Why are alkali halide surfaces not wetted by their own melt?

Alkali halide (100) crystal surfaces are anomalous, being very poorly wetted by their own melt at the triple point. We present extensive simulations for NaCl, followed by calculations of the solid-vapor, solid-liquid, and liquid-vapor free energies showing that solid NaCl(100) is a nonmelting surface, and that its full behavior can quantitatively be accounted for within a simple Born-Meyer-Huggins-Fumi-Tosi model potential. The incomplete wetting is traced to the conspiracy of three factors: surface anharmonicities stabilizing the solid surface; a large density jump causing bad liquid-solid adhesion; incipient NaCl molecular correlations destabilizing the liquid surface. The latter is pursued in detail, and it is shown that surface short-range charge order acts to raise the surface tension because incipient NaCl molecular formation anomalously reduces the surface entropy of liquid NaCl much below that of solid NaCl(100).