Visualization and Modeling of Polystyrol Colloid Transport in a Silicon Micromodel

ventional methods to investigate colloid transport often involve column studies. Here, colloid concentrations are A new experimental approach and complementary model analysis measured at the column effluent or at selected points are presented for studying colloid transport and fate in porous media. The experimental approach combines high precision etching to create along the column length. Unfortunately, such methods do a controlled pore network in a silicon wafer (i.e., micromodel), with not clearly distinguish how spatial and temporal changes epifluorescent microscopy. Two different sizes of latex colloids were in hydrochemical and hydrodynamic conditions affect colused; each was stained with a fluorescent dye. During an experiment, loid transport. For example, breakthrough curves (BTCs) water with colloids was purged through a micromodel at different obtained from column effluent represent some average flow rates. Flow paths and particle velocities were determined and behavior of colloids in the column (Baumann et al., compared with flow paths calculated using a two-dimensional (2D) 2002). Since different heterogeneous realizations can lattice Boltzmann (LB) model. For 50% of the colloids evaluated, contribute to such BTCs, the processes that control colagreement between measured and calculated flow paths and velocities loid transport in the column are obscured. were excellent. For 20%, flow paths agreed, but calculated velocities Filtration theory (Happel, 1958; Rajagopalan and were less. This is attributed to the parabolic velocity profile across the micromodel depth, which was not accounted for in the 2D flow Tien, 1976) is often used to evaluate colloid transport. model. For 12%, flow paths also agreed, but calculated velocities were It accounts for the hydrodynamic processes that lead less. These colloids were close to grain surfaces, where model errors to a contact between particles and filter surfaces. The increase. Also, particle–surface interactions were not accounted for attachment efficiency, , describes the probability that in the model; this may have contributed to the discrepancy. For the a collision between a particle and a filter grain results remaining 18% of colloids evaluated, neither flow paths nor velocities in a permanent attachment (Elimelech and O’Melia, agreed. The majority of colloids in this last case were observed after 1990). This parameter lumps the physical and chemical breakthrough, when concentrations were high. The discrepancies may interactions between colloids and surface at the pore be due to particle–particle interactions that were not accounted for scale. Often the attachment efficiency is used as a fitting in the model. Filtration efficiencies for all colloid sizes at different parameter for the inverse modeling of colloid breakflow rates were calculated from filtration theory. Attachment rates were obtained from successive images during an experiment. With through curves (Ren et al., 2000; Huber et al., 2000). these, attachment efficiencies were calculated, and these agreed with The boundary conditions, especially the pore topolliterature values. The study demonstrates that excellent agreement ogy, the local flow velocities, and the chemical heterogebetween experimental and model results for colloid transport at the neity of the surface, can vary in the pore space. In colpore scale can be obtained. The results also demonstrate that when umn tests, the packing density and the topology of the experimental and model results do not agree, mechanistic inferences pore network are unknown, and local physical and and system limitations can be evaluated. chemical heterogeneities cannot be assessed (Sugita and Gillham, 1995). Also, preferential flow inside the column is obscured. Therefore, the inverse derivation of C are ubiquitous in many groundwater aquifiltration parameters e.g., the attachment efficiency) are fers. They originate from weathering processes of representative only for the length scale of the column. the aquifer matrix, degradation of biological material, As a result, upscaling and downscaling become difficult, and precipitation of supersaturated solutions (Buffle et and experimentally derived parameters do not always al., 1998). Under certain conditions, colloids may facilitate conform to theory. A promising approach to study porethe transport of hazardous substances such as radionuscale processes involves the use of micromodels. clides (Tanaka and Nagasaki, 1997; Kersting et al., 1999), In groundwater and vadose zone studies, micromodheavy metals (Karathanasis, 1999; Kretzschmar et al., els are representations of porous media etched into sili1999), and organic substances (Roy and Dzombak, 1998; con wafers, glass, or polymers (Soll et al., 1993). SomeVillholth, 1999). It is therefore of great interest to predict times thin layers of glass beads or sand embedded colloid transport. between glass plates are also referred to as micromodels. Colloid transport is very sensitive to hydrochemical The main purposes of micromodel experiments are to and hydrodynamic conditions (Roy and Dzombak, 1997; increase spatial and temporal resolution and to provide Bergendahl and Grasso, 2000; Bradford et al., 2002). Condirect quantitative access to processes at the pore scale. An overview of micromodel applications is docuT. Baumann, Institute of Hydrochemistry, Technische Universität mented in Table 1. Each table entry notes the experiMünchen, Marchioninistr. 17, D–81377 Munich, Germany; C.J. Werth, mental approach and objectives, associated references, Dep. of Civil and Environmental Engineering, University of Illinois, 205 N. Mathews Ave. MC-250, Urbana, IL 61801. Received 10 July and the type of porous media represented. Only two of 2003. Special Section: Colloids and Colloid-Facilitated Transport of the references involved study of colloid transport (Wan Contaminants in Soils. *Corresponding author (thomas.baumann@ and Wilson, 1994b, 1996). In one study, the role of the ch.tum.de). Abbreviations: BTC, breakthrough curve; DDI, deionized distilled; Published in Vadose Zone Journal 3:434–443 (2004).  Soil Science Society of America LB, lattice Boltzmann; PTA, particle tracking algorithm; UV, ultraviolet; 2D, two-dimensional; 3D, three-dimensional. 677 S. Segoe Rd., Madison, WI 53711 USA