algorithm is used for the trajectory surface hopping calculations, and
we present a new strategy for adaptive stepsize control that dramatically
improves the convergence of the numerical propagation of the solution of th
e coupled classical and quantum mechanical differential equations. We perfo
rmed the trajectory surface hopping calculations with four prescriptions fo
r the hopping vector that is used for adjusting the momentum at hopping eve
nts. These include changing the momentum along the nonadiabatic coupling ve
ctor (d), along the gradient of the difference in the adiabatic energies of
the two states (g), and along two new vectors that we describe as the rota
ted-d and the rotated-g vectors. We show that the dynamics obtained from th
e d and g prescriptions are significantly different from each other, and we
show that the d prescription agrees better with the quantum results. The r
esults of the rotated methods show systematic deviations from the nonrotate
d results, and in general, the error of the nonrotated methods is smaller.
The nonrotated TFS-d method is thus the most accurate method for this syste
m, which was selected for detailed study precisely because it is more sensi
tive to the choice of hopping vector than previously studied systems.