Вы находитесь на странице: 1из 9

Article pubs.acs.

org/est

A Field-Validated Model for In Situ Transport of Polymer-Stabilized nZVI and Implications for Subsurface Injection
Magdalena M. Krol, Andrew J. Oleniuk, Chris M. Kocur, Brent E. Sleep, Peter Bennett, Zhong Xiong, and Denis M. OCarroll,*

Civil and Environmental Engineering, The University of Western Ontario, London, Ontario, Canada Golder Associates Ltd., 33 Alderney Drive, Suite 460, Dartmouth, Nova Scotia, Canada Civil Engineering, University of Toronto, Toronto, Ontario, Canada Haley & Aldrich, 1956 Webster Street, Suite 450, Oakland, California 94612, United States AMEC Environment & Infrastructure, 121 Innovation Drive, Suite 200, Irvine, California 92617, United States
S Supporting Information *

ABSTRACT: Nanoscale zerovalent iron (nZVI) particles have signicant potential to remediate contaminated source zones. However, the transport of these particles through porous media is not well understood, especially at the eld scale. This paper describes the simulation of a eld injection of carboxylmethyl cellulose (CMC) stabilized nZVI using a 3D compositional simulator, modied to include colloidal ltration theory (CFT). The model includes composition dependent viscosity and spatially and temporally variable velocity, appropriate for the simulation of pushpull tests (PPTs) with CMC stabilized nZVI. Using only attachment eciency as a tting parameter, model results were in good agreement with eld observations when spatially variable viscosity eects on collision eciency were included in the transport modeling. This implies that CFT-modied transport equations can be used to simulate stabilized nZVI eld transport. Model results show that an increase in solution viscosity, resulting from injection of CMC stabilized nZVI suspension, aects nZVI mobility by decreasing attachment as well as changing the hydraulics of the system. This eect is especially noticeable with intermittent pumping during PPTs. Results from this study suggest that careful consideration of nZVI suspension formulation is important for optimal delivery of nZVI which can be facilitated with the use of a compositional simulator.

INTRODUCTION Groundwater pollution by hazardous industrial chemicals, such as chlorinated solvents, is a serious problem worldwide. Although considerable advances in the understanding of the phenomena governing groundwater remediation have been made, most solutions are still not tailored to source zone removal. Consequently, there is a need for development and pilot scale testing of new and innovative remediation technologies capable of eectively treating source zone contaminants. Nanotechnology is an emerging industry with promise for application to groundwater remediation. Of particular interest to the remediation community is nanoscale zerovalent iron (nZVI) that is capable of reducing chlorinated contaminants, polychlorinated biphenyls (PCBs), and immobilizing metals in the subsurface (e.g., refs 111). Similar to microscale ZVI, originally used in permeable reactive barriers (PRB), nZVI acts as a reductant converting chlorinated contaminants to nontoxic compounds.1,2,6 However, unlike PRBs, where groundwater contaminants react with the iron as they ow through the barriers,1214 nZVI can be directly injected into source zones reducing contaminant mass or be transported with the groundwater to reach contaminated zones. Additionally, since nZVI has a higher specic surface area than microscale ZVI, it
2013 American Chemical Society

produces higher reaction rates6,15 and has been recently explored by several groups as a novel remediation option (e.g., refs 6,11,1621) Although numerous laboratory studies have shown the eectiveness of nZVI to degrade chlorinated compounds (e.g., refs 1,3,5, and 6), most lab and eld applications have suered from poor nZVI mobility.19,2225 This is due to nZVI particles aggregating quickly, resulting in most of the mass being deposited near the injection location. Since the eciency of nZVI is directly dependent on the ability to get it to contaminated areas, aggregation can pose a signicant obstacle to remediation applications. To overcome this problem, nZVI particles can be coated with polymers making them more stable in suspension.19,21,2630 Polymer coatings provide electro-static and steric repulsion that counteract the magnetic and van der Waals forces between nanoparticles resulting in well dispersed suspensions.28,3134 Carboxylmethyl cellulose (CMC) is a common stabilizer at eld installations. Laboratory studies have shown that CMC-coated nZVI particles are much more
Received: Revised: Accepted: Published:
7332

October 11, 2012 February 26, 2013 May 31, 2013 May 31, 2013
dx.doi.org/10.1021/es3041412 | Environ. Sci. Technol. 2013, 47, 73327340

Environmental Science & Technology mobile in column experiments than bare nZVI.21,27,35 However, since laboratory tests do not capture the complexities and heterogeneities present at real contaminated sites, eld scale testing is required to realistically assess the eectiveness of these coated particles. Numerous nZVI eld trials have been performed with varying success (e.g., refs 16,34,3640). Early studies using unstabilized nZVI resulted in poor mobility and well clogging due to particle aggregation and settling.16,23,37 More recently, Bennett et al.41 performed a series of pushpull tests (PPTs) to assess CMC coated nZVI mobility and reactivity, whereas He et al.26 completed a pilot study consisting of gravity-fed and pressurized injection of nZVI into the subsurface. Both sites were contaminated with chlorinated compounds and rapid dechlorination was observed upon nZVI injection. However, nZVI mobility varies depending on soil characteristics, test operations, and injection velocities, which makes nZVI transport in future eld studies dicult to predict. The ability to predict the eectiveness of a remediation strategy is a prerequisite for eective remediation design. Models can be valuable tools for site remediation practitioners in determining appropriate remedial alternatives, helping to minimize time needed for treatability studies and allowing for assessment of various remediation alternatives. To date, modeling of nZVI movement through porous media has been limited to simulation of lab scale experiments, using colloidal ltration theory (CFT).26,4345 CFT describes the attachment of colloid-sized particles due to Brownian diusion, interception, and gravitational sedimentation43,46 and has been used to quantify column transport parameters (e.g., attachment eciency and travel distances).18,26,44,45,47 Currently, no published modeling study has involved simulation of eld-scale nZVI transport. In this study a threedimensional numerical model, modied to incorporate nZVI transport described by CFT, was used to predict the movement of CMC coated nZVI particles at the eld scale. In addition, since CMC solutions have viscosities higher than water, the model was used to assess the importance of including variable viscosity when simulating nZVI eld injections. The eld test of Bennett et al.41 was used to validate the model by simulating tracer and nZVI transport and recovery. This study represents the rst validation of CFT for modeling eld scale transport of CMC stabilized nZVI (CMC-nZVI). Additionally, the validated model was used to assess nZVI transport and aquifer hydraulics due to the injection of a viscous nZVI suspension under a range of eld relevant conditions.

Article

Figure 1. Site Layout.

m below ground surface. Groundwater pore velocity ranged from 0.3 m/day to 1.5 m/day.41 Three PPTs were performed at dierent depths, corresponding to dierent water-bearing zones (PPT-1, PPT-2, and PPT3), as shown in Figure 1. A multilevel injection well was installed for injection and withdrawal of uid during the tests. For each PPT, CMC-nZVI particles were prepared on site and injected into the corresponding aquifer, along with a conservative tracer (NaBr).41 Injection and extraction followed a dierent pumping schedule for each layer (Table 1). Both Table 1. Push-Pull Test Injection/Extraction Schedulea41
PPT-1 inject for 0.6 h at 3.4 L/min no injection/extraction for 13.0 h extract for 3.4 h at 3.4 L/min inject for 1.6 h at 1.2 L/min extract for 1.9 h at 0.7 L/min no injection/extraction for 13.4 h (pump failed) extract for 8.1 h at 0.7 L/min inject for 2.1 h at 2.6 L/min extract for 6.0 h at 2.1 L/min

PPT-2

PPT-3
a

Injection and extraction rates are averages as reported by Bennett et al.41. Each well was fully screened and packed o over each water bearing layer.41

MATERIALS AND METHODS Field Test. A eld scale test of CMC-nZVI injection was completed in the spring of 2006 by Bennett et al.41 Site and test details are available in Bennett et al.41 while a brief overview follows. The site was located near San Francisco Bay, CA and consisted of a layered alluvial deposit. Silts and clays layered with coarse-grained sediments made up the sites geology, with the coarse-grained sediments representing the primary waterbearing zones. The eld test was performed in an area overlying a dissolved TCE plume. In this area, three coarse sand zones were bounded by less permeable, clayey soils (Figure 1). The rst and third layers were characterized as poorly graded sand while the middle layer was classied as a silty sand (see Supporting Information (SI) for MIP log results).48 The rst coarse-grained layer was conned and located approximately 1
7333

PPT-1 and PPT-2 had periods of inactivity (lag period) where no injection or extraction was performed, while PPT-3 was pumped continuously. Extracted volumes used to compare simulated and eld results were calculated using appropriate extraction test rates and the elapsed time between sampling events. Numerical Approach. The eld test was simulated using CompSim, a three-dimensional, three-phase, nite dierence model capable of predicting subsurface contaminant migration.49 This model has been used in a variety of laboratory and eld scale applications including pump and treat remediation, hot water ooding, steam ooding, and bioremediation.5053 To simulate nanoparticle transport, equations based on CFT were incorporated into CompSim using an attachment rate coecient (Katt) which serves as a rst order removal term in the transport equation:
dx.doi.org/10.1021/es3041412 | Environ. Sci. Technol. 2013, 47, 73327340

Environmental Science & Technology


(nC) + nK attC = (nDC) (Cq ) t

Article

Table 2. Simulation Parametersa


(1)
1

site property hydraulic conductivity (K) porosity (n) dispersivity (c) particle diameter (dp)d collector diameter (dc)e regional groundwater velocity (vp) PPT property

units m/d m nm mm m/d

sand 30 0.36 0.07 140 0.2

silty sand 5.4b 0.36 0.07 140 0.1 0.31.5 PPT-1 8.1 16.15

clay 2 104 0.36 0.07 140 0.01

where C is the aqueous phase nZVI concentration (molesL ), D is the hydrodynamic dispersion tensor (m2s1), q is the Darcy velocity (ms1), and n is porosity (). The attachment rate coecient (s1) represents the process of colloid ltration, and is dened as
K att = 3(1 n) 0vp 2dc
(2)

units cP

PPT-2 PPT-3 2.8 8.04 2.1 5.42

where dc is the collector grain size (diameter) (m), vp is the pore water velocity (p = q/n) (ms1), is the attachment or sticking eciency () dened as the ratio of immobilized nanoparticles on the collector per nanoparticle collision, and 0 is the theoretical collision eciency or the single collector contact eciency () dened as the ratio of particles striking the collector to those approaching the collector. 0 was calculated using dimensionless parameters associated with the eects of diusion, interception, and gravity, using the relationships developed by Tufenkji and Elimelech.43 This equation does not take into account particle detachment or aggregation. Aggregation is an important mechanism for unstabilized particles28 while detachment has been shown to be insignicant in similar studies.26,54 Since the three sand layers were separated by low conductivity clay (Figure 1), they were considered to be individual water bearing zones and were modeled separately with clay bounding each sand layer. For each PPT, the corresponding simulation domain was divided into three vertical soil layers (claysandclay) with a total of 11 vertical blocks (3 for clay, 5 for sand, 3 for clay), and 41 blocks in both lateral and transverse directions with a discretization of 0.09756 m in the x and y direction. Discretization in the z direction varied between 0.144 and 0.448 m, depending on the depth of the layer. The top and bottom of the domain were assumed to be no ow boundaries while hydraulic head values were prescribed (constant head boundaries) at the left and right boundaries of the modeled domain, to produce the appropriate regional groundwater velocities (ranging from 0.3 to 1.5 m/ day).41 The injection/extraction well was modeled as a source/ sink. All sand and clay layers were modeled as isotropic and homogeneous, and dispersivity was used as the tting parameter for the bromide transport. Model and site parameters are shown in Table 2 and are consistent with those quantied in the eld. CMC was modeled as a conservative species as it was added in excess in solution. Therefore, both nZVI and the polymer were modeled as separate constituents and their movement was simultaneously simulated according to the transport equation (eq 1). Several viscosity measurements using CMC (90K) at dierent polymer fractions were performed in the lab (the details of which can be found in the SI). The Grunberg and Nissan equation55 (eq 3) was used to t the lab viscosities by minimizing the root-mean-square error (RMSE) between calculated and measured viscosities, using the polymer viscosity (poly) as the tting parameter (SI Figure S1). The resulting solution viscosities are summarized in Table 2.
ln( ) = xpoly ln(poly ) + x water ln(water ) sol
(3)

polymer solution viscosity (sol) polymer molar fraction (xpoly) ( 107)


a

Unless otherwise indicated, parameters were obtained from Bennett et al.41 bSilty sand layer (PPT-2) value calculated assuming appropriate dc and Kozeny-Carman equation. cFitted parameter. d Elliott and Zhang.16 edc values assumed for sandy layers (PPT-1 and PPT-3) and sandy silt layer (PPT-2).56

During the PPTs, injection of CMC led to variations in subsurface viscosities from that of water (1 cP). Equation 3 was
7334

RESULTS AND DISCUSSION Tracer Test. In the PPTs conducted by Bennett et al.41 the nZVI and tracer (bromide) concentrations were measured at the injection/extraction well. The simulated tracer results as well as the tracer mass recovered at the PPT well were compared to the eld results (Figure 2 and Table 3). All the simulations were run with two regional groundwater velocities (simulated by imposing a regional hydraulic head gradient with constant head boundary conditions) representing the estimated upper and lower bound of the regional groundwater ow (0.3 and 1.5 m/d). The recovered mass calculated by the model at the two regional groundwater ows (Table 3) bracketed the measured mass recovery for the rst two tests. In PPT-1 and PPT-2, the tracer moved with the regional velocity during the lag periods. Thus, during these periods there was continuing tracer transport, mixing, and dilution from regional groundwater ow, resulting in lowering of bromide concentrations at the PPT well following the onset (PPT-1) or resumption (PPT-2) of extraction. The higher velocity (1.5 m/day) produced a greater reduction in bromide concentration, as expected. The simulated tracer extraction curves (Figure 2(a), Figure 2(c), and Figure 2(e)) for the two regional groundwater velocities bracket the measured extraction curves. The higher velocity results t PPT-1 data better while the PPT-2 data was better matched by the lower regional velocity through the middle layer. This is consistent with the soil classication of the layers, where PPT-1 and PPT-3 layers were characterized as poorly graded sand, and PPT-2 as a silty sand and therefore with a lower hydraulic conductivity.41,48 The PPT-3 simulations were the least aected by the regional groundwater velocity due to the lack of a lag period between injection and extraction (PPT-3 was continuously pumped). Close agreement between CompSim simulations and observed behavior suggests that conceptual model assumptions are appropriate: the aquifer soil can be modeled as homogeneous and isotropic and all sand layers were distinct water bearing zones with no ow between layers. nZVI Transport. Similarly to the tracer simulations, nZVI transport was simulated with two dierent regional groundwater velocities and was used as the tting parameter. The best t was determined by minimizing the RMSE between
dx.doi.org/10.1021/es3041412 | Environ. Sci. Technol. 2013, 47, 73327340

used to calculate this spatially and temporally changing viscosity.

Environmental Science & Technology

Article

Figure 2. Field and simulated results for PPT-1 (a and b); PPT-2 (c and d); and PPT-3 (e and f); tted values are given in Table 3.

Table 3. Observed and Simulated Tracer and nZVI Recovery, Simulated with Dierent Regional Groundwater Velocities
PPT-1 bromide recovery observed eld Br recovery (%) simulated recovery (vp = 0.3 m/d) simulated recovery (vp = 1.5 m/d) PPT-1 iron recovery observed eld Fe recovery (%) simulated recovery (vp = 0.3 m/d) simulated recovery (vp = 1.5 m/d) % 2.6 1.1 2 0.02 0.01
7335

PPT-2 % 73 76 42 PPT-2 % 21 18 12 0.006 0.006 % 31 37 36

PPT-3 % 76 72 63 PPT-3 0.007 0.006

% 61 99 50

dx.doi.org/10.1021/es3041412 | Environ. Sci. Technol. 2013, 47, 73327340

Environmental Science & Technology

Article

Figure 3. Eect of attachment eciency () on nZVI recovery.

observed and simulated nZVI concentrations. The nZVI aqueous concentrations for the three tests were reported with respect to extracted volumes, and are presented in Figure 2(b), (d), and (f), respectively, while Table 3 presents the recovered mass and values for all three tests (also presented in SI Figure S2). The nZVI colloid concentration, which is the combination of an nZVI core and an oxide shell, was assumed to be equal to the total iron concentration measured in the eld. The calibrated values were assumed to be temporally and spatially constant for simulation of eld tests that had changing injection/extraction rates as well as lag periods. In contrast, 0 was a function of the spatially variable viscosities and velocities, as described by the Tufenkji and Elimelech43 model. The lowest nZVI recovery (observed and modeled) corresponded to PPT-1 (Table 3) where a signicant lag of 13 h occurred between injection and extraction (Table 1). During the lag period, nZVI particles moved away from the injection/extraction well with regional groundwater ow and attached onto the soil grains, leading to low recovery once extraction started (SI Figure S3). PPT-1 also had the lowest tracer recovery (Table 3) but the tracer recovery for all three tests was still on the same order of magnitude, unlike nZVI recovery. This implies that under these conditions, with velocities equal to the regional groundwater velocity, it was not just nZVI movement away from the well that led to the low nZVI recovery but also signicant nZVI attachment onto the soil grains during the lag period (SI Figure S3). This was conrmed by calculating the amount of nZVI attached during dierent phases of the test. According to the model, only 9% of the injected nZVI was attached onto soil grains after injection but this amount rose to 90% after the lag period, leaving only 10% of nZVI in suspension for extraction. This led to low nZVI recovery for PPT-1. For PPT-2, 39% of the total mass was attached after injection and 70% after the rst extraction (where 15% of mass was extracted), leaving only 15% in suspension. During the lag phase, virtually all of the remaining nZVI was attached leaving no nZVI in suspension when the second extraction occurred. This is seen in Figure 2(d) where nZVI recovery drops to zero after the lag period. PPT-3 had 21% of mass attached after injection, leaving 79% in suspension for extraction, resulting in much higher nZVI recovery for PPT-3 as compared to both PPT-1 and 2. In all cases, very little mass (<1%) was left in suspension after the nal extraction period (SI Figure S4). This is also shown in SI Figure S3 which depicts
7336

the attached nZVI mass in the middle of the PPT layers following injection and extraction phases. These simulations show that the nZVI mobility is highly dependent on approach velocity (vp). In the case of PPT-3 and PPT-2 where extraction occurred right after injection, the higher injection rate (PPT-3), and therefore higher vp resulted in higher recovery, as observed by others.26,45,5759 When a lag between injection and extraction occurred and vp was equal to the regional groundwater velocity, the resulting nZVI attachment was signicantly higher as predicted by CFT, leaving little mass in suspension. This indicates that constant injection of nZVI may be better suited for nZVI applications where nZVI needs to travel far enough to reach the contaminant, whereas injections with lag periods would be benecial for remediation scenarios where nZVI emplacement into a source zone is the goal. The tted values for PPT-1 were higher than the other two tests which had lower injection solution viscosities (Table 2 and Table 3). The sensitivity to values on the resulting nZVI recovery is shown in Figure 3 (varied 50% for vp = 1.5 m/ day). For PPT-1, an increase in resulted in a decrease in initial nZVI concentration (Figure 3(a)). The eect of on PPT-2 was only apparent during the initial part of the test, before the lag period of 13.4 h, whereas in PPT-3, the initial nZVI concentration was aected (akin to PPT-1) (not shown). Values for for various lab scale transport studies of polymer coated nZVI are given in the SI (Table S1). The only direct comparison of attachment eciencies can be made with He et al.26 and Raychoudhury et al.60 who used the same nZVI synthesis methods and the same coating (CMC 90K). The calibrated range of in this study was larger than that reported by He et al.26 However, He et al.26 used a hydrodynamic particle diameter (dp) of 18 nm when calculating . Higher particle diameters have been reported for uncoated nZVI particles6,11,16 and given that the particles in this study were CMC modied and some amount of aggregation would be expected in the eld, a higher dp was assumed.21 Using a larger dp leads to smaller 0 values43 resulting in larger tted . Raychoudhury et al.60 reported higher values than those reported in this study. However,60 did not account for the solution viscosity and used a higher pore water velocity than the groundwater velocity observed in the eld. Both these factors would change the obtained.
dx.doi.org/10.1021/es3041412 | Environ. Sci. Technol. 2013, 47, 73327340

Environmental Science & Technology

Article

Figure 4. (a) Decrease in injection velocity during CH injections (b) Resulting hydraulic heads at the injection well during CF injection at dierent times with injection viscosity = 13 cP (x = 2 represents location of injection/extraction well).

Field Implications. Typical CMC solutions consist of a polymer that has a molecular weight of 90 000 g/mol (CMC 90K) or 250 000 g/mol (CMC 250K). Due to the high molecular weight, the viscosity of the nZVI/polymer solution will be greater than that of water and therefore as the suspension is injected, the uid viscosity surrounding the well increases. As greater amounts of nZVI are injected, the zone of higher viscosity increases radially outward. The higher viscosity injection solution can aect both nZVI attachment as well as the subsurface ow eld. An increase in uid viscosity will decrease 0 leading to a decrease in attachment rate (SI Figure S5) and consequently, to an increase in nZVI aqueous concentration. This change in 0 is due to decreases in Brownian diusion with increased viscosity.43 The dependency of 0 on viscosity and approach velocity will result in a non uniform 0 distribution, and thereby attachment that varies radially out from the well. The eect of viscosity on 0 and corresponding nZVI concentration can be observed in Figure 2(f) where the nZVI concentration is lower than the observed concentration when simulated with a uniform viscosity and the same attachment coecient. This change can be observed even when the injection solution viscosity of PPT-3 was 2 cP. The tracer results (Figure 2(e)) show that for this PPT, the ow eld is not aected by the injected solution viscosity due to the constant injection and extraction at the well (no lag period). Therefore, if uniform viscosity is used to model this scenario, a reasonable t to the observed data could be obtained by using as the tting parameter (Figure 2(f)). However the value tted with uniform viscosity would be lower than that with non uniform viscosity. The injection viscosity can also aect the hydraulics of the system. The eect will depend on the type of injection performed as well as the pumping schedule of the test. For PPT-1, where a signicant lag in pumping occurred (Table 1), the tracer concentration could not be correctly predicted if uniform viscosity was assumed (Figure 2(a)). This is due to the changed ow eld after injection of viscous uid into the subsurface. Once pumping stops, the viscous plume is bypassed by the regional groundwater, reducing nZVI and tracer transport and allowing for more nZVI and tracer to be extracted once pumping restarts. If uniform viscosity is assumed, the tracer and nZVI would travel further from the well resulting in lower extracted concentrations as seen in Figure 2(a). In this case, neither tracer nor nZVI transport could be accurately predicted using a uniform viscosity, even
7337

with dierent tted values for nZVI (Figure 2(b)). It is anticipated that predictive capabilities of numerical simulators would become poorer as suspension viscosity increased, if uniform viscosity was assumed. The eect of the injection solution viscosity will also dier depending on the type of eld injection performed. Injection of uids into the subsurface whether for pump tests or remediation purposes are typically performed in two ways: constant head (CH) or constant ux (CF) injections.61,62 CF injections use a constant volumetric ow rate to inject water (or remediant) into the subsurface, while CH injections specify a constant head or pressure at the well and the uid is applied to the subsurface via gravity feed. CF injections were used by Bennett et al.41 in this eld trial while others have injected nZVI into the subsurface via CH injections.16,36,37,39,42 Subsurface application of nZVI is governed by Darcys Law:
q= kg dh dr

(4a)
r

hI = hr

q dr kg

(4b)

where q is the Darcy velocity, k is the soil permeability, is the uid density (assumed to be constant), g is the gravity, is the uid viscosity, hI is the hydraulic head at the inlet (well), r is the radial distance from the well, and hr is the hydraulic head at distance r. With a CH injection, the change in head (dh, eq 4a) remains constant with time, however viscosity changes spatially and therefore the Darcy velocity decreases with time. For the CF case, Darcy velocity remains constant since the ux is controlled by the injection rate. However, the increase of viscosity will lead to an increase in inlet head with time (eq 4b). To examine the implication of both injection methods on nZVI transport, the PPT-1 domain was used to simulate the injection of CMC-nZVI suspension using CH and CF methods SI Table S2. For CH, a constant hydraulic head was prescribed at the injection well. The value of the hydraulic head was chosen so that the initial injection rate was equivalent to the PPT-1 injection rate at the beginning of the test (204 L/ min,41). However, with prolonged CMC injection, signicant reduction in Darcy velocity was observed with time. For example, the velocity at the well decreased from 8.8 105 m/s at 10 min to 4.1 105 m/s at 48 h leading to decreasing injection rates (Figure 4(a)). This decrease in velocity was only
dx.doi.org/10.1021/es3041412 | Environ. Sci. Technol. 2013, 47, 73327340

Environmental Science & Technology

Article

Figure 5. nZVI concentration with time during CH (a and c) and CF (b and d) injection at dierent injection viscosities; x = 2 represents location of injection well.

observed in this simulation and not in the eld since CF methods were used in the eld test. These reductions in Darcy velocity will result in increased attachment (Katt) with time and decreased nZVI mobility (eq 2). Therefore, if CH injections are used with CMC-nZVI particles, the decrease in velocity can lead to a decrease in nZVI transport both in terms of nZVI spread and delivery time. This is observed in Figure 5(a) and 5(b) where the nZVI concentration in the middle of the PPT-1 layer is shown with time for the two injection methods. This velocity eect on nZVI movement was examined by26 who found that reductions in pore velocity in a sand column decreased the euent nanoparticle concentration. Consequently, if a CH injection is simulated with a spatially constant viscosity distribution, the change in velocity would not be captured leading to overestimation of nZVI mobility. The constant velocity and lowered nZVI attachment are advantages of using CF to deliver nZVI to the subsurface. Using CF injection, nZVI travels further from the injection well and a greater fraction remains in suspension (Figure 5(b)). For example at 3 h 72% of the nZVI remains in suspension for the CF case whereas 65% remains in suspension for the CH case after 5 h (equivalent injected volume). However CF injections will increase the hydraulic head at the well with time (Figure 4(b). For example, the simulated hydraulic head at the well (hI) rose from 1.3 at 10 min to 1.7 m at 20 h. This type of pressure increase can lead to daylighting and cracking in the well seal.62,63 Either of these outcomes will signicantly decrease the
7338

spread of nZVI and potentially render the injection well unusable for future injections. In addition, it is desirable to inject nZVI quickly into the ground to reduce nZVI oxidation in the synthesis vessel.34 Therefore care needs to be taken so that the injection pressure does not exceed the overburden pressure at the well when injecting nZVI using CF methods. The application of nZVI stabilized in a higher viscosity suspension (e.g., CMC 250K) has been investigated by several authors21,29,45 as the nZVI will be stable for longer periods of time in the synthesis vessel and 0 is further reduced, improving mobility. Given these bene ts, higher viscosity nZVI suspensions are likely to be applied in the eld. However, although increasing uid viscosity reduces 0, for CH injections, nZVI movement can be drastically reduced due to the decreased velocity, diminishing the eect of reductions in 0 as seen in Figure 5(c). Injecting a higher viscosity nZVI solution with CF conditions (i.e., CMC 250K, measured viscosity of 72 cP at an initial polymer fraction of 0.8 wt.% 54), would substantially increase the subsurface spread of nZVI (Figure 5(d)) due to the reductions in 0 (SI Figure S5), however the head at the well would signicantly increase over that of the lower viscosity injection. For the case of PPT-1 simulated with CMC 250K, the head at the well would increase to 7 m at 20 h (compared to 1.7 m with CMC 90K). Such a head increase may not be suitable for shallow injections. Another consideration is the impact of viscosity on nZVI reactivity. Although not investigated in this study, it is possible that the contaminants will diuse more slowly to the nZVI
dx.doi.org/10.1021/es3041412 | Environ. Sci. Technol. 2013, 47, 73327340

Environmental Science & Technology particle surfaces if there is a very viscous polymer coating the surface, decreasing reaction rates.34,64

Article

* Supporting Information

ASSOCIATED CONTENT

The membrane interface probe (MIP) log results, performed by AMEC Geomatrix; Viscosity measurements; Rescaled Figure 2; Mass attached in PPTs with distance and time; ; Previous studies on values; Eect of uid viscosity on 0; and Simulation parameters of Figure 5. This material is available free of charge via the Internet at http://pubs.acs.org/.

AUTHOR INFORMATION

Corresponding Author

*Phone: 519-661-2193; fax: 519-661-3942; e-mail: docarroll@ eng.uwo.ca.


Notes

The authors declare no competing nancial interest.

ACKNOWLEDGMENTS This research was supported by the Natural Sciences and Engineering Research Council (NSERC) of Canada through a Post-Graduate Fellowship (PDF) to M.M.K., as well as NSERC and Ontario Graduate scholarships to A.J.O. Additional support has been provided by NSERC Strategic and Collaborative Research and Development Grants as well as the Ontario Centres of Excellence. REFERENCES
(1) Song, H.; Carraway, E. R. Reduction of chlorinated ethanes by nanosized zero-valent iron: Kinetics, pathways, and effects of reaction conditions. Environ. Sci. Technol. 2005, 39, 62376245. (2) Liu, Y.; Choi, H.; Dionysiou, D.; Lowry, G. V. Trichloroethene hydrodechlorination in water by highly disordered monometallic nanoiron. Chem. Mater. 2005, 17, 53155322. (3) Sarathy, V.; Salter, A. J.; Nurmi, J. T.; Johnson, G. O.; Johnson, R. L.; Tratnyek, P. G. Degradation of 1,2,3-trichloropropane (TCP): Hydrolysis, elimination, and reduction by iron and zinc. Environ. Sci. Technol. 2010, 44, 787793. (4) Boparai, H.; Joseph, M.; O Carroll, D. Kinetics and thermodynamics of cadmium ion removal by adsorption onto nano zerovalent iron particles. Journal of Hazardous Materials 2011, 186, 458465. (5) Lien, H.-L.; Zhang, W.-X. Nanoscale iron particles for complete reduction of chlorinated ethenes. Colloids Surf., A 2001, 191, 97105. (6) Liu, Y.; Majetich, S. A.; Tilton, R. D.; Sholl, D. S.; Lowry, G. V. TCE dechlorination rates, pathways, and efficiency of nanoscale iron particles with different properties. Environ. Sci. Technol. 2005, 39, 13381345. (7) Li, X. Q.; Zhang, W. X. Sequestration of metal cations with zerovalent iron nanoparticlesA study with high resolution X-ray photoelectron spectroscopy (HR-XPS). J. Phys. Chem. C 2007, 111, 69396946. (8) Liu, Z. G.; Zhang, F. S. Nano-zerovalent iron contained porous carbons developed from waste biomass for the adsorption and dechlorination of PCBs. Bioresour. Technol. 2010, 101, 25622564. (9) Ponder, S. M.; Darab, J. G.; Mallouk, T. E. Remediation of Cr(VI) and Pb(II) aqueous solutions using supported, nanoscale zerovalent iron. Environ. Sci. Technol. 2000, 34, 25642569. (10) Hoch, L. B.; Mack, E. J.; Hydutsky, B. W.; Hershman, J. M.; Skluzacek, J. M.; Mallouk, T. E. Carbothermal synthesis of carbonsupported nanoscale zero-valent iron particles for the remediation of hexavalent chromium. Environ. Sci. Technol. 2008, 42, 26002605. (11) Wang, C.-B.; Zhang, W.-X. Synthesizing nanoscale iron particles for rapid and complete dechlorination of TCE and PCBs. Environ. Sci. Technol. 1997, 31, 21542156.
7339

(12) Jin suk, O.; Jeen, S.-W.; Gillham, R. W.; Gui, L. Effects of initial iron corrosion rate on long-term performance of iron permeable reactive barriers: Column experiments and numerical simulation. J. Contam. Hydrol. 2009, 103, 145156. (13) Gillham, R.; Ohannesin, S. Enhanced degradation of halogenated aliphatics by zero valent iron. Ground Water 1994, 32, 958967. (14) Higgins, M. R.; Olson, T. M. Life-cycle case study comparison of permeable reactive barrier versus pump-and-treat remediation. Environ. Sci. Technol. 2009, 43, 94329438. (15) Nurmi, J. T.; Tratnyek, P. G.; Sarathy, V.; Baer, D. R.; Amonette, J. E.; Pecher, K.; Wang, C. M.; Linehan, J. C.; Matson, D. W.; Penn, R. L.; Driessen, M. D. Characterization and properties of metallic iron nanoparticles: Spectroscopy, electrochemistry, and kinetics. Environ. Sci. Technol. 2005, 39, 12211230. (16) Elliott, D.; Zhang, W.-X. Field assessment of nanoscale bimetallic particles for groundwater treatment. Environ. Sci. Technol. 2001, 35, 49224926. (17) Glazier, R.; Venkatakrishnan, R.; Gheorghiu, F.; Walata, L.; Nash, R.; Zhang, W.-X. Nanotechnology takes root. Civil Eng. 2003, 73, 6469. (18) Schrick, B.; Hydutsky, B. W.; Blough, J. L.; Mallouk, T. E. Delivery vehicles for zerovalent metal nanoparticles in soil and groundwater. Chem. Mater. 2004, 16, 21872193. (19) He, F.; Zhao, D. Preparation and characterization of a new class of starch-stabilized bimetallic nanoparticles for degradation of chlorinated hydrocarbons in water. Environ. Sci. Technol. 2005, 39, 33143320. (20) Li, X.-Q.; Elliott, D. W.; Zhang, W.-X. Zero-valent iron nanoparticles for abatement of environmental pollutants: Materials and engineering aspects. Crit. Rev. Solid State Mater. Sci. 2006, 31, 111122. (21) He, F.; Zhao, D. Manipulating the size and dispersibility of zerovalent iron nanoparticles by use of carboxymethyl cellulose stabilizers. Environ. Sci. Technol. 2007, 41, 62166221. (22) Saleh, N.; Phenrat, T.; Sirk, K.; Dufour, B.; Ok, J.; Sarbu, T.; Matyjaszewski, K.; Tilton, R. D.; Lowry, G. V. Adsorbed triblock copolymers deliver reactive iron nanoparticles to the oil/water interface. Nano Lett. 2005, 5, 24892494. (23) Phenrat, T.; Saleh, N.; Sirk, K.; Tilton, R. D.; Lowry, G. V. Aggregation and sedimentation of aqueous nanoscale zerovalent iron dispersions. Environ. Sci. Technol. 2007, 41, 284290. (24) Hong, Y.; Honda, R. J.; Myung, N. V.; Walker, S. L. Transport of iron-based nanoparticles: Role of magnetic properties. Environ. Sci. Technol. 2009, 43, 88348839. (25) Petosa, A. R.; Jaisi, D. P.; Quevedo, I. R.; Elimelech, M.; Tufenkji, N. Aggregation and deposition of engineered nanomaterials in aquatic environments: Role of physicochemical interactions. Environ. Sci. Technol. 2010, 44, 65326549. (26) He, F.; Zhang, M.; Qian, T.; Zhao, D. Transport of carboxymethyl cellulose stabilized iron nanoparticles in porous media: Column experiments and modeling. J. Colloid Interface Sci. 2009, 334, 96102. (27) He, F.; Zhao, D.; Liu, J.; Roberts, C. B. Stabilization of FePd nanoparticles with sodium carboxymethyl cellulose for enhanced transport and dechlorination of trichloroethylene in soil and groundwater. Ind. Eng. Chem. Res. 2007, 46, 2934. (28) Phenrat, T.; Kim, H.-J.; Fagerlund, F.; Illangasekare, T.; Tilton, R. D.; Lowry, G. V. Particle size distribution, concentration, and magnetic attraction affect transport of polymer-modified Fe0 nanoparticles in sand columns. Environ. Sci. Technol. 2009, 43, 50795085. (29) Sakulchaicharoen, N.; OCarroll, D. M.; Herrera, J. E. Enhanced stability and dechlorination activity of pre-synthesis stabilized nanoscale FePd particles. J. Contam. Hydrol. 2010, 118, 117127. (30) Tosco, T.; Sethi, R. Transport of non-newtonian suspensions of highly concentrated micro- and nanoscale iron particles in porous media: A modeling approach. Environ. Sci. Technol. 2010, 44, 9062 9068.
dx.doi.org/10.1021/es3041412 | Environ. Sci. Technol. 2013, 47, 73327340

Environmental Science & Technology


(31) Phenrat, T.; Saleh, N.; Sirk, K.; Kim, H.-J.; Tilton, R. D.; Lowry, G. V. Stabilization of aqueous nanoscale zerovalent iron dispersions by anionic polyelectrolytes: Adsorbed anionic polyelectrolyte layer properties and their effect on aggregation and sedimentation. J. Nanopart. Res. 2008, 10, 795814. (32) Kim, H.-J.; Phenrat, T.; Tilton, R. D.; Lowry, G. V. Fe0 nanoparticles remain mobile in porous media after aging due to slow desorption of polymeric surface modifiers. Environ. Sci. Technol. 2009, 43, 38243830. (33) Sirk, K. M.; Saleh, N. B.; Phenrat, T.; Kim, H.-J.; Dufour, B.; Ok, J.; Golas, P. L.; Matyjaszewsk, K.; Lowry, G. V.; Tilton, R. D. Effect of adsorbed polyelectrolytes on nanoscale zero valent iron particle attachment to soil surface models. Environ. Sci. Technol. 2009, 43, 38033808. (34) OCarroll, D.; Sleep, B.; Krol, M.; Boparai, H.; Kocur, C. Nanoscale zero valent iron and bimetallic particles for contaminated site remediation. Adv. Water Resour. 2013, 51, 104122. (35) He, F.; Zhao, D. Hydrodechlorination of trichloroethene using stabilized Fe-Pd nanoparticles: Reaction mechanism and effects of stabilizers, catalysts and reaction conditions. Appl. Catal. B 2008, 84, 533540. (36) Wei, Y.-T.; Wu, S.-C.; Chou, C.-M.; Che, C.-H.; Tsai, S.-M.; Lien, H.-L. Influence of nanoscale zero-valent iron on geochemical properties of groundwater and vinyl chloride degradation: A field case study. Water Res. 2010, 44, 131140. (37) Zhang, W.-X. Nanoscale iron particles for environmental remediation: An overview. J. Nanopart. Res. 2003, 5, 323332. (38) Quinn, J.; Geiger, C.; Clausen, C.; Brooks, K.; Coon, C.; OHara, S.; Krug, T.; Major, D.; Yoon, W.-S.; Gavaskar, A.; Holdsworth, T. Field demonstration of DNAPL dehalogenation using emulsified zero-valent iron. Environ. Sci. Technol. 2005, 39, 13091318. (39) Henn, K.; Waddill, D. Utilization of nanoscale zero-valent iron for source remediationA case study. Remediation 2006, 5776. (40) Nanotechnology for Site Remediation Fact Sheet; U.S. Environmental Protection Agency: Washington, DC, 2008; p 17. (41) Bennett, P.; He, F.; Zhao, D.; Aiken, B.; Feldman, L. In situ testing of metallic iron nanoparticle mobility and reactivity in a shallow granular aquifer. J. Contam. Hydrol. 2010, 116, 3446. (42) He, F.; Zhao, D.; Paul, C. Field assessment of carboxymethyl cellulose stabilized iron nanoparticles for in situ destruction of chlorinated solvents in source zones. Water Res. 2010, 44, 23602370. (43) Tufenkji, N.; Elimelech, M. Correlation equation for predicting single-collector efficiency in physicochemical filtration in saturated porous media. Environ. Sci. Technol. 2004, 38, 529536. (44) Johnson, R.; Johnson, G.; Nurmi, J.; Tratnyek, P. Natural organic matter enhanced mobility of nano zerovalent iron. Environ. Sci. Technol. 2009, 43, 54555460. (45) Phenrat, T.; Kim, H.-J.; Fagerlund, F.; Illangasekare, T.; Lowry, G. V. Empirical correlations to estimate agglomerate size and deposition during injection of a polyelectrolyte-modified Fe0 nanoparticle at high particle concentration in saturated sand. J. Contam. Hydrol. 2010, 118, 152164. (46) Yao, K.-M.; Habibian, M.; OMelia, C. Water and waste water filtration: Concepts and applications. Environ. Sci. Technol. 1971, 5, 11051112. (47) Taghavy, A.; Costanza, J.; Pennell, K. D.; Abriola, L. M. Effectiveness of nanoscale zero-valent iron for treatment of a PCEDNAPL source zone. J. Contam. Hydrol. 2010, 118, 128142. (48) Interim In Situ Groundwater Remediation Report; Geomatrix, 2006. (49) Sleep, B. E.; Sykes, J. F. Compositional simulation of groundwater contamination by organic-compounds.1. Model development and verification. Water Resour. Res. 1993a, 29, 16971708. (50) Sleep, B. E.; Sehayek, L.; Chien, C. C. A modeling and experimental study of light nonaqueous phase liquid (LNAPL) accumulation in wells and LNAPL recovery from wells. Water Resour. Res. 2000, 36, 35353545.
7340

Article

(51) Reinecke, S. A.; Sleep, B. E. Knudsen diffusion, gas permeability, and water content in an unconsolidated porous medium. Water Resour. Res. 2002, 38, 161. (52) OCarroll, D. M.; Sleep, B. E. Hot water flushing for immiscible displacement of a viscous NAPL. J. Contam. Hydrol. 2007, 91, 247 266. (53) Sleep, B. E. Modeling transient organic vapor transport in porous media with the dusty gas model. Adv. Water Resour. 1998, 22, 247. (54) Kocur, C.; OCarroll, D.; Sleep, B. Impact of nZVI stability on mobility in porous media. J. Contam. Hydrol. 2013, 145, 1725. (55) Poling, B. E.; Prausnitz, J. M.; OConnell, J. P. The Properties of Gases and Liquids; McGraw-Hill: Toronto, 2001. (56) Holtz, R. D.; Kovacs, D. An Introduction to Geotechnical Engineering; Prentice-Hall: Englewood Clis, NJ, 1981. (57) Raychoudhury, T.; Naja, G.; Ghoshal, S. Assessment of transport of two polyelectrolyte-stabilized zero-valent iron nanoparticles in porous media. J. Contam. Hydrol. 2010, 118, 143151. (58) Tiraferri, A.; Sethi, R. Enhanced transport of zerovalent iron nanoparticles in saturated porous media by guar gum. J. Nanopart. Res. 2009, 11, 635645. (59) Shen, C.; Huang, Y.; Li, B.; Jin, Y. Predicting attachment eciency of colloid deposition under unfavorable attachment conditions. Water Resour. Res. 2010, 46. (60) Raychoudhury, T.; Tufenkji, N.; Ghoshal, S. Aggregation and deposition kinetics of carboxymethyl cellulose-modified zero-valent iron nanoparticles in porous media. Water Res. 2012, 46, 17351744. (61) Mishra, S.; Guyonnet, D. Analysis of observationWell response during constant-head testing. Ground Water 1992, 30, 523528. (62) Simpkin, T. J.; Palaia, T.; Petri, B. G.; Smith, B. A. In In Situ Chemical Oxidation for Groundwater Remediation; Siegrist, R. L., Crimi, M., Simpkin, T. J., Ward, C. H., Eds.; SERDP ESTCP Environmental Remediation Technology; Springer: New York, 2011; Vol. 3; pp 449 480. (63) Krug, T.; OHara, S.; Watling, M.; Quinn, J. Emulsied ZeroValent Nano-Scale Iron Treatment of Chlorinated Solvent DNAPL Source Zones; Environmental Security Technology Certication Program Oce (DOD): Arlington, VA, 2010 (64) Phenrat, T.; Liu, Y.; Tilton, R. D.; Lowry, G. V. Adsorbed polyelectrolyte coatings decrease Feo nanoparticle reactivity with TCE in water: Conceptual model and mechanisms. Environ. Sci. Technol. 2009, 43, 15071514.

dx.doi.org/10.1021/es3041412 | Environ. Sci. Technol. 2013, 47, 73327340

Вам также может понравиться