Résumé : Low ionic strength waters containing significant calcium and bicarbonate are common in nature, but little literature exists on their behaviour during freezing. Modelling indicates that freezing-induced concentration of solutes (in a closed-system) would lead to progressive increase in calcite saturation index, despite rising partial pressure of CO2 (PCO2), but the consequences of CaCO3 precipitation for the distribution of matter between solid, liquid, and gas phases required experimental investigation. We studied the effects of variations in the rate of advance of an ice-water interface and in the initial degree of saturation for calcite on the behaviour of the system. Downward growth of ice in a 24-cm diameter cylindrical vessel was achieved at a constant linear rate of 3 or 8 mm/h by the progressive cooling of an overlying alcohol reservoir, and the expansion of volume accommodated by regular water sampling through side ports, together with a small expansion chamber. Initial air-saturated solutions (initial PCO2 in the range 10-3 to 10-3.2) were prepared to reflect a range from strongly undersaturated to supersaturated for calcite. Comparative blank experiments were run using deionized water. Ice growth led to enrichment in solutes at the ice-water interface and the creation of a diffusive boundary layer, calculated to be 0.6 mm thick, truncated below by convecting fluid. The first-formed ice (stage 1), was relatively solute-rich because of initial rapid ice nucleation. Where solutions were not strongly supersaturated for calcite this was followed by formation of a solute-poor (stage 2) ice. Ice-interface water segregation coefficients of stage 2 ice were calculated to be 0.0004-0.003 for various solute ions. The relative magnitude of segregation coefficients (Mg2+ > Ca2+ > Sr2+) is attributed to interstitial incorporation (coupled with HCO3-) in the ice lattice, and controlled by ion size. Air bubbles nucleated once nitrogen supersaturation had reached values of 2-2.5 in the boundary layer and were incorporated into ice. These gas inclusions had dissolved air compositions modified by the differential diffusion of O2, N2, and CO2 out of the boundary layer, an O2/N2 ratio of 0.4 being characteristic. Freezing of solutions strongly supersaturated for calcite led to formation of impure (stage 3) ice in which ions are incorporated in similar proportions to those of the parent aqueous solution. Stage 3 ice contains both solid CaCO3 and aqueous (solute-rich) inclusions, associated with an irregular ice-water interface. Gas inclusions were invariably rich in CO2, up to 63% by volume, yet represented only a small proportion of the CO2 generated as a by-product of CaCO3 precipitation. These data allow a better understanding of the expected chemical characteristics of ice that has formed from freezing of bulk water, including river icings, basal ice of glaciers, and local refrozen layers in firn and glacier ice. Generation of CO2-rich gas bubbles by re-freezing is a powerful mechanism for modification of CO2 compositions of bulk gaseous inclusions in ice.