Impact of Mineral Precipitation on the Pore Network Structure of Sediment

Abstract

Mineral precipitation and dissolution reactions in subsurface porous media may alter the porosity and permeability of the system. While the impacts of mineral dissolution on the pore network structure are generally well understood, the effects of mineral precipitation are much more complex and difficult to predict because there is little understanding of where, within a single pore and a network of pores, precipitation will occur. This thesis focuses on advancing our understanding of mineral precipitation reactions and their corresponding impacts on reactive transport in subsurface sediments. This work was conducted in application to caustic radioactive wastes leaking from tanks at the former nuclear weapons production site in Hanford, WA. Reactions there produce precipitates that can substantially alter flow as well as sequester radionuclides. Using a previously- reacted experimental column from a collaborator at PNNL, 2D Scanning Electron Microscopy (SEM) imaging and reactive transport modeling were combined with a suite of new methods developed for this project to infer impacts on flow and radionuclide release. At PNNL, the column was filled with Hanford sand and reacted with simulated tank waste, and the column was imaged periodically using 3D computed micro- tomography (CMT) at the NSLS. Afterwards, the column was shipped to Princeton where, as part of this study, the pore network structure was preserved by flowing epoxy through the column. SEM imaging of loose reacted sand grains showed two morphologies of secondary mineral precipitates and revealed these precipitates may not be chemically bonded to the grain surface. SEM imaging of column cross-sections revealed secondary cancrinite precipitates as a relatively uniform coating on all grain surfaces. To determine how mineral precipitates altered the pore network structure, a set of cross-sectional images were stitched together to cover an area of 7.5 mm2, and from these a second set of images was created by digitally removing the mineral precipitates to approximate the before-reaction condition. Both sets of images were analyzed using an erosion-dilation image analysis method to compute the pore and throat size distributions. To correct for the inherent bias in the 2D analysis, a method was developed to bias- correct the size distributions. To predict the column permeability and estimate the impact of mineral precipitation, pore network models were informed using the size distributions. The predicted permeabilities were compared to permeabilities inferred from network models that had been informed from size distributions determined from 3D X-ray CT images. Despite only one order of magnitude difference in image resolution for the 2D and 3D images, the predicted permeabilities differed by over five orders of magnitude. The sensitivity of predicted permeability to image resolution was tested and it was found that small changes in image resolution, on the order of microns, resulted in order of magnitude changes in predicted permeability. At the 4 μm 3D image resolution, small pore throats are under-estimated and the predicted permeability is over-estimated. At the 0.4 μm 2D image resolution, surface roughness features are misinterpreted as small pore throats and permeability is under-estimated. 2D SEM imaging also revealed Hanford sand grains have a large amount of intragranular porosity and that secondary mineral precipitates were observed in this intragranular space. This observation suggests radionuclides will be sequestered in intragranular regions, which may decrease their mobility in migrating water. However, as the system recovers and uncontaminated pore water contacts the reacted sediments, secondary contamination will occur as the radionuclides desorb from soils. To examine this, the long-term leaching of Cs from intragranular pores in contaminated sediments was estimated. Two reactive transport model scenarios were investigated, one accounting for Cs sorption in intragranular pores and one that includes only sorption to surfaces in contact with the bulk pore fluid. These models reveal that Cs leaching is slow and requires 105 to 591 days to desorb all exchangeable Cs from a soil segment. In addition, Cs sorbed to soil downstream will not desorb until the depletion of upstream exchangeable Cs. Intragranular pores will prolong the period of secondary contamination at the site, as desorption of Cs sorbed in intragranular pores will not begin until the depletion of Cs sorbed to the bulk soil, further elongating the time to begin downstream Cs desorption.