ARTICLE pubs.acs.org/est
Remobilization of Residual Non-Aqueous Phase Liquid in Porous Media by FreezeThaw Cycles Kamaljit Singh,†,‡ Robert K. Niven,*,† Timothy J. Senden,§ Michael L. Turner,§ Adrian P. Sheppard,§ Jill P. Middleton,§ and Mark A. Knackstedt§ †
School of Engineering and Information Technology, The University of New South Wales at ADFA, Northcott Drive, Canberra, ACT, 2600, Australia § Department of Applied Mathematics, Research School of Physics and Engineering, Australian National University, ACT, 0200, Australia
bS Supporting Information ABSTRACT: The pore-scale behavior of a nonaqueous phase liquid (NAPL) trapped as residual contamination in a porous medium, subject to freezethaw cycles, was investigated by X-ray microcomputed tomography. It is shown that freezethaw cycles cause significant NAPL remobilization in the direction of the freezing front, due to the rupture and transport of a significant proportion of (supposedly entrapped) larger multipore NAPL ganglia. Significant NAPL remains in place, however, due to substantial ganglion fragmentation into single- and subpore ganglia. The contraction of branched ganglia into more rounded forms, especially near the top surface, is also observed. Three freezing-induced mechanisms are proposed to explain the results. The findings have important implications for NAPL contamination in cold regions, and for the behavior of waterhydrocarbon systems on the Earth and other planets.
’ INTRODUCTION In recent years, there has been significant interest in the monitoring and remediation of non-aqueous phase liquid (NAPL) contamination in cold regions. Such contamination is often associated with the accidental surface/subsurface release of hydrocarbon fuels, oils and solvents associated with fuel storage facilities, pipelines, vehicle maintenance operations, and the chemical industry.14 Once NAPLs are spilt on a typical (unfrozen) soil, they percolate downward due to gravity, leaving behind socalled residual saturation consisting of disconnected NAPL blobs or globules called ganglia, occupying less than one to several pore cavities.5 If entrapped in the saturated zone, such NAPL ganglia are held in position by strong interfacial forces, and are generally regarded as immobile.6,7 These become a long-term secondary source of pollution by slow dissolution of the NAPL into the surrounding groundwater by diffusion-limited processes, over periods of years to decades.8 The entrapment of a residual gas phase (such as air or methane) follows an identical mechanism, differing primarily in the higher buoyancy, surface tension, and compressibility of the (bubble) ganglia compared to NAPL systems.9 The presence of trapped NAPL in soils which undergo diurnal or seasonal freezethaw cycles creates a four-phase (water iceNAPLsoil) system, of which the pore-scale behavior is still poorly understood. It is well-known that natural (NAPL-free) soils, subject to freezethaw cycles, exhibit complicated, history-dependent r 2011 American Chemical Society
behaviors due to the 9% volumetric expansion of water on freezing, which induces irreversible changes to the soil fabric, stratigraphy, moisture content, and engineering properties.1015 Most freeze thaw studies on organic contaminants are limited to low-level (nonimmiscible) contamination or solvent exsolution effects,16,17 NAPL spills on frozen soils, permafrost and ice/snow,3,1823 and bulk NAPL behavior in freezing and thawing soils.22,2428 A few nondestructive studies have been conducted on pore-scale effects, including visualization studies in two-dimensional cells,16,24,2932 which provide a useful understanding of mechanisms but are unable to simulate three-dimensional effects, and one X-ray tomography study at lower resolution,33 in which the NAPL ganglia were not resolved. The effects of successive freezethaw cycles on the pore-scale NAPL spatial distribution, ganglia morphology, and availability for dissolution—and hence on contaminant fate and transport—are virtually unknown. Interpreting these pore-scale phenomena will provide important new insights to our understanding of multiphase (fluidfluid/solid) systems in the vicinity of the waterice phase boundary, including (i) the behavior of natural waterice hydrocarbon systems on the Earth and other planetary bodies,34 and (ii) the entrapment and release of liquid hydrocarbons and methane Received: January 16, 2011 Accepted: March 16, 2011 Revised: March 14, 2011 Published: March 28, 2011 3473
dx.doi.org/10.1021/es200151g | Environ. Sci. Technol. 2011, 45, 3473–3478
Environmental Science & Technology gas from continental permafrost and suboceanic gas hydrates due to natural processes and global warming.3538 We here report the findings of high-resolution three-dimensional visualization of a model porous medium containing residual NAPL, subject to successive freezethaw cycles, using X-ray microcomputed tomography (XCT). This method allows precise and accurate quantitative analysis of phase distributions, and characterization of the size and shape distributions of NAPL ganglia and the porous medium.39,40 A light NAPL (LNAPL), less dense than water, was selected as the immiscible liquid. We observed substantial freezing-induced rupture of NAPL ganglia and NAPL transport in the direction of the freezing front; significant fragmentation of NAPL during freezing to form numerous small ganglia; and the contraction of many branched ganglia to form more rounded shapes, primarily in the upper part of the sample. Three freezing-induced mechanisms, involving pore expulsion, ganglion cutoff, and ganglion contraction, are proposed to explain these results.
’ MATERIALS AND METHODS Materials. A 50% by mass mixture of dodecane (g99%, Sigma-Aldrich, St Louis, MO) and 1-iodononane (95%, SigmaAldrich), here termed ID-50, was selected as the NAPL. This was chosen for its contrast in relative X-ray intensity (1.25) compared to water (0.37) and glass (0.87), permitting phase segmentation by XCT, as well as its lower density than water (950 kg m3), comparable interfacial tension (44.9 ( 2.2 mN m1) to most hydrocarbonwater systems, low aqueous solubility, and low volatility. Deionized and deaired water (electric conductivity e 1.0 μS/cm, Purite Ltd., Oxon) was used as the aqueous phase. The porous medium consisted of 0.5-mm diameter soda lime glass beads (Biospec Products, Inc., OK), prewashed and of sieve interval 425600 μm. Methodology Overview. Experiments were conducted in a cylindrical glass cell (16.4 mm internal diameter, 49 mm height), filled with glass beads and water-saturated. ID-50 was then established at an overall mean residual saturation (volume fraction of NAPL-occupied pores) of 15.4%, to simulate “naturally” trapped NAPL. The cell was then insulated and subjected to unidirectional (top downward) freezing and thawing cycles, to simulate natural meteorologically driven temperature changes. The system was imaged by XCT at a voxel side length resolution of 11.83 μm, at residual saturation and after freezethaw cycles 1, 5, and 10. Due to XCT schedule limitations, only the upper 1900 voxel interval (22.48 mm height) of the cell could be imaged. After preprocessing and phase identification,41 the segmented tomographic data were visualized in three dimensions using the Drishti volume rendering software,42 and analyzed quantitatively to determine the spatial distributions and volumes of NAPL, water, and solid phases, using the Mango software toolkit developed at the Australian National University, Canberra.43 Further details of the experimental and characterization methods are described in ref 31 and summarized below. Residual Saturation. The cell was filled with glass beads to bed height 42.12 mm and total internal volume 8.897 mL (including XCT-imaged volume of 4.748 mL), which was water-saturated, and a residual saturation of ID-50 then established. To attain this, CO2 gas (laser grade, BOC Ltd., Botany, Australia) was first passed through the air-filled cell from the base for 30 min to displace air. More than 10 pore volumes of water were then pumped into the cell from the base using a digital
ARTICLE
controller (GDS Instruments Ltd., Hampshire, UK) at 2.0 mL/ min and then at 0.2 mL/min. The water was then drained by pumping ID-50 using a syringe pump (Harvard PHD 2000 Programmable Model) into the top of the cell to obtain a stable downward displacement at 0.2 mL/min. When the NAPL displacement front had reached the lower nylon filter, the flow was reversed by supplying water into the base of the cell, first at 2.0 mL/min until it reached the top of the cell, and then at 0.2 mL/min for another 10 pore volumes. The mean residual NAPL saturation in the entire cell, determined gravimetrically, was 15.4% (consistent with ref 44); this differed slightly from the XCT-quantified NAPL saturation of 16.3% in the imaged region. After residual saturation was established, the water level was maintained at the surface of beads throughout the experiment by a constant head technique.45 The cell was then imaged by XCT (see below). FreezeThaw Cycles. The outside of the cell was insulated carefully with cut pieces of high density polystyrene, with a hole in the top to permit freezing and thawing from the top downward. The cell was then placed inside an environmental chamber (Challenge 600 series, Angelantoni Industrie S.p.A., Italy), and subjected to freezing and thawing cycles by control of the chamber temperature. In all cycles, the temperature of the cell was first maintained at a uniform temperature of 2.64.0 °C for several hours, then set in the range 2.0 to 2.5 °C, causing cooling in the range 4.765.84 °C/h and 1.842.48 °C/h, respectively, at 10 mm and 36 mm from top of the cell, which diminished over the period of 35 h. In all cases, the two temperature probes indicated nucleation commenced at the top of the cell and progressed downward. The temperature was then maintained at this temperature for sufficient time to permit freezing, and then set in the range þ8.0 to þ9.0 °C, to induce thawing. X-ray Microcomputed Tomography. The cell was imaged at residual saturation and after freezethaw cycle 1, 5, and 10, using a purpose-built XCT instrument.39,40 This facility is built on a 3-m optical rail and consists of three parts: a microfocus X-ray source (X-Tek RTR-UF225 operated at 80 kV and 120 μA), a rotation stage, and an X-ray camera (140-μm thick CsI crystal scintillator and 20842 pixel grade 1 Kodak KAF4302E CCD). A 3-mm silica filter was used to minimize beam hardening artifacts. The cell was clamped on the rotation stage, and a series of radiographs (2-dimensional arrays of horizontal lines through the specimen) were captured by 14.5-s X-ray exposures, with rotation of the cell in 2880 steps of 0.125°, giving a total acquisition time of 1012 h. The limit of resolution was a voxel (cube) of side length 11.83 μm, equivalent to a volume of 1.7 109 mL. Phase Quantification. The XCT projection data were first preprocessed and reconstructed using the standard cone-beam FeldkampDavisKress algorithm.46 The image was then processed for phase identification using the Mango software toolkit,43 including tomogram alignment and tilting, anisotropic diffusion filtering, cropping (masking) to remove the temperature probes and external walls, and a two-stage segmentation of the NAPL, water, and solid phases. Further details are given in Singh.31 The spatial distributions and volumes of each phase were quantified, and the segmented data were used for visualization purposes and further analyses. A sensitivity analysis of the choice of X-ray intensity threshold used to discriminate between NAPL and other phases, to examine the effect of the “halo” around the NAPL, indicated possible errors of þ4.65% and 4.75% in the quantified total NAPL volume. In contrast, the 3474
dx.doi.org/10.1021/es200151g |Environ. Sci. Technol. 2011, 45, 3473–3478
Environmental Science & Technology
Figure 1. Spatial distribution of NAPL in XCT imaged region (diameter 16.4 mm, height 22.48 mm) at (a) residual saturation and (b) after freezethaw cycle 10, viewed from the same position (glass and water phases not shown). The core boundary location is approximate.
phase volumes were much less sensitive to the sharper water solid X-ray intensity threshold, giving an error in porosity of (0.83%. Integration of all phases indicated a total volume of 4.623 mL in the imaged region, consistent—allowing for tomogram cropping—with its calculated volume of 4.748 mL. Tomogram Quantitative Analysis. The XCT data were further processed using a watershed algorithm applied to the Euclidean distance transform of the segmented pore space to identify common regions occupied by each phase and the properties of the porous medium.47 A subsequent region-merging step was required to connect adjacent regions that were considered insufficiently distinct. A new ganglion counting algorithm was also developed to obtain information on the size distribution and types of NAPL ganglia, which the authors believe has not been reported in the literature. First, the water and solid phases were merged in a separate data file. Second, the frequency distribution of NAPL ganglia was analyzed by counting the numbers of adjoining NAPL voxels in each pore and assessing the connectivity between each pore and its 26 nearest neighbors (all faces, edges and corners). By this process, each ganglion was identified and labeled, and its volume and the number of pores it occupied were determined. For this study, a criterion of 10% of the volume of a pore was used to determine whether it was “occupied” by a ganglion, thereby sorting the data into integer-valued pore occupancy bins. Analyses using different criteria (up to 90%) indicated only small differences (almost always e2) in the number of pores occupied by each ganglion. The raw ganglion volumes were binned on a logarithmic scale. Analysis of the raw data indicated virtually no discrepancy (
