Isn't this all the wrong way around? Shouldn't the dipole moments perpendicular to the interface cancel, but those parallel to it reinforce?