Environ. Sci. Technol. 2006, 40, 6064-6069
Epi-Fluorescence Imaging of Colloid Transport in Porous Media at Decimeter Scales P E N G F E I Z H A N G * .†.‡ A N D Y O N G G A N G W A N G †,‡ Department of Earth & Atmospheric Sciences, City College of New York, 138th Street and Convent Avenue, New York, New York 10031, and Department of Earth and Environmental Sciences, Graduate School & University Center, City University of New York, 365 Fifth Avenue, New York, New York 10016
A noninvasive epi-fluorescence imaging technique was developed for real-time observation of colloid transport in porous media at decimeter scales. Fluorescent latex microspheres and translucent quartz sand were used as a model colloid-porous medium system. Various calibrations were performed for accurate conversion of fluorescence intensities to microsphere concentrations. Fluorescence intensities were found to linearly increase with microsphere concentrations (5 × 105-5 × 108 spheres/mL in saturated sand) and with camera exposure time. Fluorescence intensities also increased with sand thickness (saturated with microsphere solution), indicating that the fluorescence signals detected by the imaging system were integrated signals from the entire thickness (10 mm) of the sand. A set of microsphere transport experiments was conducted to demonstrate the versatility of the imaging system. Excellent mass recoveries (93-103%) were achieved in all transport experiments, demonstrating the robustness of the imaging system for quantitative study of colloid transport. The system allowed the change of flow velocity, ionic strength, and flow direction within one transport experiment and the realtime, quantitative monitoring of the movement of microspheres in packed sand, greatly reducing the time and effort needed for similar work with traditional column experiments.
Introduction Colloid transport in porous media has been traditionally studied via column experiments, where colloids are injected and eluted, and colloid breakthrough-elution curves and/or colloid distributions within the media are obtained and analyzed (e.g., refs 1-3). Data obtained from column experiments are restricted to one dimension (1D), and colloid distribution profiles are limited to the end of the column experiments due to destructive sampling. Such column experiments are clearly inadequate for studying colloid transport in heterogeneous media, where direct and continuous observation of the two-dimensional (2D) movement of colloids within the media at macroscopic scales is essential. Recently, a light transmission technique was used to directly observe the reactive transport of solutes and colloids * Corresponding author phone: (212)650-5609; fax: (212)650-6482; e-mail:
[email protected]. Corresponding author address: Department of Earth & Atmospheric Sciences, City College of New York, 138th Street and Convent Avenue, New York, NY 10031. † City College of New York. ‡ City University of New York. 6064
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 19, 2006
in translucent porous media at decimeter scales (4-9). In this technique, a light source was placed on one side of a flow chamber, and bioluminescence or fluorescence from the flow chamber was captured by a charge-coupled device (CCD) camera placed on the opposite side of the flow chamber. Fluorescent (4, 7-9) or ultraviolet (UV) (5, 6) tubes were employed as a diffuse light source, and a large gel filter sheet (decimeters in one dimension) was typically placed in front of the flow chamber to restrict excitation wavelengths. Large gel filter sheets tend to have defects and allow some source light to leak through (8), leading to increased light background and decreased sensitivity. A large distance between the flow chamber and the camera lens (0.9-4 m, depending on the size of the imaging area (6, 8)) was required for a typical light transmission imaging system. As such, a darkroom was necessary to house the entire imaging system for reduced light background. The flow chamber was often fixed at a vertical direction (camera in horizontal) due to the space needed to lay out the entire system, making it difficult to pack porous media with various types and levels of heterogeneity when such heterogeneity is desired (9). Image intensities recorded by the CCD camera must be converted to real solute or colloid concentrations for quantitative analyses of solute/colloid transport. The application of Beer-Lambert law (10) to a system with quartz sand and fluorescent colloids (e.g., ref 8 and current study) is not straightforward, due to light refraction, light reflection, and the lack of information on fluorescence quantum yield. Therefore, empirical relationships between image intensities and colloid concentrations (i.e., calibrations) must be developed. Calibrations are particularly important for porous media with physical heterogeneity (e.g., hydraulic conductivity contrasts), since light transmission through areas of different hydraulic conductivity may be different (9) and light intensity is no longer conserved during transport. As such, calculating colloid concentrations based on the principle of light conservation (8) is no longer valid. The objective of this study was to explore an alternative imaging technique that would provide a compact, easy to calibrate, and highly sensitive imaging system for quantitative study of colloid transport in porous media at decimeter scales. A light reflection imaging technique (both the light source and the CCD camera were placed on the same side of the sample) was used for a compact configuration. A relatively small flow cell was employed for easy calibrations. Optical emission and excitation filters of common sizes were used for a large selection of filter pairs that may reduce light background and enhance sensitivity. Thorough calibration processes were performed, including intensity vs concentration in porous medium, intensity vs camera exposure time, intensity vs medium thickness, intensity vs irradiation duration, and intensity vs aqueous concentration. A few transport experiments were carried out to demonstrate the utility of the epi-fluorescence imaging system for quantitative monitoring of the movement of fluorescent microspheres in translucent sand under changing ionic strengths, flow rates, and flow directions. A small subset of the transport data is presented in this paper for illustration purposes. The paper does not purport to present new models of colloid transport based on real-time imaging data.
Materials and Methods Imaging System. The imaging system consists of four major components: (1) an illuminator, (2) a closed optical path image (COPI) chamber, (3) a capture system chamber, and 10.1021/es061104y CCC: $33.50
2006 American Chemical Society Published on Web 08/24/2006
FIGURE 1. Schematic diagram of the epi-fluorescence imaging system: (1) illuminator, (2) closed optical path image (COPI) chamber, (3) capture system chamber, and (4) sample chamber. (4) a sample chamber (Figure 1). The illuminator uses a 150 W halogen lamp as the light source and is interfaced with the COPI chamber by a filter slider that holds up to four optical excitation filters (18 mm diameter). The COPI chamber has a set of fiber optics and small mirrors that distribute the filtered light from the light source evenly onto a platen (a 20 × 20 cm viewing window made of optical grade UV transmittable acrylic), where a sample is excited and fluorescence is emitted. A large mirror in the COPI chamber, positioned 45° from the platen and the camera, reflects the downward fraction of the emitted fluorescence to the capture system chamber (i.e., epi-fluorescence) (Figure 1). The capture system chamber is composed of a filter wheel that holds up to four emission filters (72 mm diameter), a 10× zoom lens that is capable of capturing images from 20 × 20 cm (200 µm/pixel) to 2 × 2 cm (20 µm/pixel), and a cooled 14-bit CCD camera with a chip size of 1024 × 1024 pixels. The sample chamber (working area underneath the lid) has fittings that connect the flow cell inside to the pump and fraction collector outside (Figure 1). The optical arrangement of epi-fluorescence (light reflection) made it possible to place all necessary components of the imaging system into a compact (91 cm long by 48 cm wide by 43 cm high), light-tight box, eliminating the need for a darkroom. The entire imaging system (Figure 1) can be easily rotated so the orientation of the flow chamber can be changed (horizontal or vertical) as desired. Raw images were processed using the Kodak 1D Image Analysis Software (Eastman Kodak, Rochester, NY). The software provides the intensity information on each pixel and gives the total and mean intensity of a selected region of interest. Illumination nonuniformity is highly reproducible with this imaging system and can be corrected by dividing a sample image by an illumination reference image using the reference correction function in the Software. Flow Cell. The flow cell has internal dimensions of 180 mm (length) by 100 mm (width) by 10 mm (depth), with three inlets and three outlets evenly spaced across the width of the cell (Figure 1). The bottom acrylic sheet is glued to an acrylic frame to form the base, and the top acrylic sheet is fastened to the base with a set of screws. A rubber gasket is placed between the top sheet and the frame to prevent fluid leaking. The flow cell was fully packed with sand (10 mm thick) in subsequent experiments, except as noted otherwise. Sand and Microspheres. Translucent pure quartz sand (Accusand, 40-60 mesh, Le Sueur, MN) was used in this study. The sand was thoroughly cleaned according to the procedures of Litton and Olson (11). Briefly, the sand was soaked in concentrated hydrochloric acid (Trace Metal Grade, Fisher Scientific, Pittsburgh, PA) for 24 h, rinsed thoroughly with ultrapure Milli-Q water (Millipore Co., Billerica, MA),
dried in an oven at 100 °C overnight, roasted in a furnace at 810 °C for 8 h, then cooled, and stored in precleaned glass jars under argon atmosphere. Carboxylate-modified latex fluorescent microspheres (FluoSpheres, Molecular Probes, Eugene, OR) with 1.0 µm diameter and red fluorescence (excitation/emission wavelengths of 580/605 nm) were used in this study. The microsphere stock solution (3.0 × 1010 spheres/mL) was sonicated for 30 s and was diluted to desired concentrations. A few excitation/emission filter combinations were tried, and the pair with excitation/emission wavelengths of 535/600 nm gave the best result and therefore was selected for observing the red fluorescent microspheres. Calibrations. In order to convert fluorescence intensity to the actual number of microspheres in saturated quartz sand, two types of calibrations were carried out: (1) intensity vs concentration for a given camera exposure time and (2) intensity vs camera exposure time for a given concentration. The second calibration was intended to increase the dynamic range of the imaging system (limited to 214 or 16,384 intensity levels for a given exposure time) by reducing the exposure time for high concentration samples, if a linear relationship exists between intensity and exposure time. A third test, fluorescence intensity vs the thickness of sand (saturated with a given concentration of microspheres), was performed to confirm that the fluorescence signals detected were integrated signals across the whole depth (10 mm) of the flow cell, not just from the bottom of the flow cell. A fourth test, intensity vs irradiation duration for a given concentration and camera exposure time, was also carried out to examine whether fluorescence from microspheres would decay under constant irradiation (e.g., a few hours during transport experiments). Such decay, if occurs, will lead to artifacts in concentration measurements and poor mass balances, limiting the application of this imaging technique to colloid transport studies. To obtain the first calibration, microsphere solution (5.0 × 105, 1.0 × 106, 1.0 × 107, 2.0 × 107, or 5.0 × 108 spheres/mL) and clean quartz sand were gradually added to the flow cell (all inlets and outlets sealed) and gently mixed using a glass rod until the flow cell was full. The flow cell was then capped and imaged with a 30-s camera exposure time (2 s for the 5 × 108 spheres/mL concentration). The volume (67 mL) of the microsphere solution and the mass (328 g) of sand added to the flow cell were predetermined to ensure full saturation and consistency between different experiments. The flow cell was cleaned and repacked for each microsphere concentration to avoid cross contamination. The packed flow cell with the 107 spheres/mL concentration was also imaged with various camera exposure times (5-60 s) to obtain the second calibration. After that, the flow cell was irradiated by VOL. 40, NO. 19, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6065
the filtered light (535 nm) constantly and imaged with a 30-s camera exposure time every 15 min for up to 7 h for the fourth test (intensity vs irradiation duration). The procedures for the third test (intensity vs thickness) were similar to those of the first calibration except that the flow cell was packed with different amounts of microsphere solution and sand to yield layers (quartz sand saturated with 107 spheres/mL microsphere suspension) of different thickness (2.0 mm, 4.0 mm, 7.0 mm, and 10.0 mm). An additional calibration of fluorescence intensity vs aqueous microsphere concentration (no sand) was also carried out for measuring influent and effluent microsphere concentrations in transport experiments. To obtain this calibration, 0.5 mL of microsphere solution (105-107 spheres/ mL) was pipetted into a 24-well culture plate, the plate was imaged with a 30 s exposure time, and the fluorescence intensity in each well was determined. Transport Experiments. A series of microsphere transport experiments was conducted to demonstrate the versatility of the imaging system for studying colloid transport. A syringe pump (Harvard Apparatus, Holliston, MA) with 50-mL gastight glass syringes was used to inject microsphere solution, and a peristaltic pump (Cole Parmer, Vernon Hills, IL) was used to inject microsphere-free solution. A three-way valve was used to connect the two pumps and the flow cell. In this series of experiments, only the center inlet and outlet of the flow cell were used in an attempt to obtain 2D tracer plumes. To better observe the deposition and transport of microspheres around the injection point, we extended the inlet horizontally about 15 mm into the flow cell by inserting a 1/16′′ (inner diameter) Teflon tubing into the original center inlet. The flow cell was then wet packed with the clean quartz sand (porosity ) 0.37) and conditioned with degassed microsphere-free solution for 3 PVs at an average linear velocity of 16.0 m/d. The ionic strength of the conditioning solution was the same as the ionic strength of the microsphere solution (Milli-Q water, 0.1, 1.0, or 10.0 mM NaCl) that would follow. The microsphere concentration was 107 spheres/mL, and the pH was 6.5 for all four ionic strengths. After conditioning the packed flow cell, microsphere solution with a particular ionic strength was injected into the flow cell at a linear velocity of 8.0 m/d for 3 PVs. Microspherefree solution of the same ionic strength was then injected at the same velocity for 3 PVs to elute unattached microspheres. The flow rate was then doubled to examine the impact of increased velocity (increased hydrodynamic shear) on microsphere re-entrainment. After that, Milli-Q water was injected (3 PVs at 8.0 m/d) to examine the re-entrainment of microspheres due to decreased ionic strength. Three more PVs of Milli-Q water were injected at an increased flow rate (16.0 m/d). Finally, the flow direction was reversed (16.0 m/d for 1 PV) to examine the impact of flow direction on microsphere re-entrainment. Effluent samples were collected with a fraction collector for the entire duration of each experiment except for the reversed flow part of the experiment. The flow cell was thoroughly cleaned and repacked for a different ionic strength microsphere solution. The flow cell was imaged at a frequency of 20 pictures per PV for the entire experiment. The camera exposure time was 30 s for most images and was reduced whenever saturation of the CCD-chip was noticed. Effluent microsphere concentrations were determined to yield breakthrough-elution curves. Mass balance of microspheres in each transport experiment at a particular stage was calculated by summarizing the total number of microspheres in the entire flow cell and in the effluent.
Results and Discussion Calibrations. Fluorescence intensity was uniform across the flow cell after illumination reference correction (a flow cell 6066
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 19, 2006
FIGURE 2. Calibration curve of fluorescence intensity vs microsphere concentration within the packed flow cell. The solid line is the linear regression line for the first four data points, whereas the dashed line is the linear regression line for the entire data set.
FIGURE 3. Calibration curve of fluorescence intensity vs camera exposure time at a given microsphere concentration of 107 spheres/ mL within the flow cell. packed with clean sand and saturated with Milli-Q water was used as the reference). Therefore, the mean intensity (per pixel) of the entire flow cell was used to construct the calibration curves. Fluorescence intensities and microsphere concentrations (in saturated sand) showed excellent positive linear relationship (r2 ) 1.00, Figure 2). The detection limit was about 2.5 × 105 spheres/mL in pore water for the 1 µm red fluorescent microspheres. The low detection limit allowed the use of relatively low concentrations (e.g., 107 spheres/ mL) of microspheres in transport experiments for realistic representation of natural colloid concentrations (105-1014 particles/mL with microbial concentrations close to 105 particles/mL (12)) and for reduced cost for microspheres. As expected, fluorescence intensity linearly increased with camera exposure time for a given microsphere concentration (r2 ) 1.00, Figure 3). The second linear relationship allowed us to quantify high microsphere concentrations using reduced exposure time, thereby increasing the dynamic range of the imaging system. For instance, an exposure time of 2 s (as opposed to 30 s) was used for the 5.0 × 108 spheres/mL concentration to avoid saturation of the CCD chip. The fluorescence intensity was then normalized to the 30 s exposure time (i.e., raw intensity multiplied by a factor of 15) and plotted together with the rest data. The extended calibration curve (Figure 2) was still perfectly linear (r 2 ) 1.00), and the slope only changed slightly (
