The Flow Behaviour of Elemental Mercury DNAPL in Porous Media

This thesis focuses on the characterization of the flow behaviour of elemental mercury (Hg0) DNAPL (Dense NonAqueous Phase Liquid) in porous media. In the subsurface, Hg0 DNAPL can act as a long lasting source of contamination, causing detrimental consequences for the human health and the environment. Therefore, insight into the flow behaviour of elemental mercury in porous media is needed and is critical in assessing the control of contaminant spreading as well as remediation approaches. However, the scientific literature on Hg0 DNAPL is still very limited and, to date, it remains unclear to what extent the validity of the classical constitutive relations, used to describe DNAPLs flow behaviour, as well as the scaling theory, holds for elemental mercury. These issues become crucial in partially water saturated porous media, where liquid Hg0 is likely to behave as a nonwetting phase with respect to both air and water. To address these knowledge gaps, experimental and numerical analysis were performed. In particular, the properties affecting the constitutive relations governing liquid Hg0 infiltration behaviour were explored first, using capillary pressure-saturation, Pc(S), experiments in different granular porous media, and in two- and three-phase fluid systems. Then, the infiltration and (re)distribution behaviour of Hg0 DNAPL was studied in variably water saturated stratified porous media with flow container experiments and dual gamma ray measurements of porosity and fluid saturations. Experimental results indicated that elemental mercury infiltration is strongly controlled by the porous medium water content and can be triggered by its changes due to, for example, rain events. Finally, a new theoretical formulation of elemental mercury retention properties in variably water saturated porous media was proposed and, to assess to what extent numerical modelling can predict elemental mercury migration in porous media, the flow container experiments were simulated using GDAn, the code developed by the Author.