Evaluating the effect of grain size and salts on liquid water content in frozen soils of Antarctica by combining NMR, chemical equilibrium modeling, and scattered diffraction analysis

•We used 2H NMR to determine liquid water content in soils of Antarctica.•We compared the results from NMR and simulations based on chemical equilibrium.•We correlated freezing points with grain size measurements.•Results point to a large effect of soil texture.•Reaction kinetics of salt formation is also an important factor.

Recent studies point to the possible existence of liquid water on Mars surface. While many efforts have focused on experimental determination and modeling of liquid water content in inorganic-salt rich soils, additional factors that govern liquid water formation remain largely unexplored. The purpose of this study is to determine main soil properties that will affect unfrozen water content. We analyzed liquid water content and bulk freezing points using high field deuterium NMR measurements and compared the results to the amount of liquid water content at the same temperatures based on salt concentration and composition using an empirical chemical equilibrium model (FREZCHEM). We selected soils with a wide range of salt concentrations and composition collected from the McMurdo Dry Valleys, Antarctica. The comparison indicates that at 50% unfrozen water content the freezing temperatures deviate substantially from water content modeled based on salt composition. In particular, the NMR freezing points are lower than those calculated with FREZCHEM for all low to medium salt content samples. The results indicate that soil properties other than aqueous chemical composition add to liquid water content. Particle size does not necessarily correlate with the liquid water content, revealing that more detailed investigation including surface area and composition may be necessary to describe the dependence of unfrozen water content on soil properties. The study highlights the necessity of experimental measurements of water content due to effects of multiple factors that are difficult to take into account in simulations.