General explanation of geometric phase effects in reactive systems: Unwinding the nuclear wave function using simple topology.

We describe a simple topological approach which was used recently to explain geometric phase (GP) effects in the hydrogen-exchange reaction [Juanes-Marcos et al., Science 309, 1227 (2005)]. The approach is general and applies to any reactive system in which the nuclear wave function encircles a conical intersection (CI) and is confined to one adiabatic surface. The only numerical work required is to add and subtract nuclear wave functions computed with normal and GP boundary conditions. This is equivalent to unwinding the nuclear wave function onto a double cover space, which separates out two components whose relative sign is changed by the GP. By referring to earlier work on the Aharanov-Bohm effect, we show that these two components contain all the Feynman paths that follow, respectively, an even and an odd number of loops around the CI. These two classes of path are essentially decoupled in the Feynman sum, because they belong to different homotopy classes (meaning that they cannot be continuously deformed into one another). Care must be taken in classifying the two types of path when the system can enter the encirclement region from several different start points. This applies to bimolecular reactions with identical reagents and products, for which our approach allows a symmetry argument developed by Mead [J. Chem. Phys. 72, 3839 (1980)] to be generalized from nonencircling to encircling systems. The approach can be extended in order to unwind the wave function completely onto a higher cover space, thus separating contributions from individual winding numbers. The scattering boundary conditions are ultimately what allow the wave function to be unwound from the CI, and hence a bound state wave function cannot be unwound. The GP therefore has a much stronger effect on the latter than on the wave function of a reactive system.