A method is shown for calculating vapor pressures over a CMAS droplet in a gas of any composition. It is applied to the problem of the evolution of the chemical and Mg and Si isotopic composition of a completely molten droplet having the composition of a likely refractory inclusion precursor during its evaporation into the complementary, i.e. modified solar, gas from which it originally condensed, a more realistic model than previous calculations in which the ambient gas is pure H 2(g). Because the loss rate of Mg is greater than that of Si, the vapor pressure of Mg (g) falls and its ambient pressure rises faster than those of SiO (g) during isothermal evaporation, causing the flux of Mg (g) to approach zero faster and MgO to approach its equilibrium concentration sooner than SiO 2. As time passes, δ 25Mg and δ 29Si increase in the droplet and decrease in the ambient gas. The net flux of each isotope crossing the droplet/gas interface is the difference between its outgoing and incoming flux. δ 25Mg and δ 29Si of this instantaneous gas become higher, first overtaking their values in the ambient gas, causing them to increase with time, and later overtaking their values in the droplet itself, causing them to decrease with time, ultimately reaching their equilibrium values. If the system is cooling during evaporation and if mass transfer ceases at the solidus temperature, 1500 K, final MgO and SiO 2 contents of the droplet are slightly higher in modified solar gas than in pure H 2(g), and the difference increases with decreasing cooling rate and increasing ambient pressure. During cooling under some conditions, net fluxes of evaporating species become negative, causing reversal of the evaporation process into a condensation process, an increase in the MgO and/or SiO 2 content of the droplet with time, and an increase in their final concentrations with increasing ambient pressure and/or dust/gas ratio. At cooling rates <∼3 K/h, closed-system evaporation at P tot ∼ 10 −3 bar in a modified solar gas, or at lower pressure in systems with enhanced dust/gas ratio, can yield the same δ 25Mg in a residual CMAS droplet for vastly different evaporated fractions of Mg. The δ 25Mg of a refractory residue may thus be insufficient to determine the extent of Mg loss from its precursor. Evaporation of Mg into an Mg-bearing ambient gas causes δ 26Mg and δ 25Mg of the residual droplet to fall below values expected from Rayleigh fractionation for the amount of 24Mg evaporated, with the degree of departure increasing with increasing fraction evaporated and ambient pressure of Mg. δ 26Mg and δ 25Mg do not depart proportionately from Rayleigh fractionation curves, with δ 25Mg being less than expected on the basis of δ 26Mg by up to ∼1.2‰. Such departures from Rayleigh fractionation could be used in principle to distinguish heavily from lightly evaporated residues with the same δ 25Mg.