Carbon dioxide in the martian seasonal polar caps can metamorphose into an annealed polycrystalline layer, provided freshly condensed grains are smaller than (10 \mu \mathrm{m}). Densification is not expected for coarse-grained (grain size larger than (10 \mu \mathrm{m}) ) (\mathrm{CO}{2}) snow. These conclusions are based on an analysis of the microphysical processes involved in pressureless sintering. The presence of low porosity (\mathrm{CO}{2}) deposits over large areas could lead to the millimeter- to centimeter-sized path lengths required to explain several weak absorption features observed in the (1.5-) and (2.3-\mu \mathrm{m}) spectral regions, provided grain boundaries scatter light very inefficiently and the layer is sufficiently flawless, i.e., devoid of scattering centers such as air bubbles, cracks, and impurities. The latter condition may well be satisfied in the seasonal caps on Mars, where (\mathrm{CO}{2}) condenses in the near absence of an inert gas, the cap temperature is almost constant, and the average abundance of impurities is not expected to exceed (0.1 %) by volume. A nonporous layer may also explain the low thermal emissivity inferred from Viking measurements. In addition, an annealed layer could become sufficiently transparent during the later stages of spring sublimation to allow the albedo dichotomy between the northern and southern cap to be associated with the underlying substrate and thus be permanent; this would explain the exposure of the northern residual (\mathrm{H}{2} \mathrm{O}) cap even during years when no great dust storms occur. The measurements to be carried out by instruments onboard the Mars Observer spacecraft may help to distinguish locations where fine-grained frost is forming on the ground from areas covered by coarse-grained (\mathrm{CO}_{2}) snow. (1993 Academic Press, Inc.