I compare the diffusion coefficients of different substances in polymers. While D in glassy polymers tend to correlate better with the kinetic diameter (size dominates), rubberlike polymers tend to correlate better with the boiling point (solubility dominates). I noticed that in both cases H2O and CO2 are unusual outliers ... When I plot the kinetic diameter of many different molecules against the boiling point, I see (as you would expect) an approximately linear correlation. Carbon dioxide (if you take the sublimation point) and water seem to be clear outliers. Why?