Role of potential structure in the collisional excitation of metastable O(<SUP>1</SUP>D) atoms

This paper considers the collisional excitation of O(1D) modeled by the crossing of two valence 1 3&#928;g curves dissociating to O(3P)+O(3P) [V11(R)] and O(3P)+O(1D) [V22(R)] which in turn are further crossed by the 3&#928;g Rydberg curve dissociating to O(3P)+O(5S) [V33(R)]. The role of structure in the potential curves and coupling matrix elements is quantitatively probed by the first-order functional-sensitivity densities &#948; ln&#963;12(E)/&#948; lnVij(R) of the excitation cross section &#963;12(E) obtained from close-coupling calculations. The results reveal that, in spite of the well-separated nature of the crossing between the two valence curves from their crossings with the Rydberg potential curve, the excitation cross section &#963;12 displays considerable sensitivity to the Rydberg curve V33(R) at all energies in the range 3.0-9.0 eV. For relative collisional energies corresponding to the higher closely spaced vibrational energy levels of the Rydberg state, the excitation cross section is found to be much more sensitive to the Rydberg state than to the two valence states themselves. At all energies, the sensitivity of the excitation cross section &#963;12 to the coupling V12(R) between the valence states is much larger than the sensitivity to the couplings V13(R) or V23(R) with the Rydberg state. At higher energies, the large increase in the sensitivity of the cross section to the Rydberg potential is mirrored by a similar increase in sensitivity to its coupling V23(R) with the upper valence state. Due to the weak coupling between the three curves, a qualitative similarity exists between the sensitivity profiles and those predicted by the Landau-Zener-Stueckelberg (LZS) theory. Quantitative departures witnessed in earlier work are, however, more pronounced for the multilevel curve crossings investigated here. Implications of the results for attempts to extend the LZS-type treatment to multilevel curve crossings and for functional-sensitivity-based algorithms for the inversion of cross-section data are discussed