Evaluation of Equilibrium Solubilities: Statistics Versus Selected Experiments

Abstract Using the examples of the solubility of NaCl and NaNO 3 in water, the statistical treatment of all published data is compared with a selection of reports of 2 – 4 authors of particularly carefully performed and described solubility determinations. The uncertainties in the latter case are up to ten times smaller. For t < 100 °C, such accurate work was published at the end of the 19th century. The temperature, when anhydrite, CaSO 4 , in contact with water starts to form gypsum, CaSO 4 ·2H 2 O, and vice versus can be predicted by the temperature of intersection of the solubility curves of both minerals. Exact knowledge of this temperature is of interest for the geoscience of evaporitic rocks and tunnel construction planning through sulfate-containing rocks. However, the uncertainty resulting from separate statistical treatment of the solubility data of gypsum and anhydrite an uncertainty, which is too large for fixing the temperature of gypsum/anhydrite transition within a 2–3 K range. Experimental solubility determinations with focus on the intersection temperature are superior to statistical treatments and yield a temperature of (42.1 ± 1.5) °C, which is supported by independent caloric measurements. In the system MgSO 4 – H 2 O, a series of stable hydrates occurs along the solubility curve of magnesium sulfate in dependence on temperature. Above 68 °C, the monohydrate represents the stable phase, known as mineral kieserite, which is found in evaporitic rocks and was formed at ambient temperatures in solutions rich in MgCl 2 . Large amounts of magnesium sulfate hydrate on the surface of the planet Mars raise the question of whether the monohydrate represents a primary factor for water distribution control. Due to kinetic difficulties in achieving solubility equilibrium in the laboratory at low temperatures, thermodynamic modelling is applied to predict the low temperature limit for kieserite formation. It is shown that experimental evidence is still missing to confirm the model’s predictions.