A modeling work has been conducted on a phenomenon called post ion milling (PIM), a post-treatment of Ar+ ion sputtering to modify nanostructures on solid surface. It was found by experiments that for PIM with a sufficiently low ion flux, both the average dot size and the surface roughness of Si nanodot arrays on Si(1 0 0) decline steadily against milling time. However, the usually adopted Kuramoto-Sivashinsky (KS) model involving the Bradley-Harper (BH) theory failed to explain the experimental results, nor the KS model that combines both the BH and Ehrlich-Schwoebel (ES) effects. We reexamined the ES term in the KS equation, and derived new terms reflecting the ES contribution. With such a modification, the KS model involving both the BH and the refined ES effects finally gave a qualitative explanation to the PIM result. (c) 2007 Elsevier B.V. All rights reserved.