surface modifications enhance nanoiron transport and napl ......all water was deionized (di) by...

ENVIRONMENTAL ENGINEERING SCIENCEVolume 24, Number 1, 2007© Mary Ann Liebert, Inc.

Surface Modifications Enhance Nanoiron Transport and NAPLTargeting in Saturated Porous Media

Navid Saleh,1 Kevin Sirk,2 Yueqiang Liu,1 Tanapon Phenrat,1 Bruno Dufour,3Krzysztof Matyjaszewski,3 Robert D. Tilton,2,4 and Gregory V. Lowry1*

1Department of Civil & Environmental Engineering2Department of Chemical Engineering

3Department of Chemistry4Department of Biomedical Engineering

Carnegie Mellon UniversityPittsburgh, PA 15213-3890

ABSTRACT

Rapid in situ degradation of chlorinated solvents present as nonaqueous phase liquids (NAPL) can be ac-complished using reactive zerovalent nanoiron particles. Prior studies have shown that nanoiron transportin the subsurface is limited, and successful delivery of the nanoiron is essential for effective remediation.Here, the physical properties of bare and modified nanoiron are measured, and laboratory column reac-tors are used to compare the transport of three types of surface-modified nanoiron; triblock polymer-mod-ified, surfactant-modified, and a commercially available polymer-modified nanoiron. The effect of parti-cle concentration and solution ionic strength on the transport of each modified nanoiron is evaluated, andthe filtration mechanisms for bare and modified particles are determined in microfluidic flow cells andquartz crystal microbalance (QCM) experiments that probe the particle–collector grain interaction. Theeffect of surface modification on nanoiron reactivity is evaluated in batch experiments. Transport of mod-ified nanoiron does not directly correlate with �-potential or colloidal stability, but rather correlates to par-ticle–grain interactions. Filtration of bare nanoiron is caused by straining and subsequent clogging ratherthan by deposition to clean sand grains, suggesting that filter ripening models rather than clean bed fil-tration models should be used to describe nanoiron transport at high particle concentrations. Surface mod-ification decreased nanoiron reactivity by two to four times, but as high as a factor of nine depending onthe modifier used. Amphiphilic triblock copolymer modified nanoiron with a high hydrophobe/hydrophileratio shows promise for in situ targeting of NAPL, but requires further optimization.

Key words: environmental nanotechnology; colloid transport; groundwater remediation; NZVI

45

*Corresponding author: Civil and Environmental Engineering, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh,PA 15213. Phone: 412-268-2948; Fax: 412-268-7813; E-mail: [email protected]

INTRODUCTION

GROUNDWATER REMEDIATION is challenging whennonaqueous phase liquids (NAPL), particularly

chlorinated dense nonaqueous phase liquids (DNAPL),are present in the subsurface. No large DNAPL-impactedsites have been restored to meet drinking water standards(U.S. EPA, 1993; National Research Council of the Na-tional Academies, 2004). Subsurface heterogeneity andcomplex NAPL architecture make remediation difficult(Illangasekare et al., 1995; Dai et al., 2001; Daus et al.,2001), and currently available technologies includingpump-and-treat (Kerr, 1987), permeable reactive barriers(Sai and Anderson, 1992), in situ oxidation (InterstateTechnology and Regulatory Cooperation Work Group,2000), bio-enhanced natural attenuation (Sharma andMcCarty, 1996; Yang and McCarty, 2000), or thermaltreatment (Interstate Technology and Regulatory Coop-eration Work Group, 2000), do not provide efficientNAPL source zone remediation in some cases.

Recent investigations have demonstrated that nanoiron(Schrick et al., 2004; Liu et al., 2005a, 2005b; Song andCarraway, 2005), Fe0-based bimetallic nanoparticles (El-liott and Zhang, 2001) and emulsified zerovalent iron(EZVI) (Quinn et al., 2005) can effectively degrade sub-surface chlorinated solvents such as trichloroethylene(TCE). However, it has also been demonstrated that de-livering the nanoparticles to subsurface DNAPL poses asignificant challenge due to filtration by the soil (Schricket al., 2004). For nanoiron transporting through saturatedporous media containing DNAPL ganglia, the particlescan collide with and attach to soil grains, aggregate, andlead to pore plugging and straining, or can collide withand stick to entrapped DNAPL (Fig. 1). Filtration bystraining and attachment to soil grains will limit nano-iron transport and is undesirable. Nanoiron collisionswith and subsequent attachment to entrapped DNAPL isdesirable and should be promoted.

For nanoparticles (dp � 100 nm), Brownian diffusiontransports particles from the pore water to the vicinity ofa soil grain. Subsequent attachment leads to particle re-moval. Brownian diffusion is the dominant filtrationmechanism operable for nanoparticles (Yao et al., 1971).Nanoparticle aggregation into larger particles (100s ofnm to micron-sized particles) may make interception andgravity settling operable transport mechanisms as well(Tufenkji and Elimelech, 2004). Thus, formation of largernanoiron aggregates, and nanoiron affinity for the sur-faces of soil grains (e.g., as dictated by surface charge)would be expected to affect their transportability in thesubsurface.

It has been suggested that nanoiron is difficult to de-liver through a porous matrix, because of its tendency to

rapidly aggregate, that is, it is colloidally unstable(Hiemenz and Rajagopalan, 1997; Schrick et al., 2004).Nanoiron, whose surfaces are predominantly iron oxides,may also have a high affinity for the surfaces of soilgrains so low transport efficiency may also be due to ahigh attachment efficiency of nanoiron. Surface modifi-cation of the nanoiron can enhance transport through theporous media by increasing colloidal stability and de-creasing the affinity for the surfaces of soil grains (Salehet al., 2005a). The use of surfactant or polyelectrolytecoatings (Pincus, 1991; Rosen, 2002) as surface modi-fiers is a well-established technique to enhance colloidalstability. Adsorbed ionic surfactants or polyelectrolytesmay increase the surface charge of the nanoiron and leadto electrostatic stabilization. Adsorption of large poly-mers can also provide steric stabilization, and in the caseof adsorbed polyelectrolytes, a combination of electro-static and steric, that is, electrosteric stabilization arises.Appropriately designed amphiphilic triblock copolymercan also promote adsorption (i.e., target) of nanoiron atthe NAPL/water interface (Lin et al., 2003; Saleh et al.,2005a).

Targeting concept

The concept of targeting nanoiron to the NAPL sourcezone mimics targeted drug delivery systems (Allen andCullis, 2004), where drugs are coupled to biospecific li-gands that promote preferential accumulation in the dis-eased target tissue. Nanoiron NAPL source zone target-ing is similarly based on a bound delivery agent thatoffers preferential affinity for the target zone. It differs

46 SALEH ET AL.

Figure 1. Conceptual model of nanoiron transport showingfiltration, straining, and DNAPL targeting.

from drug delivery in that NAPL source zone targetinghas to be achieved merely through nonspecific thermo-dynamic interactions. The triblock copolymers used tomodify the nanoiron surfaces have three blocks (Fig. 2A),which are synthesized with three specific functionalities;that is, a poly(methacrylic acid) PMAA block that servesto anchor the polymer to the nanoiron surface, a hy-drophobic poly (methyl methacrylate) PMMA block thatimparts the desired thermodynamic affinity for NAPL,and an anionic, hydrophilic poly(styrene sulfonate) PSSblock that provides strong electrosteric repulsions andlimits aggregation and interaction with soil grains (Fig.2B). The hydrophobic block, which is collapsed in wa-ter, will swell when in contact with NAPL (the PSS blockcollapses in NAPL) to provide the particles with a ther-modynamic affinity to localize at NAPL/water interface(Saleh et al., 2005a) (Fig. 2C).

Objectives and approach

Here we investigate the ability of different surfacemodifications to improve nanoiron dispersion stability

and its transportability in water-saturated porous media,and assess the effect of modification on the reactivity andDNAPL/water partitioning behavior in situ (Saleh et al.,2005a). The study goals are to (1) better understand thefiltration mechanisms limiting nanoiron transport in sat-urated porous media, (2) demonstrate the ability of thesurface modifications to enhance transport of nanoironthrough porous media and elucidate the reasons for theenhanced transport, (3) assess the potential for DNAPLtargeting in situ, and (4) determine if surface modifiersaffect nanoiron reactivity with TCE. Bare nanoiron andnanoiron modified by surfactant or polymer adsorptionare characterized in terms of their surface charge, size,and dispersion stability to understand the effect of mod-ifiers on particle–particle interactions. A commercialpolymer-modified nanoiron is also evaluated. Bench-scale column experiments are performed to evaluate thetransportability of the particles as a function of particleconcentration and ionic strength. To elucidate the oper-able filtration mechanisms, the pore-scale transport be-havior of the particles is observed using microfluidic flow

NAPL IN SATURATED POROUS MEDIA 47

ENVIRON ENG SCI, VOL. 24, NO. 1, 2007

Figure 2. The targeting mechanism. (A) Adsorbed blockcopolymers contain a polyacrylic acid anchoring block with n de-gree of polymerization (black circles on B and C) that has highaffinity for iron oxide surfaces, a hydrophobic block with m de-gree of polymerization (gray circles on B and C) that has an affin-ity for DNAPL, and a polyelectrolyte block with p degree of poly-merization (open circles on B and C) that has an affinity for water;(B) the polyelectrolyte block is large enough to stably suspendparticles in water without aggregating, and the strong negativecharge in the polyelectrolyte block minimizes particle adhesionto negatively charged mineral or natural organic matter surfacesin the soil before reaching the DNAPL; (C) in water, the poly-electrolyte block swells and the hydrophobic block collapses, thereverse happens in the DNAPL phase and this amphiphilicity an-chors the particle at the DNAPL/water interface.

C

A B

cells, and direct nanoiron–silica sand grain interactionsare probed by using a quartz crystal microbalance (QCM)to measure nanoiron adsorption to a silica-coated sub-strate. The potential for in situ NAPL targeting is evalu-ated in column experiments by measuring the retentionof polymer-modified nanoiron in columns containing do-decane-coated sand relative to clean sand. Finally, the re-activity of modified nanoiron with TCE in batch reactorsis compared to that of bare nanoiron

EXPERIMENTAL

Materials

All water was deionized (DI) by reverse osmosis fol-lowed by final purification using ion exchange (U.S. Fil-ter Corp., Lowell, MA). Concentrated hydrochloric acid(trace metal grade, 35�%), nitric acid (reagent grade,69�%), and reagent grade anhydrous sodium bicarbon-ate, sodium nitrite, and sodium chloride were purchasedfrom Fisher (Fairlane, NJ). Dodecane (99 �%),trichloroethylene (99 �%), and dodecylbenzene-sulfonicacid, sodium salt (SDBS, 88%, MW 348.48 g/mol) wereobtained from Acros (Geel, Belgium).

Silica sand of approximately 300-�m average diame-ter (d50) was used as model soil grains (Agsco Corp., Has-brouck Heights, NJ). The sand was acid washed beforeuse with concentrated hydrochloric acid to remove thebackground iron oxide and rinsed with DI water until thepH returned to near neutral �7.0. Reactive nanoiron par-ticles (RNIP-Lot# 050401, Toda Kogyo Corp., Onoda,Japan) were used bare as received, or were modified withtriblock copolymer or surfactants as described below. Theparticles were shipped and stored in water at pH 10.6 andapproximately at 300-g/L concentration until used. A so-dium polyaspartate (MW of 2,000–3,000 g/mol) stabilizednanoiron, modified RNIP (MRNIP-Lot# 041002-1), wasalso supplied by Toda Kogyo Corp. and used as received.For in situ targeting experiments, acid-washed sand wasmodified with dodecane by homogenously coating thesand surfaces with 10% (vol/vol) dodecane containingOil-red-O (Acros) to make it visible. Dodecane was cho-sen as a model NAPL phase to avoid working with toxicDNAPL compounds such as TCE. Two triblock copoly-mer architectures were used (see Table 1, where subscriptsdenote block degrees of polymerization). The polymerswere prepared by atom transfer radical polymerization(ATRP) as described elsewhere (Matyjaszewski and Xia,2001; Lee et al., 2003; Saleh et al., 2005a, 2005b).

Methods

Preparation of particle suspensions. RNIP was modi-fied with the amphiphilic PMAA–PMMA–PSS triblock

copolymers or with SDBS surfactant. A 0.2-mL aliquotof the 300 g/L stock RNIP slurry was dispersed in 20 mLof 2 g/L polymer or surfactant in 1 mM NaHCO3 back-ground electrolyte to provide a 3-g/L slurry of polymer-modified or SDBS-modified RNIP. The mixture was thensonicated using an ultrasonic probe (Fisher Model 550)for 30 min, and the samples were rotated using an end-over-end rotator at 30 rpm to equilibrate for at least 72 h prior to use.

Particle characterization. Bare RNIP, polymer- orSDBS-modified RNIP, and MRNIP were characterized forsize and electrophoretic mobility using dynamic light scat-tering (DLS) and electrophoretic light scattering, respec-tively, using a Malvern Zetasizer Nano ZS (Table 1). TheCONTIN algorithm was used to convert intensity autocor-relation functions to intensity-weighted particle hydrody-namic diameter distributions, assuming the Stokes-Einsteinrelationship (Hiemenz and Rajagopalan, 1997) for spheri-cal particles. Measured electrophoretic mobilities were con-verted to apparent �-potentials using the Smoluchowski re-lationship (Hiemenz and Rajagopalan, 1997). The nanoirondispersion stability was determined by measuring the sed-imentation rate of nanoparticle suspensions (Table 1). Theoptical density (at � � 508 nm) of 0.08 wt% suspensionswas monitored for 2.5 h in a UV-Vis spectrophotometer.The dispersions were gently shaken by hand to break upany loosely formed flocs immediately prior to the floccu-lation/sedimentation measurements. The surface area ofbare RNIP was determined using a NOVA 2200 BET sur-face area analyzer after drying the samples in an inert atmo-sphere (Quantachrome, Boynton Beach, FL).

Bench-scale transport experiments. Transport studieswere performed in bench-scale column experiments.Stainless steel, 12.5-cm long (1.27-cm o.d. and 1.09-cmi.d.) columns with 1/16� end fittings were used. Columnswere packed wet (Roy and Dzombak, 1996; Lowry andReinhard, 2000) and then flushed with a 1-mM NaHCO3

solution for at least 10 pore volumes (PV) to obtain auniform surface charge on the sand. The nominal porewater velocity was maintained at 0.11 cm/s (93 m/day)with an HPLC pump (Alltech, Waukegan, IN, 301 HPLCpump). Porosity (0.33 � 0.01) and mean fluid residencetime (1.5 min) were determined from a NaNO2 tracer testusing an in-line UV-Vis spectrophotometer (Spectra-Physics 100, � � 230 nm, Spectra-Physics, Stratford,CT). The column was placed horizontally to simulate nat-ural groundwater conditions.

Prior to introducing nanoiron into the column, the par-ticles were suspended in a 1-mM NaHCO3 backgroundelectrolyte and sonicated using the ultrasonic probe for�30 min to disperse the particles. A 1.5-min (3-mL)

48 SALEH ET AL.

square pulse of particles was then introduced to the col-umn at 0.11 cm/s. The feed tank containing the nanoironsuspension was continuously sonicated in a sonicatingbath during the pulse to prevent any aggregation prior toentering the column. Eluted particles were collected atthe outlet of the column for 13 pore volumes. Particlesin the eluent sample were dissolved in HCl (1 to 4 vol/volratio), diluted with 1% HNO3 and analyzed for total ironusing flame atomic absorption spectrometry (GBC 908AA). Errors for the measured total iron concentration av-eraged �4%. The influent particle concentration wasmeasured the same way by assaying an aliquot of the son-icated particle suspension. Each transport experiment wasperformed in duplicate.

The ability of the particles to target the NAPL/wa-ter interface in situ followed the same method withsome modifications. Dodecane was added to dry sand(10% vol/vol) ex situ and was homogenized beforepacking the sand into the column. To improve detec-tion of nanoiron holdup in the columns, a lower nano-iron concentration was used (�120 mg/L). Experimentswere conducted with nanoiron suspensions that hadbeen presettled for 60 min, so only the most stable par-ticles were injected to the column to better distinguishholdup due to NAPL affinity from other holdup mech-anisms. The particles were introduced to the columnsimilar to the above, but at a velocity of 5.4 � 10�3

cm/s (4.66 m/day). After injecting 1 PV of particle sus-pension, the pump was turned off for 24 h. The lowerflow velocity and stagnation time was used to allow theparticles time to diffuse and attach to the dodecane-coated sand grains. Control experiments using cleansand were run under identical conditions. The pumpwas restarted after 24 h to flush the column with a 1-mM NaHCO3 solution and collect the eluent for totaliron analysis as above.

Quartz crystal microbalance (QCM). The instrumentused in these experiments was the quartz crystal mi-crobalance with dissipation (QCM-D) D300 from Q-Sense (Göteborg, Sweden), which is described in detailelsewhere (Rodahl et al., 1995). Briefly, in a QCM-D ex-periment, there are four separate resonant frequencies(overtones, n) used to drive the oscillation of the shearwave through the crystal: �5 MHz (n � 1), �15 MHz(n � 3), �25 MHz (n � 5), and �35 MHz (n � 7). Asmall mass added to the crystal induces a decrease in fre-quency. Where the Sauerbrey relation (Sauerbrey, 1959)holds, the change in frequency (f) is directly propor-tional to the adsorbed mass (m) give by Equation (1),

m � �C

nf (1)

where C is a constant based upon the physical properties ofthe quartz crystal [in this case C � 17.7 ng/(cm2 Hz)]. TheSauerbrey relation (Sauerbrey, 1959) is valid for thin, rigidfilms with negligible internal friction; 14-mm AT cut goldcrystals coated with SiO2 were used for all experiments. Thecrystals were cleaned using the following protocol: eachcrystal was immersed in a 2% sodium dodecylsulfate (SDS)solution for 1 h. The crystals were then rinsed with deion-ized H2O and dried with N2. The crystals were placed in aUV/Ozone chamber (Jetline Co., Irving, CA) for 15 minprior to use. All experiments were performed at T � 25°C �0.02°C. The f values for the third overtone were the mostconsistent, and were used in this paper.

Microfluidic cell experiments. A rectangular microflu-idic flow cell was made with poly (dimethylsiloxane)PDMS and glass having a dimension of 1.5� � 0.25�. Thecell had a depth of 300 �m to create a uniform mono-layer of sand. The cell had an inlet and an outlet, and asyringe pump was used to inject the particles. An opti-



Table 1. Physicochemical properties of unmodified and modified RNIP in 1 mM NaHCO3.

Electrophoreticmobility � potential Average Dia Stability

Particle type (�mcm/Vs) (mV) (nm) timea (s)

RNIP �2.32 � 0.22 �29.6 � 2.8 146 � 4 500PMAA48– �3.31 � 0.08 �42.3 � 1.5 212 � 21 5,031

PMMA17–PSS650

�RNIPPMAA42– �3.73 � 0.16 �47.6 � 2.0 178 � 11 1130

PMMA26–PSS462 �RNIP

SDBS � RNIP �3.00 � 0.05 �38.25 � 0.9 220 � 4 1,465MRNIP �2.95 � 0.07 �37.6 � 1.1 66 � 3 1,548

aStability time is defined as time to reach I/I0 � 0.5.

50 SALEH ET AL.

cal microscope (Nikon 2000U) was used to observe thepore-scale transport of the iron particles, and the imageswere taken using an IDT XS-4 high-resolution camera.

Reactivity study. Batch experiments were conducted in60-mL serum bottles capped by Teflon Mininert™ valves.Each contained 32 mL of deoxygenated water with 3 g/Lof bare RNIP or modified RNIP, and 30 mL of headspace.The bare and surface modified RNIP suspensions were pre-pared in the same manner as that for the bench-scale trans-port experiment except that the suspensions were deoxy-genated by purging with argon, and thus the reactorheadspace is argon. A 175-mL aliquot of saturated TCE so-lution (1,100 mg/L) was added to provide an initial TCEconcentration of 6 mg/L in solution. The reactors were ro-tated on an end-over-end rotator at 30 rpm at 22 � 2°C.TCE degradation, and the formation of products were mon-itored by periodically removing and analyzing a 100 �Lheadspace sample by GC/FID as previously described (Liuet al., 2005b). Replicate reactors were analyzed for eachparticle. TCE transformation rates were evaluated using akinetic modeling software package, Scientist, v.2.01 (Mi-cromath, St. Louis, MO). The loss of TCE and formationof products were fit concurrently using reaction pathwayspreviously proposed for RNIP (Liu et al., 2005b). Errorsreported for the observed reaction rate constants are 95%confidence intervals for the data fits.

RESULTS AND DISCUSSION

Effect of modification on particle properties

Particle modification by polymer or surfactant adsorp-tion generally increased the particle surface charge and sta-

bility, and to a lesser degree altered the particle hydrody-namic diameter (Table 1). The highly charged poly(styre-nesulfonate) block of the triblock copolymer (MW �91,000 to 125,000 g/moL) was designed to increase thenet charge of the bare RNIP, and it lead to the higher ob-served surface charge as intended. The anionic surfactant,SDBS, also has sulfonated head groups which increasedthe surface charge of RNIP, but to a lesser degree than thetriblock copolymer. MRNIP is stabilized using sodiumpolyaspartate (wt% ratio of 1:6 polymer:RNIP) with anMW of 2,000–3,000 g/moL. Polyaspartate is a biodegrad-able polycarboxylic dispersant that increases the net sur-face charge of RNIP (Wang et al., 2002).

The measured particle-size distributions are shown inFig. 3. Bare RNIP particles have a geometric mean hy-drodynamic diameter of 146 nm, which is larger than theaverage primary particle size determined from TEM mea-surements (Nurmi et al., 2005), indicating that bare nano-iron exists as aggregates of a few particles even at lowconcentration. Polymer modification of RNIP increasedthe particle diameter to �200 nm. This increase in hydro-dynamic diameter is consistent with the expected thick-ness for the adsorbed PSS brush (layer conformation of PSS polymer when adsorbed to nanoiron), given the PSS degrees of polymerization used here (Zhulina et al., 1992). SDBS-modified RNIP has a bimodal size dis-tribution, with peaks at 36 and 220 nm. A 2-g/L SDBSsuspension (no particles) had a peak near 40 nm, suggest-ing that SDBS micelles were present in the solution alongwith SDBS-modified particles with an average size of 220nm. MRNIP has an average particle diameter of 66 nm.

Colloidal suspensions of the sort used here are ther-modynamically unstable, and the term “stable” refers toa suspension that is sustained over a certain period oftime (Czigany et al., 2005), that is, kinetic stability. Thestability of bare RNIP is very low. Bare RNIP rapidlyaggregates and sediments from the suspension (Fig. 4).This is due to the strong van der Waals attraction asso-ciated with the high Hamaker constant of the magnetite(10�19 J) (Riew et al., 2005) shell as well as the mag-netic attractions between the particles (Viota et al., 2005;Phenrat et al., in press). Suspensions of MRNIP, or poly-mer- or SDBS-modified RNIP were significantly morestable (Fig. 4), indicating that the strong van der Waalsattractive forces that exist between the nanoiron particlesare overcome by the electrostatic repulsions from theSDBS charged head groups and electrosteric repulsionsfrom the polyelectrolyte of the triblock copolymers orpolyaspartic acid for MRNIP. The initial rate of aggre-gation of SDBS-stabilized RNIP and MRNIP were sim-ilar and more rapid than polymer-modified RNIP. Thismay be due to the somewhat higher surface charge ofpolymer-modified RNIP, but this is not likely, since the

Figure 3. Size distribution of particles (�30 mg/L) in 1 mMNaHCO3 (DLS intensity distribution).



polymer-modified RNIP charge is only slightly largerthan SDBS-stabilized RNIP. The slower aggregation ofpolymer-modified RNIP could also be attributed in partto its having the largest hydrodynamic diameter (lowestdiffusion coefficient), but again, the difference in sizecompared to SDBS-modified RNIP is small. Most likely,the improved colloidal stability is because both MRNIPand SDBS-modified RNIP are electrostatically stabilized,while polymer-modified RNIP is electrosterically stabi-lized (both electrostatic and steric). Electrosteric stabi-lization by adsorbed polyelectrolyte brushes is known tobe highly effective (Pincus, 1991; Biesheuvel, 2004).

Transport of bare and modified RNIP

RNIP (bare and modified) transport through sand-packed columns was evaluated at high (3-g/L) particleloadings and low ionic strength (1 mM) to mimic injec-tion into a contaminated sandy aquifer. The eluted massfor each of the particles is shown in Fig. 5. At 3 g/L, bareRNIP has very low transportability (1.4 � 3% mass elu-tion) through a saturated sand column at low ionicstrength. Retrieving the sand from the column and ana-lyzing for iron revealed that most particles were trappedwithin the first 1 to 2 cm of the column. MRNIP, poly-mer, and SDBS-modified RNIP elution was much higher,with the triblock copolymer and MRNIP elution at 95and 98%, respectively. SDBS was not as effective as thepolymer but still improved RNIP elution to 48%. Theseresults indicate that surface modification is essential forreasonable transport, even at low ionic strength. Inter-estingly, particle transport did not directly correlate withthe surface charge and stability of the particles, for ex-ample, SDBS had a higher zeta potential than MRNIP

but was transported less efficiently. The lower transportefficiency of SDBS-modified RNIP cannot be attributedto the desorption of SDBS from the RNIP surface dur-ing transport; as transport of SDBS-modified RNIP incolumns with sand surfaces that were precoated withSDBS and having 2 g/L SDBS during particle elution(hence, eliminating the possibility of desorption from thenanoiron surfaces due to no concentration gradient forthe SDBS to desorb) was also around 50%. Thus, parti-cle–particle interaction alone cannot predict transporta-bility and particle–collector grain interactions must alsobe considered as discussed below.

Effect of particle concentration on transport

Transport of bare RNIP depends highly on the parti-cle concentration (Fig. 6). At low concentration (180mg/L) more than half of the particles elute from the col-umn, but elutability decreases as the particle concentra-tion increases, and is highly inefficient (�2%) at the con-centration required to make field application economical(3 g/L) (Quinn et al., 2005). At this concentration, thehigher particle collision frequency makes RNIP aggre-gation rapid, and may promote cake filtration (Hunt etal., 1993; Mays and Hunt, 2005). Polymer-modifiedRNIP is elutable even at 3 g/L. Polymer modification de-creases particles aggregation, which could be the reasonfor the improved transport, but it is also possible that thepolymer coating improves elution efficiency by decreas-ing RNIP adhesion to the sand grains. Although it is notpossible to distinguish between these two effects in col-umn studies, the effect of polymer on RNIP–silica inter-actions is addressed by microfluidic cell and QCM ex-periments discussed below.

Figure 4. Sedimentation profiles of bare and modified RNIPmeasured by the optical density (I) at 508 nm. Optical densitydecreases as particles sediment below the incident light beam.

Figure 5. Percent mass of bare and modified RNIP elutedthrough a 12.5-cm silica sand column with porosity of 0.33.Modifying agents, that is PMAA–PMMA–PSS polymer orSDBS were added at 2 g/L concentration in each case. The ap-proach velocity was 93 m/day.

52 SALEH ET AL.

Effect of ionic strength on transport

Coatings that improve colloidal stability primarily byincreasing the surface charge of the particles (electro-static stabilization) are susceptible to changes in the so-lution ionic composition and strength. Groundwater typ-ically contains a variety of cations and anions due todissolution of minerals or anthropogenic inputs, andgroundwater ionic strengths can range from a few mil-limoles up to several hundred millimoles (Liu et al.,1995; Orth and Gillham, 1996; McCarthy et al., 2002;Sparks, 2003). Increasing the ionic strength of the sus-pension screens the surface charge of colloidal particlesand thus reduces EDL repulsion (Hiemenz and Ra-jagopalan, 1997). This results in lower colloidal stabil-ity and higher filtration probability (Elimelech, 1994;McCarthy et al., 2002). The elution data for bare andmodified RNIP at different ionic strength is shown inFig. 7. Adding a monovalent 1:1 electrolyte (NaCl) toincrease the ionic strength of the 1 mM NaHCO3 solu-tion from 1 to100 mM significantly decreased the trans-portability for the nanoiron particles. Bare RNIP is non-transportable at any ionic strength greater than 1 mM.SDBS-modified RNIP and MRNIP transport both de-crease significantly as the ionic strength is increased.Both of the PMAA–PMMA–PSS polymer-modifiedRNIP particles, however, eluted (�50%) even at 100mM [Na�]. Polymer-modified RNIP elution is superiorto either MRNIP or SDBS-modified RNIP at any ionicstrength examined. As the ionic strength increases,charge shielding occurs for all of the modifiers, but thesteric repulsions afforded by the high molecular weightpolymer allow it to be transportable even as the EDL

repulsions are screened. Even though MRNIP also of-fers electrosteric repulsions, polyaspartic acid has alower molecular weight than the PMAA–PMMA–PSSpolymer, and therefore provides less electrosteric re-pulsions than the larger molecular weigh PMAA–PMMA–PSS polymer. This result demonstrates thatelectrosteric repulsions are preferable for nanoirontransport at moderate or high ionic strength, and that thelarger molecular weight polymers are more effective.Divalent cations, such as Mg2� and Ca2� present ingroundwater will have an even greater effect as they aremore efficient at compressing the EDL. This is beingaddressed in ongoing research.

Pore-scale understanding of thetransport/filtration mechanisms

Column experiments can be used to measure the ef-fects of surface modifiers and geochemistry (ionicstrength) on the overall particle transport, but thismacroscopic method cannot distinguish between strain-ing and filtration due to attachment to sand grains. Tobetter understand the filtration mechanisms and the rea-sons for the enhanced transport of modified nanoironrelative to bare nanoiron, saturated sand-filled mi-crofluidic flow cells equipped with a digital camerawere used to observe RNIP filtration. QCM, where par-ticle adsorption to a silica surface is monitored as afunction of time, is used to qualitatively determine the“stickiness” of bare and modified RNIP to quartz sur-faces. Together, these techniques can be used to dis-tinguish between straining due to aggregation, and fil-tration due to attachment to sand grains.

Figure 6. Effect of nanoiron concentration on elution of bareand PMAA41–PMMA26–PSS462 modified RNIP through a 12.5-cm silica sand column with porosity of 0.33. Pore water ve-locity for the experiment was 93 m/day.

Figure 7. Effect of ionic strength on elution of bare and mod-ified RNIP through a 12.5-cm silica sand column with poros-ity of 0.33. Particle mass concentration is 3 g/L. All samplesalso contained 1 mM NaHCO3 to control pH at 7.4. The ap-proach velocity was 93 m/day.



Straining behavior of nanoiron in a microfluidic cell

Images from the microfluidic cell during transport ofbare RNIP indicate that RNIP particles stick to the silicasurface (Fig. 8a). The size of primary RNIP particles is be-low the optical resolution of the microscope, so it is evi-dent that RNIP aggregates are being observed to adhere tothe sand grains. A time series of microscope images (Fig.8b–c) indicates that after the initial attachment of RNIP,RNIP subsequently attaches to already attached RNIP ag-gregates, rather than to bare regions of the sand surface.This implies that the particle–particle interactions for bareRNIP are stronger than for RNIP–sand grain interactions.This is consistent with the sedimentation curves, whichshowed that bare RNIP aggregates rapidly. Continued

RNIP attachment eventually results in pore plugging (Fig.8d). Polymer-modified RNIP was also used in such flowcell experiments, and the particles did not stick to sandsurfaces and pore plugging did not occur. This behaviorsuggests that clean bed filtration models that typically con-sider only particle–sand grain interactions may not be ap-propriate for describing nanoiron transport at high particleconcentrations. Rather, filter ripening models that considerparticle–particle interactions and attachment are more suit-able (O’Melia and Ali, 1978).

Particle–collector grain interaction from QCM

The importance of particle–sand grain interactions inpredicting nanoiron transport is further demonstrated bythe QCM data for particle adhesion to a silica-coated

Figure 8. RNIP straining as an aqueous suspension of 3 g/L of PMAA41–PMMA26–PSS462 modified RNIP flows through awater-saturated monolayer of silica sand; (A) t � 1 min, (B) t � 5 min, (C) t � 10 min, and (D) example of pore plugging thatoccurred after longer times.

BA

D

C

54 SALEH ET AL.

crystal in 1 mM NaHCO3 solutions (Fig. 9). The down-ward shift in crystal resonant frequency change (indicat-ing direct particle attachment to the silica) is significantlyhigher for bare RNIP than for either MRNIP, PMAA–PMMA–PSS polymer-modified RNIP, or SDBS-modi-fied RNIP. Thus, the sticking coefficient and attachmentefficiency of bare RNIP should be higher than for theother particles, which is consistent with the elution data.Polymer-modified RNIP does not appear to adsorb to sil-ica (zero frequency change), a result of the strong elec-trosteric repulsions between the adsorbed polyelectrolytebrush and the negatively charged silica surface. SDBS-modified RNIP exhibit more adsorption to quartz withMRNIP showing some adsorption, but both are substan-tially less than for bare RNIP. The relative affinity forsilica surfaces as determined by QCM (polymer–RNIP �MRNIP � SDBS–RNIP �� bare RNIP) correlates wellwith the transport results and demonstrates that parti-cle–collector grain attachment is important but effec-tively eliminated by electrosterically stabilizing RNIP.

In situ targeting of NAPL

Delivering the particles to the NAPL source zone notonly demands the transportability of the particles but alsorequires the particles to have functionality to localize atthe NAPL–water interface. Surface modifications that im-part more hydophobicity to the particle should provide bet-ter NAPL targeting. MRNIP, which is modified withpolyaspartic acid did not demonstrate any affinity for do-

decane-coated sand grains and eluted completely (100 �1.9%), as it did for clean sand columns. Modifying RNIPparticles with the amphiphilic triblock copolymers showedsome promise for in situ targeting. For polymer-modifiedRNIP, the polymer with the low hydrophobe/hydrophile(PMMA/PSS) ratio (PMAA48–PMMA17–PSS650) did notdemonstrate any affinity for the dodecane-coated sand and eluted completely (100 � 2.1%). For the polymer with a high hydrophobe/hydrophile (PMMA/PSS) ratio(PMAA42–PMMA26–PSS462), particle elution was lower(90 � 2.4% elution) than it was for a clean sand column(100 � 2.6%) presumably due to RNIP adsorption to thedodecane–water interface. This adsorption is likely to beenhanced for a better solvent for PMMA (e.g., chloroformor TCE), but this must be verified. Using a higher hy-drophobe/hydrophile ratio, or changing the middle hy-drophobic block from methyl-methacrylate to butyl-methacrylate to lower the glass transition temperature andthus promote swelling of the hydrophobe in contact withNAPL may further enhance targeting. These targeting ex-periments indicate potential for in situ targeting, but addi-tional research to optimize the block size and type and hy-drophile/hydrophobe ratio are needed.

Effect of RNIP surface modification on reactivity

In comparison to the bare RNIP, surface modificationby SDBS, triblock copolymers with PSS-462 and withPSS-650 appears to decrease TCE dechlorination rate bya factor of 4 for SDBS and PSS-462, and a factor of 9for PSS-650 (Fig. 10). This implies that adsorbed layersand perhaps free molecules of the surface modifiers dis-solved in the suspension adversely interfere with the

Figure 10. Effect of polymer and surfactant modification onRNIP reactivity with TCE. Measured pseudofirst-order rateconstants are reported.

Figure 9. Downward shift in resonant frequency caused byadsorption of bare RNIP, MRNIP, or PMAA–PMMA–PSSpolymer-modified RNIP to a silica-coated QCM crystal,demonstrating that surface modification of RNIP significantlyreduces particle–silica adhesion relative to bare RNIP. Elec-trosterically stabilized polymer-modified RNIP adhesion to sil-ica is negligible.

dechlorination process. The decline of the TCE dechlo-rination rate by the surface modifiers correlates with in-creasing colloidal stability, PSS-650 � PSS-462~SDBS.This is consistent with differences in the structure of theadsorbed layers, which may be hindering TCE diffusionto the RNIP surface. Because the dose of SDBS used inthis study is above the critical micelle concentration(CMC � 1.4g/L, Rouse and Sabatini, 1993), partitioningof TCE into hydrophobic cores of SDBS micelles mightcompete with sorption of TCE onto reactive sites of RNIPand decrease the dechlorination rate at the RNIP surface(Loraine, 2001). It is possible that similar TCE interac-tions with the hydrophobic blocks of the triblock copoly-mers in solution may be occurring as well. These resultsindicate that there will ultimately be a tradeoff betweeneffective stabilization and transport, and the maximumreactivity afforded by nanoiron.

SUMMARY AND CONCLUSION

Surface modification of nanoiron is a promisingmethod to enhance its transport in saturated porous me-dia. Surfactant-modified nanoiron or electrostatically sta-bilized nanoiron can enhance transport, but becomes lesseffective as the ionic strength increases. Nanoiron that iselectrosterically stabilized by adsorbing amphiphilicPMAA–PMMA–PSS triblock copolymers demonstratedthe best ability to transport through saturated sand columnsat high ionic strength and high initial particle concentra-tion. Straining was the dominant mechanism of filtrationof bare RNIP, and therefore clean-bed filtration modelsmay not be appropriate for describing particle transport ofbare nanoiron. Polymer-modified RNIP showed no ten-dency to adhere to silica grain surfaces at low ionicstrength. Only polymer-modified RNIP with a high hy-drophobe/hydrophile ratio showed some evidence of in situtargeting. Surface modification enhanced transport butlowered reactivity by a factor of 2 to 10, depending on themodifier used indicating a tradeoff between transportabil-ity and reactivity. This laboratory-scale demonstration in-dicates the potential for nanoiron treatment of NAPL-im-pacted sites pending due optimization of polymer design.

ACKNOWLEDGMENTS

This research was funded in part by the Office of Sci-ence (BER), U.S. Department of Energy (DE-FG07-02ER63507), the U.S. EPA (R830898), and the NSF(CTS-0521721). We thank the Royal Thai Governmentfor the fellowship to Tanapon Phenrat, and the Wa-terQUEST Center at Carnegie Mellon University for lab-

oratory equipment and supplies. Any opinions, findings,and conclusions or recommendations expressed in thismaterial are those of the authors, and do not necessarilyreflect the views of the Department of Energy or UnitedStates Environmental Protection Agency.

REFERENCES

ALLEN, T.M., and CULLIS, P.R. (2004). Drug delivery sys-tems: Entering the mainstream. Science 303, 1818.

BIESHEUVEL, P.M. (2004). Ionizable polyelectrolyte brushes:Brush height and electrosteric interaction. J. Colloid Inter-face Sci. 275, 97.

CZIGANY, S., FLURY, M., and HARSH, J.B. (2005). Colloidstability in vadose zone Hanford sediments. Environ. Sci.Technol. 39, 1506.

DAI, D., BARRANCO, F.T., JR., and ILLANGASEKARE,T.H. (2001). Partitioning and interfacial tracers for differen-tiating NAPL entrapment configuration: Column-scale in-vestigation. Environ. Sci. Technol. 35, 4894.

DAUS, D.A., KENT, B., and MOSQUERA, G.C.B. (2001). Acase study of DNAPL remediation in Northwestern Brazil.J. Environ. Sci. Health. 36, 1505.

ELIMELECH, M. (1994). Particle deposition on ideal collec-tors from dilute flowing suspensions: Mathematical formu-lation, numerical solution, and simulations. Separat. Tech-nol. 4, 186.

ELLIOTT, D.W., and ZHANG, W.-X. (2001). Field assessmentof nanoscale bimetallic particles for groundwater treatment.Environ. Sci. Technol. 35, 4922.

HIEMENZ, P.C., and RAJAGOPALAN, R. (1997). Principlesof Colloid and Surface Chemistry. New York: MarcelDekker.

HUNT, J.R., HWANG, B.-C., and MCDOWELL-BOYER,L.M. (1993). Solids accumulation during deep bed filtration.Environ. Sci. Technol. 27, 1099.

ILLANGASEKARE, T.H., YATES, D.N., and ARM-BRUSTER, E.J. (1995). Effect of heterogeneity on transportand entrapment on nonaqueous phase waste products inaquifers: An experimental study. J. Environ. Eng. 121, 572.

INTERSTATE TECHNOLOGY AND REGULATORY CO-OPERATION WORK GROUP. (2000). Dense Non-AqueousPhase Liquids (DNAPLs): Review of Emerging Characteri-zation and Remediation Technologies. Washington, DC: In-terstate Technology and Regulatory Council.

KERR, R.S. (1987). Practical Limits to Pump and Treat Tech-nology for Aquifer Remediation. Ada, OK: EnvironmentalResearch Laboratory.

LEE, S.B., RUSSELL, A.J., and MATYJASZEWSKI, K.(2003). ATRP synthesis of amphiphilic block, gradient andrandom copolymers of 2-(dimethylamino)ethyl methacrylate



and n-butyl methacrylate in aqueous media. Biomacromole-cules 4, 1386.

LIN, Y., SKAFF, H., EMRICK, T., DINSMORE, D., and RUS-SELL, T.P. (2003). Nanoparticles at liquid–liquid interfaces:Assembly, displacement and transport. Science 299, 226.

LIU, D., JOHNSON, P.R., and ELIMELECH, M. (1995).Colloid deposition dynamics in flow-through porous me-dia: Role of electrolyte concentration. Environ. Sci. Tech-nol. 29, 2963.

LIU, Y., CHOI, H., DIONYSIOU, D., and LOWRY, G.V.(2005a). Trichloroethene hydrodechlorination in water byhighly disordered monometallic nanoiron. Chem. Mater. 17,5315.

LIU, Y., MAJETICH, S.A., TILTON, R.D., SHOLL, D.S., andLOWRY, G.V. (2005b). TCE dechlorination rates, pathwaysand efficiency of nanoscale iron particles with different prop-erties. Environ. Sci. Technol. 39, 1338.

LORAINE, G.A. (1993). Effects of alcohols and nonionic sur-factants on the reduction of PCE and TCE by zero-valentiron. Wat. Res. 35, 1453.

LOWRY, G.V., and REINHARD, M. (2000). Pd-catalyzedTCE dechlorination in groundwater: Solute effects, biologi-cal control, and oxidative catalyst regeneration. Environ. Sci.Technol. 34, 3217.

MATYJASZEWSKI, K., and XIA, J. (2001). Atom transfer rad-ical polymerization. Chem. Rev. 101, 2921.

MAYS, D.C., and HUNT, J.R. (2005). Hydrodynamic aspectsof clogging in porous media. Environ. Sci. Technol. 39,577.

MCCARTHY, J.F., MCKAY, L.D., and BRUNER, D.D.(2002). Influence of ionic strength and cation charge on trans-port of colloidal particles in fractured Shale Saprolite. Envi-ron. Sci. Technol. 36, 3735.

NATIONAL RESEARCH COUNCIL OF THE NATIONALACADEMIES. (2004). Contaminants in the Subsurface:Source Zone Assessment and Remediation. Washington, DC:National Academic Press.

NURMI, J.T., TRATNYEK, P.G., SARATHY, V., BAER,D.R., AMONETTE, J.E., PECHER, K., WANG, C., LINE-HAN, J.C., MATSON, D.W., PENN, R.L., et al. (2005).Characterization and properties of metallic iron nanoparti-cles: Spectroscopy, electrochemistry, and kinetics. Environ.Sci. Technol. 39, 1221.

O’MELIA, C.R., and ALI, W. (1978). The role of retained par-ticles in deep bed filtration. Prog. Water Technol. 10, 167.

ORTH, W.S., and GILLHAM, R.W. (1996). Dechlorination oftrichloroethene in aqueous solution using Fe0. Environ. Sci.Technol. 30, 66.

PHENRAT, T., SALEH, N., SIRK, K., TILTON, R., andLOWRY, G.V. Aggregation and sedimentation of aqueousnanoiron dispersions. Environ. Sci. Technol. (in press).

PINCUS, P. (1991). Colloidal stabilization with grafted poly-electrolytes. Macromolecules 24, 2912.

QUINN, J., GEIGER, C., CLAUSEN, C., BROOKS, K.,COON, C., O’HARA, S., KRUG, T., MAJOR, D., YOON,W.-S., GAVASKAR, A., et al. (2005). Field demonstrationof DNAPL dehalogenation using emulsified zero-valent iron.Environ. Sci. Technol. 39, 1309.

RIEW, C.K., ROWE, E.H., and SIEBERT, A.R. (1976). Tough-ness and brittleness of plastics. Adv. Chem. Series, Tough-ness Brittleness Plast., Symp. 1974. 154, 326.

RODAHL, M., HOOK, F., KROZER, A., BRZEZINSKI, P.,and KASEMO, B. (1995). Quartz crystal microbalance setupfor frequency and Q-factor measurements in gaseous and liq-uid environments. Rev. Sci. Instrum. 66, 3924.

ROSEN, M.J. (2002). Surfactants and Interfacial Phenomena.New York: Wiley-Interscience.

ROUSE, J.D., and SABATINI, D.A. (1993). Minimizing sur-factant losses using twin-head anionic surfactants in subsur-face remediation. Environ. Sci. Technol. 27, 2072.

ROY, S.B., and DZOMBAK, D.A. (1996). Colloid release andtransport processes in natural and model porous media. Col-loids Surf.: Physicochem. Eng. Aspect. 107, 250.

SAI, J.O., and ANDERSON, D.C. (1992). Barrier wall materi-als for containment of Dense Nonaqueous Phase Liquid(DNAPL). Hazard. Waste Hazard. Mater. 9, 317.

SALEH, N., PHENRAT, T., SIRK, K., DUFOUR, B., OK, J.,SARBU, T., MATYJASZEWSKI, K., TILTON, R.D., andLOWRY, G.V. (2005a). Adsorbed triblock copolymers de-liver reactive iron nanoparticles to the oil/water interface.Nanoletters 5, 2489.

SALEH, N., SARBU, T., SIRK, K., LOWRY, G.V., MATY-JASZEWSKI, K., and TILTON, R.D. (2005). Oil-in-wateremulsions stabilized by highly charged polyelectrolyte-grafted silica nanoparticles. Langmuir 21, 9873.

SAUERBREY, G.Z. (1959). Use of quartz crystal vibrator forweighting thin films on a microbalance. Zeitschrift FurPhysik 155(2):206–222.

SCHRICK, B., HYDUTSKY, B., BLOUGH, J., and MAL-LOUK, T. (2004). Delivery vehicles for zerovalent metalnanoparticles in soil and groundwater. Chem. Mater. 16,2187.

SHARMA, P.K., and MCCARTY, P.L. (1996). Isolation and characterization of facultative aerobic bacteria that reductively dehalogenate tetrachloroethylene to cis-1,2-dichloroethane. Appl. Environ. Microbiol. 62, 761.

SONG, H., and CARRAWAY, E.R. (2005). Reduction of chlo-rinated ethanes by nanosized zero-valent iron: Kinetics,pathways, and effects of reaction conditions. Environ. Sci.Technol. 39, 6237.

SPARKS, D.L. (2003). Environmental Soil Chemistry. SanDiego, CA: Academic Press.

56 SALEH ET AL.



TUFENKJI, N., and ELIMELECH, M. (2004). Correlationequation for predicting single-collector efficiency in physico-chemical filtration in saturated porous media. Environ. Sci.Technol. 38, 529.

U.S. ENVIRONMENTAL PROTECTION AGENCY (U.S.EPA). (1993). Guidance for Evaluating the Technical Im-practicability of Groundwater Restoration. Washington, DC:Author, p. 2.

VIOTA, J.L., VICENTE, J. DE, DURAN, J.D.G., and DEL-GADO, A.V. (2005). Stabilization of magnetorheologicalsuspensions by polyacrylic acid polymers. J. Colloid Inter-face Sci. 284, 527.

WANG X.Y., LEE B.I., and MANN L. (2002). Dispersion ofbarium titanate with polyaspartic acid in aqueous media. Col-loids Surf., A. 202, 71.

YANG, Y.R., and MCCARTY, P.L. (2000). Biologically en-hanced dissolution of tetrachloroethylene DNAPL. Environ.Sci. Technol. 34, 2979.

YAO, K.M., HABIBIBIAN, T., and O’MELIA, C.R. (1971).Water and waste water filtration. Concepts and applications.Environ. Sci. Technol. 5, 1105.

ZHULINA, E.B., BORISOV, O.V., and BIRSHTEIN, T.M.(1992). Structure of grafted polyelectrolyte layer. J. Phys. IIFrance 2, 63.

surface modifications enhance nanoiron transport and napl ......all water was deionized (di) by...

Documents