We generalize and apply a microscopic force level statistical
mechanical
theory of activated spherical penetrant dynamics in glass-forming
liquids to study the influence of semiflexible polymer connectivity
and penetrant–polymer attractive interactions on the penetrant
hopping rate. The detailed manner that attractions of highly variable
strength and spatial range modify the penetrant size and polymer melt
density (from the rubbery state to slightly beyond the kinetic glass
transition) dependences of penetrant activation barriers is established.
Of special interest are possible nonadditive consequences of physical
bonding and steric caging, the degree of coupling of penetrant hopping
and the Kuhn segment scale alpha relaxation process, the relative
importance of local caging and long-range matrix collective elasticity
as a function of penetrant size, and implications for optimizing transport
selectivity. With increasing attraction strength, the repulsive caging-restriction
effect on penetrant mobility is predicted to grow, in contrast to
the effect of the equilibrium penetrant–matrix solvation shell
size, which decreases. The former dynamical effect results in a significant
enhancement of the importance of the local cage barrier, while the
latter effect results in a decrease of the importance of the nonlocal
collective elastic barrier. These two competing effects have a very
strong influence on selective penetrant transport for different sized
penetrants: selectivity varies nonmonotonically with attraction strength
in the deeply supercooled state but decreases monotonically in the
rubbery state and at fixed attraction strength, exhibits a nonmonotonic
variation with the matrix packing fraction. By comparing results based
on modeling the matrix as semiflexible polymer chains with analogous
calculations using the same dynamical theory but for a disconnected
hard sphere matrix, the effect of chain connectivity is revealed and
found to have quantitative, but not qualitative, consequences on penetrant-activated
dynamics.