Environ. Sci. Technol. 2009 43, 8916–8922
Geomembranes Containing Powdered Activated Carbon Have the Potential to Improve Containment of Chlorinated Aromatic Contaminants ERIN M. SURDO,† EDWARD L. CUSSLER,‡ PAIGE J. NOVAK,† AND W I L L I A M A . A R N O L D * ,† Department of Civil Engineering, University of Minnesota, 500 Pillsbury Drive Southeast, Minneapolis, Minnesota 55455, and Department of Chemical Engineering and Materials Science, University of Minnesota, 421 Washington Avenue Southeast, Minneapolis, Minnesota 55455
Received May 12, 2009. Revised manuscript received October 4, 2009. Accepted October 6, 2009.
Breakthrough across high-density polyethylene (HDPE) was measured for 2,3′,4′,5-tetrachlorobiphenyl and a higher-solubility surrogate, 1,2,4-trichlorobenzene. Addition of powdered activated carbon (0.14 g carbon/cm3 membrane) reduced pseudosteady-state flux through thin HDPE membranes by approximately 60%. Breakthrough curves for activated carbon-containing membranes were best described by a model in which sorption to the carbon was limited by the rate of diffusion from the bulk membrane to the carbon particle surfaces. Field-scale estimates based on this model show a substantial (over 10 orders of magnitude) reduction in flux for the activated carboncontaining HDPE compared with pure HDPE. The flux of 2,3′,4′,5tetrachlorobiphenyl through a composite membrane with thin layers of poly(vinyl alcohol) (PVA) with 0.05 g carbon/cm3 and pure HDPE was 69% lower than expected for a similar layered membrane without the sorptive scavenger. This flux reduction was achieved with less than a third of the carbon used in the HDPE case, an improvement that is likely the result of better solute uptake in the hydrophilic PVA layer.
Introduction Exposure to polychlorinated biphenyls (PCBs) via equilibrium partitioning and through the food web leads to bioaccumulation in aquatic organisms (1, 2). Consequently, high PCB concentrations in fish pose a human health threat that has led to fish consumption advisories for many lakes and rivers (3, 4). While surface water PCBs originate from a variety of sources (including atmospheric deposition) (4), contaminated sediment “hot spots” are the target of many remediation efforts. Traditionally, dredging has been used to reduce the exposure of aquatic species to PCB-contaminated sediments by physically removing them from lake- and riverbeds (5). Dredged sediments are commonly stored on nonporous highdensity polyethylene (HDPE) geomembrane liners, either for * Corresponding author phone: (612) 625-8582; fax: (612) 6267750; e-mail:
[email protected]. † Department of Civil Engineering. ‡ Department of Chemical Engineering and Materials Science. 8916
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 23, 2009
permanent landfilling or for temporary dewatering before transport/treatment (6, 7). HDPE geomembranes effectively contain water and heavy metals (8) and are regularly used for contaminant isolation (e.g., landfill liners, silt screens). Hydrophobic molecules (trichloroethylene, benzene, xylene, etc.), however, diffuse more easily through HDPE (9-11). For example, Shimotori et al. (11) estimated that carbon tetrachloride would break through a 1.5-mm thick HDPE liner in 32 days. These observations highlight the potential for other hydrophobic contaminants, such as PCBs, to diffuse through HDPE liners. This study provides the first attempt at measuring PCB diffusion through HDPE. Sediment caps are increasingly preferred over dredging and ex situ treatment, because dredging stirs the contents of the river and its sediments, creating a slug of turbid and polluted water immediately following the dredging process (5, 12). Sand caps physically separate contaminated sediments from the bioactive region and reduce the upward flux of aqueous PCBs from sediment porewater to the overlying surface water (13, 14). Alternative materials, such as crushed limestone, bentonite-cement composites, and a proprietary clay-composite material called AquaBlok offer reduced effective diffusion coefficients compared with sand (7, 15, 16). Effective sorption of PCBs to activated carbon has led to research on in situ sequestration of PCBs in contaminated sediments. One strategy mixes activated carbon into sediments to reduce PCB bioavailability (2, 17, 18). In laboratory tests, PCB bioaccumulation was up to 89% lower in Macoma bathica (clams) exposed to PCB-contaminated sediment treated with activated carbon than in those exposed to the same sediment without activated carbon (2). Less dramatic results were observed at a field site (18). Others have focused on the incorporation of thin layers of activated carbon or coke into sediment caps (14, 19). Equilibrium sorption tests combined with advection/diffusion modeling highlight the potential of “active” materials to extend the PCB isolation time in capped sediments (14). Coke incorporated into a permeable Reactive Core Mat was deployed under a sand cap at a test site in the Anacostia River (19). To further the effectiveness of activated carbon sediment treatment/capping, using Fe/Pd-doped activated carbon to both sorb and dechlorinate PCBs was recently proposed (20). HDPE is used to line hazardous waste landfills because of its durability and resistance to degradation in the environment. Activated carbon has the potential to sorb PCBs as they migrate through an “active” sediment cap. The current study combines these two proven strategies to develop an HDPE barrier with improved barrier performance. Such a barrier might be used to line landfills and confined disposal facilities used to store PCB-contaminated sediments, to cap contaminated sediments in situ in special cases (when groundwater upwelling and gas production within capped sediments are not significant), and to cap terrestrial sources of atmospheric PCBs, such as transformer storage yards (4). Experimentally, the first objective of this study was to characterize diffusion of a model PCB, 2,3′,4′,5-tetrachlorobiphenyl (2,3′,4′,5-PCB), through HDPE, poly(vinyl alcohol) (PVA), and composite membranes via breakthrough experiments. Experiments with 1,2,4-trichlorobenzene (1,2,4-TCB), a higher-solubility PCB surrogate, were used to evaluate the differences between conditions in our experiments and those expected in the environment. The second objective of this study was to determine the potential of powdered activated carbon (PAC) to improve the barrier performance of HDPE. By comparing experimental breakthrough curves to existing model predictions, this work also provides results that further 10.1021/es9014162 CCC: $40.75
2009 American Chemical Society
Published on Web 10/29/2009
our understanding of how sorptive scavengers affect membrane breakthrough for hydrophobic pollutants and identifies the appropriate parameters for scaling results to field conditions.
Theory Fick’s laws describe solute diffusion through membranes. By solving Fick’s second law, the solute concentration on the initially clean (downstream) side of the membrane as a function of time is expressed as (21):
(
∞
Cdown DHA L2 (-1)n -Dn2π2t/L2 2L2 ) te - 2 Cup LVdown 6D π D n)1 n2
∑
)
(1)
where Cdown and Cup (mol/m3) are the downstream and upstream (contaminated) solute concentrations, respectively, D (m2/s) is the solute diffusion coefficient in the membrane, H (dimensionless; (mol/Lmembrane)/(mol/Lwater)) is the membrane-solution partition coefficient of the solute, A (m2) is the area of the membrane normal to the direction of diffusion, L (m) is the thickness of the membrane, Vdown (m3) is the downstream volume, and t (s) is time. The permeability of the membrane, P (m2/s), is equivalent to DH. Equation 1 is often simplified to a linear form: Cdown DHA ) (t - tlag) Cup LVdown
(2)
Equation 2 is valid after the lag time, tlag (s), during which the downstream concentration remains zero while the solute diffuses across the membrane toward the downstream side. For pure polymer membranes (21): tlag )
L2 6D
DHCup L
(4)
An effective contaminant barrier will have a long lag time, a low flux, or both. Experimentally, a finite downstream volume is used in conjunction with downstream concentration measurements to measure the cumulative solute mass passing through the membrane, CdownVdown. Equations 1, 2, and 4 were derived assuming Cup . Cdown, which is expected in the field and is valid in the lab during periods of low Cdown. With a finite downstream volume, however, flux eventually slows as Cdown approaches Cup. Several models describe how tlag and j are affected by scavenger particles in membranes. Yang et al. (22) described the effect of adding scavenger particles that react rapidly with diffusing solutes. Solute molecules that enter the upstream surface are immediately consumed by reaction with membrane-embedded particles. The reactive particles are also consumed, allowing new solute molecules to diffuse deeper and advancing the “reactive front”. Breakthrough only occurs when the reactive particles in the membrane have been exhausted, extending tlag. This model effectively described breakthrough data for membranes containing Fe0 nanoparticles (11, 23). Warta et al. (24) extended the model for membranes with particles that sorb solutes according to the Langmuir isotherm: q)
qmaxbCsolution 1 + bCsolution
tlag )
L2Cs0qmax 2DH(Cup - Ce)
(6)
where Cs0 (kg/m3) is the initial scavenger concentration in the membrane and Ce ) 1/b (mol/m3), the minimum dissolved solute concentration at which the sorption capacity of the particles is reached (24). A second model predicts breakthrough for dissolved concentrations below Ce, where equilibrium sorption to particles is instantaneous and a linear function of C. Diffusion through the membrane is described by a mass balance (25): ∂q ∂2C ∂C + Cs0 )D 2 ∂t ∂t ∂x
(7)
With K ) qCs0/C ) qmaxbCs0 (dimensionless), eq 7 becomes: D ∂2C ∂C ) ∂t 1 + K ∂x2
(8)
Equation 8 mimics Fick’s second law but with a reduced effective diffusion coefficient. With Deffective ) D/(1 + K), eq 1 predicts the initial curvature of breakthrough curves under the fast equilibrium assumption (25). A third model for scavenger-containing membranes argues that if the concentration of particles within the membrane is low, diffusion from the bulk membrane to the particle surface limits the kinetics of sorption. The reactiondiffusion equation for a membrane with mass transfer-limited sorption kinetics is (25):
(3)
After the lag, a constant flux, j (mol/m2s), develops (21): j)
for solute sorption, Csolution (mol/m3) is the equilibrium concentration of solute in solution, and b (m3/mol) is a constant. When sorption saturates the particles, tlag is described by:
(5)
where q (mol/kg) is the concentration of solute sorbed to particles, qmax (mol/kg) is the total capacity of the particles
∂2C ∂C ) D 2 - kC ∂t ∂x
(9)
The diffusion-limited sorption rate constant, k (s-1), is derived using the steady-state Fick’s law expression for uptake by spherical particles (26): d 2 dC r )0 dr dr
( )
(10)
where r (m) is the direction of diffusion toward the surface of a spherical particle. Equation 10 is solved using boundary conditions C(R) ) 0 and C(S) ) Cm(x,t), where R (m) is the radius of the particle, S (m) is the radius of diffusion around the particle, and Cm(x,t) (mol/m3) is the concentration of solute in the membrane at a given depth x (m) and time t. The steady-state solution to eq 10 gives k as a function of R and the volume fraction of particles, φ (dimensionless): 3D
k) 2
1
(11)
( φ1 - 1)
R (1 - φ /3)
Equation 11 is similar to that derived previously for particles in a cubic unit cell (27). A complete derivation of k is provided in the Supporting Information (SI) (Section S1). Equation 9 is solved analytically for the steady-state case to calculate the pseudo-steady-state flux of solute through the membrane. The term pseudo-steady-state indicates that the steady-state solution of the diffusion-limited kinetics model is an intermediate value --the derivation assumes an infinite capacity of the sorbent particles for the diffusing solute. In reality, the particles slow diffusion by intercepting and immobilizing solute molecules within the membrane. VOL. 43, NO. 23, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8917
TABLE 1. Breakthrough Curve Models: Assumptions and Relationships
model 0: pure polymer (21)
model 1: fast, irreversible sorption (moving reactive front model) (22)
model 2: fast equilibrium sorption (25)
tlag
pseudo-steady-state j
L2 6D
DHCup L
L2Cs0qmax 2DH(Cup - Ce)
DHCup L
L2(1 + K) 6D
DHCup L
HCup√kD model 3: diffusion-limited sorption kinetics (25)
When the particles are saturated, flux through the membrane increases to the pure polymer value. Table 1 summarizes lag times and pseudo-steady-state flux expressions for the models presented.
Experimental Section The materials used are described in the SI (Section S2). PAC particle size was measured with a Multisizer 3 (Beckman Coulter). Membrane Preparation. Appropriate ratios of HDPE beads and PAC were mixed at 160 °C in a Haake PolyLab OS RheoDrive 4 batch mixer (Thermo) for ∼20 min. The resulting mixtures were pressed to films 100-400 µm thick between two polytetrafluoroethylene (PTFE) sheets using a hot press (Wabash Metal Products) at 160 °C with 3300-6700 N force. Membrane thicknesses varied as an artifact of variations in the membrane preparation process and were sometimes varied intentionally to test reproducibility. Thicknesses were measured with a micrometer (Mitutoyo). PVA membranes were prepared by adding 1.5 g of PVA and the desired mass of PAC to 15 mL of water on a hot plate (
