Aggregation and Deposition Kinetics of Carboxymethyl Cellulose-Modified

water research 46 (2012) 1735e1744

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Aggregation and deposition kinetics of carboxymethyl cellulose-modiﬁed zero-valent iron nanoparticles in porous media

Trishikhi Raychoudhury a, Nathalie Tufenkji b, Subhasis Ghoshal a,* a Department of Civil Engineering, McGill University, Montreal, Quebec H3A 2K6, Canada b Department of Chemical Engineering, McGill University, Montreal, Quebec H3A 2B2, Canada article info abstract

Article history: Transport and deposition of carboxymethyl cellulose (CMC)-modified nanoparticles of Received 6 August 2011 zero-valent iron (NZVI) were investigated in laboratory-scale sand packed columns. Received in revised form Aggregation resulted in a change in the particle size distribution (PSD) with time, and the 18 December 2011 changes in average particle size were determined by nanoparticle tracking analysis (NTA). Accepted 19 December 2011 The change in PSD over time was influenced by the CMC-NZVI concentration in suspen- Available online 30 December 2011 sion. A particleeparticle attachment efficiency was evaluated by fitting an aggregation model with NTA data and subsequently used to predict changes in PSD over time. Changes Keywords: in particle sizes over time led to corresponding changes in single-collector contact effi- Nanoparticle transport ciencies, resulting in altered particle deposition rates over time. A coupled aggregation- Groundwater remediation colloid transport model was used to demonstrate how changes in PSD can reduce the Colloid aggregation transport of CMC-NZVI in column experiments. The effects of particle concentrations in Colloid deposition the range of 0.07 g L 1 to 0.725 g L 1 on the transport in porous media were evaluated by Ionic strength comparing the elution profiles of CMC-NZVI from packed sand columns. Changes in PSD Chlorinated solvents over time could reasonably account for a gradual increase in effluent concentration between 1 and 5 pore volumes (PVs). Processes such as detachment of deposited particles also likely contributed to the gradual increase in effluent concentrations. The particleecollector attachment efficiency increased with CMC-NZVI particle concentration due þ to a rise in dissolved Na concentration with increased addition of Na-CMC. This inad- vertent change in ionic strength led to decreased effluent concentrations at higher CMC- NZVI concentrations. ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction eliminate or transform certain pollutants rapidly, its transport to target locations and its reactivity may be limited by its An emerging remediation technique for in situ groundwater rapid aggregation (Phenrat et al., 2007). Several studies have remediation of chlorinated solvent and heavy metal- demonstrated improved colloidal stabilization of NZVI by contaminated aquifers is the injection of reactive nano- coating them with polymers or polyelectrolytes (Kanel et al., particles of zero-valent iron (Tratnyek and Johnson, 2006; 2007; Phenrat et al., 2008; Saleh et al., 2007; Schrick et al., Zhang, 2003). Although at appropriate doses, NZVI can 2004). Polyelectrolyte or polymer coatings can inhibit NZVI

* Corresponding author. Tel.: þ1 514 398 6867; fax: þ1 514 398 7361. E-mail address: [email protected] (S. Ghoshal). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.12.045 1736 water research 46 (2012) 1735e1744

1 Nomenclature kdep particle deposition rate coefﬁcient [T ]

3 kdep,i deposition rate coefficient of the ith sized particles C effluent particle mass concentration [M L ] 1 3 [T ] C0 influent particle mass concentration [M L ] 1 3 kdet particle detachment rate coefficient [T ] Ci mass concentration of ith size aggregate [M L ] 2 1 massi mass of ith particles [M] D hydrodynamic dispersion coefficient [L T ] n particle number concentration [L 3] Df fractal dimension of aggregates ni particle number concentration of ith aggregates Ni number of single particles forming the aggregates 3 [L ] S solid phase deposited mass density [M M 1] dep t time [T] T temperature 1 2 2 v Darcy’s velocity [L T ] Tadhesive adhesive torque [M L T ] x distance along column length [L] T torque applied [M L2 T 2] applied a particleecollector attachment efficiency Z maximum number of size categories pc a particleeparticle attachment efficiency d number-averaged mean diameter measured by pp Navg b particleeparticle collision frequency function NTA [L] [L3 T 1] d number-averaged mean diameter calculated from Navg 3 porosity model [L] h single-collector contact efficiency d mean collector diameter [L] 0 c h single-collector contact efficiency of ith sized d particle diameter of ith aggregates [L] 0,i p(i) aggregate d ¼ diameter of the smallest particle size [L] p(i 1) r bulk density of sand [M L 3]

aggregation by a combination of electrostatic and steric particles. The observed deposition behavior was quantita- repulsion forces when these interactions are large enough to tively explained by particle-size dependent deposition rate overcome inter-particle magnetic and van der Waals attrac- coefficients. Although those studies have demonstrated that tive forces. The colloidal stabilization of NZVI by polymers changes in particle size resulting from aggregation influence has been shown to reduce the deposition of NZVI on granular transport, the role of aggregation kinetics or time-varying media and thus increase its mobility, relative to that of bare PSD on nanoparticle transport behavior has not been NZVI (He et al., 2009; Kanel et al., 2007; Phenrat et al., 2008). examined. Polymeric coatings of NZVI are unlikely to provide The objectives of this study were to assess the rates of complete stabilization because of incomplete or insufficient aggregation of CMC-NZVI particles, as well as the effects of surface coverage (Cirtiu et al., 2011). The extent of colloidal aggregation kinetics on CMC-NZVI transport and deposition in stabilization is also dependent on the polymer characteristics, granular porous media at different nanoparticle concentra- the thickness of the polymer layer on the particle surface, and tions. Changes in the equivalent mean diameter of a CMC- on the solution chemistry (Phenrat et al., 2008; Saleh et al., NZVI suspension due to aggregation were measured over 2008, 2007). Even at g L 1 doses of polymers and poly- a time period equivalent to the particle residence time during electrolytes there is evidence of incomplete stabilization a transport experiment. An aggregation kinetics model was within minutes (Phenrat et al., 2008, 2007; Raychoudhury fitted to this data to obtain the particleeparticle attachment a et al., 2010; Saleh et al., 2007). Thus, it is important to assess efficiency ( pp) and calculations were performed to determine the changes in PSD of polymer-stabilized NZVI due to the change in the PSD with time during the transport experi- aggregation. ments. In an effort to better understand the role of nano- The influence of aggregation of polyelectrolyte-modified particle aggregation kinetics on nanoparticle transport, NZVI on the deposition and transport of these particles in a modified colloid transport model which couples the gov- granular media has been discussed in recent studies (Petosa erning transport equation to the aggregation kinetics equation et al., 2010; Phenrat et al., 2010, 2009; Raychoudhury et al., was developed. A schematic of the multiple concurrent 2010). Those studies suggested that aggregation of the nanoparticle aggregation, deposition and transport processes nanoparticles will influence their deposition in granular occurring in granular porous media is shown in Fig. 1. Column media and that the particle size determines the magnitude of transport experiments were performed for different nano- h the single-collector contact efficiency ( 0), which is defined particle concentrations, and the effluent CMC-NZVI concen- as the ratio of the number of collisions between the particle trations over time were fitted to the modified colloid transport and the collector and the total number of particles model to assess whether particle-size dependent deposition approaching the collector. The effect of colloid aggregate size rates could account for the observed shape of the experi- on colloid transport in porous media has been examined by mental breakthrough curves. The potential role of other Chatterjee and Gupta (2009). In their study, injection of nanoparticle deposition mechanisms such as blocking of a polydisperse suspension of inert micron-sized particles deposition sites, detachment of nanoparticles and the composed of silica and metal oxides in a column packed with changes in electrolyte composition with CMC-NZVI concen- borosilicate glass beads resulted in larger particles being tration, were also considered for explaining the observed deposited at shallower bed depths compared to the smaller breakthrough curves. water research 46 (2012) 1735e1744 1737

The particle sizes of CMC-NZVI over time in aqueous 2. Materials and methods suspensions were determined by nanoparticle tracking analysis (NanoSight LM10). Prior to CMC-NZVI size measurement, 2.1. Synthesis of CMC-NZVI the NTA accuracy was verified by checking the size of stan- dard NanoSight polystyrene latex particles of 100 nm and CMC-NZVI was prepared according to the methods described 200 nm diameter. For CMC-NZVI size measurement, suspen- by He and Zhao (2007). Briefly, an aqueous solution of 0.065 M sions of 0.07, 0.2 and 0.725 g L 1 CMC-NZVI (as total Fe) were $ FeSO4 7H2O (AlfaAesar, purity greater than 99%) was added to m prepared in N2 atmosphere with filtered (0.22 m) degassed a5gL 1 of Na-CMC (90 K, Aldrich) solution and mixed thor- 0.1 mM NaHCO3 and sonicated by a 40 kHz ultrasonic cleaner oughly for 30 min. The solution was then reduced by the drop- (Cole-Palmer 8891) for 10 min to ensure homogeneity of the e wise addition of a solution of NaBH4 (Sigma Aldrich) at a rate suspensions, and the suspension was filled in a vial to zero 1 of 55.5 mg min under an N2 atmosphere. The ratio of [BH4 ]/ headspace. The pH of each particle suspension was 7.4 0.4. 2þ [Fe ] was set at 2.0. The mixture was then stirred for an Aliquots were withdrawn at various time points from vials additional 30 min before being dried overnight and finally the subjected to slow stirring, and then diluted to 5 mg L 1. The 2þ dry powder was stored under N2. The ratio of [CMC]/[Fe ] was mean square displacements of single particles were deter- 3 1.6 10 . This is the lowest CMC/Fe ratio that provides mined by tracking the path of scattered light using the a unimodal particle size distribution and a mean particle NanoSight software. With NTA, the diffusion coefficients are diameter of less than 100 nm. Other details of the synthesis determined from a series of single-particle Brownian motion conditions and characterizations are described in our tracks over time, rather than the collective light scattering previous study (Cirtiu et al., 2011). response from all nanoparticles in dynamic light scattering which results in more biased toward the larger-sized particles 2.2. Particle size analysis of CMC-NZVI for assessment in the latter technique (Domingos et al., 2009). of aggregation kinetics 2.3. Packed column experiments The above process yielded CMC-NZVI particles with a mean diameter of 70 nm as determined by high resolution (Philips A Kontes Chromoflex (Fisher Scientific) column of 1 cm i.d. CM200) transmission electron microscopy (TEM) operated at and bed depth of 9 cm was used. A nylon mesh (100 mm 200 kV and 120 kV. A typical TEM image of CMC-NZVI parti- opening size) was placed at the bottom of the packed column cles is presented in the Supplementary data (Fig. S1). Samples to prevent sand grains from being displaced into the tubing. for TEM analysis were prepared by depositing two droplets of The column was packed with graded silica sand (Unimin nanoparticle suspension onto carbon-coated 400 mesh Corp., #4045) sieved through an F30 sieve and retained on a F50 copper grids under N2 in order to avoid oxidation, and the sieve. The sand was acid-washed with concentrated HCl, diameters were measured using the General Image Manipu- rinsed with deionized (DI) water and oven dried for 3 h at m lating Software (GIMP Software, GNU) by averaging 100 550 C. The mean size of the graded sand (d50) was 375 m. The particles on different TEM grids. From the TEM images, the column was dry-packed with sand at several stages with CMC-layer thickness was approximated to be in the range of intermittent vibration to ensure uniform packing and then e 5 10 nm, which is in agreement with CMC-layer thicknesses saturated with CO2 to remove air bubbles. Next, 12 PVs of reported in prior studies (Fatisson et al., 2010; Raychoudhury background electrolyte (0.1 mM NaHCO3) were injected in et al., 2010). For calculating the single-collector contact effi- column using a syringe pump (KDS 200, KD Scientific) for h e ciency ( 0) and the particle particle collision frequency conditioning of the column prior to injecting a suspension of function (b), an average CMC-layer thickness of 7 nm was CMC-NZVI. The porosity of the packed granular medium in considered. the column was determined to be 0.32 using methods

Fig. 1 e Schematic of concurrent aggregation and particle deposition processes in porous media. 1738 water research 46 (2012) 1735e1744

described elsewhere (Lambe and Robert, 1991). The value of pooled from duplicate experiments conducted for each the hydrodynamic dispersion coefﬁcient (D) was estimated to concentration, with particle diameter measurements taken at be 0.3 cm2 min 1 from breakthrough curves obtained with an different time points. The extent of aggregation of CMC-NZVI inert tracer (KNO3). over approximately 1 h resulted in an increase in dNavg of

CMC-NZVI suspensions of different nanoparticle mass 28 nm (from dNavg(t¼0 min) 190 nm to dNavg(t¼65 min) 218 nm) for concentrations (0.07 g L 1, 0.2 g L 1 and 0.725 g L 1 as total Fe) the 0.2 g L 1 CMC-NZVI suspension and 76 nm (from 1 in aqueous solution of 0.1 or 10 mM NaHCO3 were prepared in dNavg(t¼0 min) 194 nm to dNavg(t¼60 min) 270 nm) for the 0.725 g L aN2 atmosphere with sonication as described above for the suspension. The change in particle size with time of the NTA analyses. The column influent suspension was pumped 0.07 mg L 1 CMC-NZVI particle suspension was found to be through the column at a rate of 0.445 cm min 1 using a syringe negligible but there was aggregation at higher particle pump. This velocity was chosen as it is intermediate between concentrations. Increasing particle concentrations result in typical groundwater and NZVI injection velocities. The greater number of particleeparticle collisions and an increase groundwater velocity beyond the point of injection would be in the extent of aggregation. lower than the injection velocity and higher than the back- The NTA data in Fig. 2a for the 0.2 and 0.725 g L 1 ground groundwater velocity, and thus an intermediate suspension was fitted to particle aggregation kinetics model velocity is relevant. The particle residence time in the reservoir proposed by Lee et al. (2000) and described in Equation (1). syringe ranged from 0 to 48 min, and the average residence X Xz dn 1 time in the column was 6 min. Settling of the CMC-NZVI in the k ¼ a bði; jÞn n a n bði; kÞn (1) dt 2 pp i j pp k i syringe was prevented by gentle stirring as in the aggregation iþj¼k i¼1 experiments. The concentration of nanoparticles entering the where, ni, nj and nk are the i, j and k-mer particle number column was thus maintained constant over time as verified by concentrations, and z is the maximum number of size cate- monitoring of the column influent suspension using flame gories. The first term on the right-hand side of Equation (1) atomic absorption spectrometry (AAS, Perkin Elmer 3110). accounts for the formation of aggregates of size k as a result For AAS analyses of Fe, the collected effluent samples were dissolved in concentrated HCl (1:2 v/v ratio), diluted with 5%

HNO3 and analyzed for total Fe at a wavelength of 249.7 nm. a d -1 The detection limit of AAS for Fe measurement was Navg (C0=0.07gL ) 1 d -1 0.001 mg L and all the samples have concentrations above Navg (C0=0.20gL ) the detection limit. The presented data is the average of d -1 Navg (C0=0.725gL ) replicate experiments. The experimental conditions are * d Navg summarized in Table 1.

3. Results and discussion Mean particle size (nm)

3.1. Aggregation kinetics of CMC-NZVI 180 210 240 270 300 0 102030405060 The change in the mean diameter (dNavg) of the CMC-NZVI Time (min) over time in a 0.07 g L 1, 0.2 g L 1 and 0.725 g L 1 suspension 20 prepared in 0.1 mM NaHCO3, as determined by NTA, is pre- b sented in Fig. 2a. The data points shown in the ﬁgure are

Table 1 e Specifications of experimental conditions. 10 Column length L 9cm CMC-NZVI concentration 0.07, 0.2 or 0.725 g L 1 concentration (mM) concentration 5 of the column feed C0 + 1 Interstitial velocity v 0.445 cm min Na Packing density of dry 1.95 g cm 3 r 0 sand in the packed-bed b 0 200 400 600 800 1000 1200 1400 Column void fraction 0.32 -1 (or porosity) 3 Concentration of CMC-NZVI (mgL ) Mean diameter of the 375 mm e sand particles dc Fig. 2 (a) Changes in mean particle size over time for Longitudinal dispersion 0.3 cm2 min 1 suspensions of CMC-NZVI obtained by nanoparticle coefficient D tracking analysis (NTA). The lines show the best fit of the 3 3 L L Density of the nanoparticles 6.7 10 kg m aggregation kinetics model to the 0.2 g L 1 and 0.725 g L 1 Electrolyte solution for 0.1 mM NaHCO D 3 CMC-NZVI. (b) Change in Na concentration with increase preparation of particle suspensions pH 7.4 0.4 in CMC-NZVI concentration in suspensions prepared in 0.1 mM NaHCO3. water research 46 (2012) 1735e1744 1739

of aggregation of particles of size i and j. The second term on Table 2 e Fitted values of particleeparticle attachment the right hand-side accounts for the loss of particles of size a e efficiency ( pp), particle collector attachment efficiency k as a result of collisions that lead to further aggregation. a k ( pc) and detachment rate constant ( det) for different Aggregation kinetics is controlled by the particleeparticle experimental conditions. a attachment efficiency, pp, which is the fraction of parti- a a Experimental Effective pp pc kdet cleeparticle collisions resulting in attachment and is depen- condition IS (mM) (min 1) dent on solution chemistry, and by the particleeparticle ¼ 1 e b b C0 0.07 g L 1.2 0.05 0.005 collision frequency function, . The parameter accounts for ¼ 1 4 C0 0.20 g L 3.2 1.85 10 0.07 0.005 Brownian motion, fluid shear and differential sedimentation. ¼ 1 4 C0 0.725 g L 9.6 3.08 10 0.12 0.005 In Equation (1), b was computed using the Coalesced Fractal Sphere model (Lee et al., 2000), which assumes that only binary collisions occur, and that the shape of the aggregates associated with the particle suspension. A linear increase in are defined by their fractal dimension, D , which remains þ f Na concentration with CMC-NZVI concentration is shown in unchanged during aggregation. The coalesced fractal sphere Fig. 2b. The total Na measured in the aqueous phase, after model is a valid approach for solving the Smoluchowski separating the CMC-NZVI by centrifugation at 10000 rpm for equation and used to evaluate the coagulation kinetics for þ 15 min, was approximated as the Na concentration. Ther- a wide range of colloids (Gierczycki and Al-Rashed, 2008; Ira- mogravimetric analyses showed that more than 94% of the nipour et al., 2004; Maximova and Dahl, 2006). A generally CMC was bound to the NZVI and thus only a small amount of accepted value of D is in the range of 2.0 for reaction-limited f undissociated Na-CMC could have been present in the CMC- aggregation that occurs under conditions of low ionic strength NZVI free solution. (IS) aqueous solutions and/or inter-particle steric hindrance a The values of pp reported above were relatively small, (Lin et al., 1989; Trinh et al., 2009). The aggregate sizes indicating a probability of up to 3 successful collisions out of generated with time, are defined as follows (Lee et al., 2000): a 10,000. The small values of pp are not unusual given the low = effective IS of less than 10 mM (Petosa et al., 2010) and the ¼ 1 Df dpðiÞ Ni dpði¼1Þ (2) predicted high repulsive energy barrier for particleeparticle interactions (Fig. S2a in Supplementary data). Saleh et al. where N is the number of single particles forming an aggre- i a (2010) reported pp of carbon nanotubes coated with humic gate of diameter dp(i) and dp(i¼1) is the diameter of the smallest acid to be in the range of 10 3 to 10 4 in 1 mM IS solutions. Hu particle (45 nm) present in the suspension and considered as et al. (2010) also observed particleeparticle attachment effi- the primary monomer. a ciencies of similar magnitude for humic acid-coated magne- The pp was determined by fitting the calculated particle tite nanoparticles at IS ranging from 5 to 10 mM. number-averaged mean diameter ðd Þ over time predicted Navg The particle size distributions of CMC-NZVI suspensions by Equation (1) with the number-averaged mean diameter over time were calculated using Equations (1) and (2) with the (d ) measured over time by NTA. Navg a fitted values of pp. Changes in the PSD with time as calculated The fit of Equation (1) to NTA data is shown in Fig. 2a. The 1 ¼ by Equation (1) for particle concentrations of 0.2 g L and initial (t 0) PSD was obtained from TEM image analyses of 0.725 g L 1 are presented in Fig. S3 in Supplementary data. The CMC-NZVI in a freshly-prepared suspension, which yielded results show that at higher particle concentrations, there is a dNavg of 70 nm, and a minimum diameter, dp(i¼1), of 45 nm. a greater abundance of relatively large-sized aggregates. The mean particle diameter at t ¼ 0 based on NTA was 192 nm. The difference in the mean particle sizes determined by TEM and NTA can be explained by the fact that NTA calculates 3.2. Framework for assessing effect of aggregation on particle hydrodynamic diameters based on light scattering transport of CMC-NZVI (Domingos et al., 2009). The difference in mean particle diameters determined from TEM and NTA at the initial time The effects of aggregation on transport of CMC-NZVI in point (t ¼ 0 min) was assumed to be constant over time, and packed columns were assessed by accounting for the changes the value of dNavg was adjusted accordingly. The fitting of the in PSD over the time course of the transport experiments. The 0.2 g L 1 and 0.725 g L 1 CMC-NZVI suspensions prepared in classical colloid transport theory is described in the a 4 4 0.1 mM NaHCO3 yielded a pp of 1.7 10 and 3.0 10 , Supplementary data. The particle deposition rate coefficient a a respectively. The different values of pp occurs because it was (kdep) is a function of the attachment efficiency ( ) between found that the two suspensions had different IS even though suspended particles and the collector surface and the single- h prepared with the same electrolyte solution of 0.1 mM collector contact efficiency ( 0). The single-collector contact þ NaHCO3. The effective IS accounting for the increase in Na efficiency depends on the porous medium properties (grain salt concentration for different CMC-NZVI suspensions are size, porosity), fluid properties (fluid velocity and viscosity) reported in Table 2. With an increase in CMC-NZVI concen- and the particle properties (size and density of particles) þ tration, the dissolved Na concentration released from NaBH4 (Rajagopalan and Tien, 1976; Tufenkji and Elimelech, 2004; h added during particle synthesis and from the dissociation of Yao et al., 1971). Calculations of 0 using the Tufenkji and Na-CMC (Shimabayashi et al., 1992) also increases. CMC-NZVI Elimelech Equation (Tufenkji and Elimelech, 2004) show that particles were used without washing to prevent excessive loss the value of the single-collector contact efficiency decreases of particles and to minimize oxidation of these reactive rapidly as the particle diameter increases up to 300 nm particles and thus any Na added at synthesis remained (Fig. 3a) due to the decreasing contribution of Brownian 1740 water research 46 (2012) 1735e1744

h 2 Xn diffusion. The changes in 0 are smaller for particle diameters vC v C vC ¼ D v k ; C ðtÞ (5) between 300 and 500 nm. For particles greater than 500 nm, vt vx2 vx dep i i i¼1 h the value of 0 again increases due to the contributions of interception and sedimentation to particle deposition. Thus, where kdep,i is the deposition rate coefficient, massi is the changes in the nanoparticle suspension PSD with time as mass, Ci is the mass concentration in the aqueous suspension a result of aggregation will result in time-varying changes in and ni is the number concentration of the ith sized particles. Calculations using Equations (1) and (3e5), show that kdep. The magnitude of the overall attachment efficiency for particle deposition is given by Kuhnen et al. (2000) as the shape of the breakthrough curve (BTC) was influenced a ¼ a q þ a q q by the extent of aggregation. Calculated BTCs for CMC-NZVI pcB( ) pp , where is the fraction of collector surface a coverage and B(q) is the blocking function. The deposition of suspensions with different values of pp are shown in suspended particles onto previously deposited particles can Fig. 3b. These predicted BTCs were computed assuming the a initial PSD obtained from TEM image analysis with a dNavg of be deemed negligible based on the small values of pp 1 a (¼1.7 10 4 3 10 4) reported in Table 2, and q 70 nm as discussed above, a concentration of 0.3 g L , and pc a (¼1.5 10 4 2.4 10 3). The values of q were calculated from of 0.1. This value of pc is approximately an average value the mass of particles retained in the column as described in determined from fitting experimental data obtained in this Supplementary data. Also, B(q) was assumed to be unity for study (Table 2). The results in Fig. 3b show that aggregation a very small value of q (Kuhnen et al., 2000). Thus, here, a can can lead to enhanced transport of CMC-NZVI over time. With a 4 be assumed to be equal to the particleecollector attachment a pp value of 10 , there was a 13 nm increase in average a particle size over the time period within which transport efficiency ( pc) alone. The expressions for the deposition rate ¼ ¼ coefficient of ith particle size and the resulting modified occurred (from dNavg 70 nm at t 0 min to dNavg 83 nm at ¼ h colloid transport equations are: t 60 min). This resulted in relatively small decreases in 0 and a corresponding small increase in C/C0 over time (Fig. 3b). a ¼ 3 ð 3 Þ However, with a higher degree of aggregation when pp 10 , ¼ 3 1 va h kdep;i pc 0;i (3) the d increased from 70 nm at t ¼ 0 min to 162 nm at 2dc 3 Navg ¼ h t 60 min. This resulted in a more extensive decrease in 0

and a corresponding higher increase in C/C0 over time. As Ci ¼ massi ni (4) a ¼ 3 shown in Fig. 3b, for pp 10 , the value of C/C0 increased a ¼ 4 from 0.42 at 1.5 PV to 0.6 at 8 PV. In contrast, for pp 10 ,

C/C0 increased from 0.42 at 1.5 PV to 0.48 at 8 PV.

3.3. Breakthrough curves of CMC-NZVI from transport experiments

Column transport experiments were conducted using CMC- NZVI suspensions prepared at concentrations of 0.07 g L 1, 0.2 g L 1 and 0.725 g L 1, in an electrolyte solution of 0.1 mM

NaHCO3. Representative BTCs shown in Fig. 4 indicate that at the highest inﬂuent CMC-NZVI concentration of 0.725 g L 1, = C C0 at 8 PVs is 0.61 which is lower compared to C/C0 of 0.75 for the experiment conducted at 0.07 g L 1. The BTCs indicate that even after 2 PVs, steady state concentrations were not achieved as the efﬂuent concentrations slowly but continuously increase with time. The tracer

(KNO3) BTC showed steady state effluent concentrations after 2 PVs and thus the particle effluent concentrations were not influenced by artefacts arising from column packing. The CMC-NZVI BTCs are typical of blocking of collector surfaces, where deposited particles inhibit additional particle deposition once they have extensively covered collector surfaces (Bradford and Bettahar, 2006; Brown, 2007; Johnson and Elimelech, 1995; Kuhnen et al., 2000; Song and Elimelech, 1993). However, blocking is unlikely in this system given that: (i) the surface coverage of the collector grain surfaces was comparatively small in the range of 0.015e0.24% (at 8 PV), as calculated from the mass of particles eluted based on the Fig. 3 e (a) Change in single-collector contact efficiency BTC and the mass injected, and (ii) effluent concentrations with particle size in the range 40e700 nm, and (b) change decreased with increasing CMC-NZVI influent concentrations. in CMC-NZVI breakthrough curves computed for varying Blocking generally requires more collector surface coverage by a d [ a [ pp, with Navg 70 nm, pc 0.1 and particle deposited particles than observed in our study. Ko et al. (2000) L concentration of 0.3 g L 1. observed blocking with as low as 5% surface coverage for sub- water research 46 (2012) 1735e1744 1741

a 012345678 particle concentrations (beyond 5 PVs) cannot be attributed to 1.0 1.0 aggregation or blocking as described above. The increasing efﬂuent concentrations beyond what can 0.8 0.8 be accounted for by aggregation may be attributable to detachment of deposited particles. Several studies have

0 0.6 0.6 demonstrated detachment of colloids, including nano- -1 C/C C =0.07gL 0.4 0 0.4 particles, when deposited on silica surfaces (Bradford et al., -1 C0=0.2gL 2007; Phenrat et al., 2009; Quevedo and Tufenkji, 2009; Tong -1 0.2 C0=0.725gL 0.2 and Johnson, 2006). The results of extended Derja- Tracer data Fitted Curve guineLandaueVerweyeOverbeek (XDLVO) (de Gennes, 1987; 0.0 0.0 012345678 Gregory, 1975) calculations (Fig. S2b in Supplementary data) for the interaction between a CMC-NZVI particle and a sand Pore volume (PV) surface indicate a very high energy barrier with a secondary 012345678 energy minimum (on the order of 2.7 kT) where particles b 1.0 1.0 might be retained under this unfavorable condition for deposition (Chen and Elimelech, 2008; Phenrat et al., 2009). For 0.8 0.8 particles deposited under unfavorable conditions, detach- 0.6 0.6 ment may occur when the applied torque (Tapplied) exceeds the 0 adhesive torque (Tadhesive). The Tadhesive acts on a deposited -1

C/C C =0.07gL 0.4 0 0.4 particle due to particle-collector XDLVO total interaction -1 C0=0.2gL energy, and was calculated for different-sized particles in this -1 0.2 C0=0.725gL 0.2 study according to Torkzaban et al. (2007). The Tapplied arises Fitted Curve from the shear force and hydrodynamic drag acting on 0.0 0.0 012345678 deposited particles and was calculated based on the method described by Bergendahl and Grasso (2000). Our calculations Pore volume (PV) show that even for the smallest size particle (45 nm), > Fig. 4 e Experimental breakthrough curves of CMC-NZVI Tapplied Tadhesive, and with increasing particle size, Tapplied/ with model fit considering (a) particle aggregation and Tadhesive increases (Fig. S4 in Supplementary data). Therefore transport but no detachment, and (b) Particle aggregation, detachment of deposited particles due to hydrodynamic drag transport and detachment. The BTCs were obtained using is likely to influence CMC-NZVI transport behavior. L particle suspensions of concentrations of 0.07 g L 1, To better understand the potential influence of particle L L 0.2 g L 1 and 0.725 g L 1. The column bed depth was 9 cm detachment on the shape of the particle BTC, the aggregation- L and flow velocity 0.445 cm min 1. based transport model described by Equation (5) was extended to incorporate particle detachment from the collector surface (Bradford et al., 2003) as follows:

v v2 v Xn r micron sized particles at low IS, but the surface coverage C ¼ C C ð Þþ kdetSdep D v kdep;iCi t (6) vt vx2 vx 3 required for blocking increased with decreasing particle size. i¼1 In addition, blocking results in higher efﬂuent concentrations X dSdep with increasing inﬂuent particle concentrations (Brown, 2007; r ¼ 3 ð Þ r d kdep;iCi t kdetSdep (7) Kuhnen et al., 2000). t

Breakthrough curves for CMC-NZVI were calculated using where kdet is the detachment rate coefficient, Sdep is the the modified colloid transport model (Equations (1) and (3e5)) concentration of particles deposited on the solid phase and are shown with the experimental data in Fig. 4a. For the (mg kg 1) and r is the bulk density of the porous medium. 1 a 0.07 g L CMC-NZVI suspension, pp was set to 0. Fig. 4a As shown in Equations (6) and (7), the cumulative CMC- shows that the fit of the predicted BTCs matches reasonably NZVI mass deposited increases over successive PVs, and so well with the measured effluent concentrations up to 5 PVs. does the detachment of this deposited mass (Bradford and The increase in the effluent concentrations over time is Bettahar, 2006; Phenrat et al., 2010; Torkzaban et al., 2007). greater for the BTC for the 0.725 g L 1 suspension than that for The net effect is a gradual increasing trend in effluent 1 the 0.2 g L suspension. This is because the increase in dNavg concentrations over time. Fig. 4b shows that the transport h over time, and the corresponding decrease in 0 was greater model that accounts for aggregation kinetics as well as for the 0.725 g L 1 suspension compared to the 0.2 g L 1 particle detachment can be fitted well (r2 > 0.95) to the a suspension. The particle-collector attachment efficiency ( pc) experimental data obtained from column experiments. The a was determined by fitting Equations (3)e(5) to the experi- fitted parameters in Equations (6) and (7) are kdet and pc, and mentally determined effluent concentrations, and the fitted the values are presented in Table 2. The single fitted value of 1 values are presented in Table 2. Beyond 5 PVs, the observed kdet of 0.005 min in this study is within the range gradual increase in effluent concentrations for the different (0.001e0.015 min 1) reported in other studies conducted with CMC-NZVI suspensions shown in Fig. 4a is slightly greater latex particles of submicron to micron-sized spheres than what is obtained from fitting of the modified colloid (Bradford et al., 2003; Li et al., 2005; Tong and Johnson, 2006). transport model. Hence, the observed increase in effluent Additional experimental data confirming detachment of CMC- 1742 water research 46 (2012) 1735e1744

volumes, and the non-steady state change in effluent concentration was more rapid with increasing particle concentrations. The increases in effluent concentrations beyond approximately 5 PV were however higher than what was accounted for by fitting the coupled aggregation- transport model and it is proposed that detachment of deposited particles contributed to the increasing effluent concentrations, especially at higher PVs. Particle deposition increased with higher CMC-NZVI influent concentrations and þ was attributable to increases in Na in solution from the

higher doses of Na-CMC and NaBH4 present in more concentrated particle suspensions. Such effects of CMC concentration on aqueous chemistry and transport have not been e Fig. 5 Breakthrough curves of CMC-NZVI for column reported to date. L experiments with 0.2 g L 1 particle suspensions prepared in 0.1 mM and 10 mM IS. The column bed depth was 9 cm L and ﬂow velocity 0.445 cm min 1. Acknowledgment

This research was funded by the Strategic Project Grant NZVI are however needed to demonstrate particle detach- Program of the Natural Sciences and Engineering Research ment, and will be addressed in future work. Council of Canada and by Golder Associates Ltd., as well as by a The value of pc is expected to be constant for different Le Fonds que´be´cois de la recherche sur la nature et les tech- concentrations of the CMC-NZVI suspensions. However, the nologies Team Grant program. T.R. was supported in part by a observed increase in pc with particle concentration suggests a McGill Engineering Doctoral Fellowship. enhanced deposition at higher inﬂuent concentrations and can be attributed to the effective increase in IS with increasing CMC-NZVI concentrations as explained in Section 3.1. The a Appendix Supplementary material values of pc obtained in this study range from 0.05 to 0.12 as reported in Table 2, and are in the range of values of 0.065e0.072 reported in other studies conducted at a similar IS Supplementary material associated with this article can be (Phenrat et al., 2009; Saleh et al., 2008). found, in the online version, at doi:10.1016/j.watres.2011. þ The effect of Na concentration on transport and deposi- 12.045. tion was assessed by additional column transport experiments where particle suspensions of 0.2 g L 1 CMC-NZVI were references prepared in electrolyte solutions of 0.1 mM and 10 mM

NaHCO3 and all other experimental parameters unchanged. The IS of 10 mM was chosen as it was similar to the effective IS þ of the 0.725 g L 1 suspension based on Na concentrations Bergendahl, J., Grasso, D., 2000. Prediction of colloid detachment in a model porous media: hydrodynamics. Chemical (Table 2). Fig. 5 shows that the relative effluent CMC-NZVI Engineering Science 55, 1523e1532. mass concentrations (C/C0) at 8 PVs are 0.66 and 0.40 for IS Bradford, S.A., Bettahar, M., 2006. Concentration dependent of 0.1 mM and 10 mM, respectively. Indeed, the deposition of transport of colloids in saturated porous media. Journal of CMC-NZVI increased in the presence of higher amounts of Contaminant Hydrology 82, 99e117. þ Na in the electrolyte. Bradford, S.A., Simunek, J., Bettahar, M., Van Genuchten, M.T., Yates, S.R., 2003. Modeling colloid attachment, straining, and exclusion in saturated porous media. Environmental Science and Technology 37, 2242e2250. 4. Conclusion Bradford, S.A., Torkzaban, S., Walker, S.L., 2007. Coupling of physical and chemical mechanisms of colloid straining in e The effects of particle concentration on aggregation and saturated porous media. Water Research 41 (13), 3012 3024. Brown, D.G., 2007. Adaptable method for estimation of transport of CMC-NZVI was investigated. With increases in parameters describing bacteria transport through porous particle concentration, more rapid increase in average particle media from column effluent data: optimization based on data size with time was observed because of increased parti- quality and quantity. Colloids and Surfaces A: cleeparticle attachment efficiency and number of collisions. Physicochemical and Engineering Aspects 296, 19e28. A coupled aggregationetransport model demonstrated that Chatterjee, J., Gupta, S.K., 2009. An agglomeration-based model the initial particle size distribution of a CMC-NZVI suspension for colloid filtration. Environmental Science and Technology e determines the magnitude of the single-collector contact 43 (10), 3694 3699. Chen, K.L., Elimelech, M., 2008. Interaction of fullerene (C60) efficiency as a result of aggregation and leads to a time- nanoparticles with humic acid and alginate coated silica dependent deposition rate coefficient. Column experiments surfaces: measurements, mechanisms, and environmental with CMC-NZVI at different concentrations showed a gradual implications. Environmental Science and Technology 42, increase in effluent particle concentration over eight pore 7607e7614. water research 46 (2012) 1735e1744 1743

Cirtiu, M.C., Raychoudhury, T., Ghoshal, S., Moores, A., 2011. Maximova, N., Dahl, O., 2006. Environmental implications of Systematic comparison of the size, surface characteristics aggregation phenomena: current understanding. Current and colloidal stability of zero valent iron nanoparticles pre Opinion in Colloid & Interface Science 11, 246e266. and post-grafted with common polymers. Colloids and Petosa, A.R., Jaisi, D.P., Quevedo, I.R., Elimelech, M., Tufenkji, N., Surfaces A: Physicochemical and Engineering Aspects 390, 2010. Aggregation and deposition of engineered 95e104. nanomaterials in aquatic environments: role of de Gennes, P.G., 1987. Polymers at an interface: a simplified view. physicochemical interactions. Environmental Science and Advances in Colloid and Interface Science 27, 189e209. Technology 44 (17), 6532e6549. Domingos, R.F., Baalousha, M.A., Ju-Nam, Y., Reid, M.M., Phenrat, T., Kim, H.-J., Fagerlund, F., Illangasekare, T., Tufenkji, N., Lead, J.R., Leppard, G.G., Wilkinson, K.J., 2009. Lowry, G.V., 2010. Empirical correlations to estimate Characterizing manufactured nanoparticles in the agglomerate size and deposition during injection of environment: multimethod determination of particle sizes. a polyelectrolyte-modified fe0 nanoparticle at high particle Environmental Science and Technology 43 (19), 7277e7284. concentration in saturated sand. Journal of Contaminant Fatisson, J., Ghoshal, S., Tufenkji, N., 2010. Deposition behavior of Hydrology 118 (3e4), 152e164. carboxymethylcellulose coated zero-valent iron nanoparticles Phenrat, T., Kim, H.-J., Fagerlund, F., Illangasekare, T., Tilton, R.D., onto silica: role of solution chemistry and organic Lowry, G.V., 2009. Particle size distribution, concentration, and macromolecules. Langmuir 26 (15), 12832e12840. magnetic attraction affect transport of polymer-modified fe0 Gierczycki, A., Al-Rashed, M., 2008. Aggregation and breakage of nanoparticles in sand columns. Environmental Science and solid particles in suspensions agitated in a vibrating mixer: Technology 43 (13), 5079e5085. a fractal approach. Chemical Engineering Communications Phenrat, T., Saleh, N., Sirk, K., Kim, H.-J., Tilton, R.D., Lowry, G.V., 195 (4), 427e434. 2008. Stabilization of aqueous nanoscale zerovalent iron Gregory, J., 1975. Interaction of unequal double layers at constant dispersions by anionic polyelectrolytes: adsorbed anionic charge. Journal of Colloid and Interface Science 51 (1), 45e51. polyelectrolyte layer properties and their effect on aggregation He, F., Zhang, M., Qian, T., Zhao, D., 2009. Transport of and sedimentation. Journal of Nanoparticle Research 10 (5), carboxymethyl cellulose stabilized iron nanoparticles in 795e814. porous media: column experiments and modeling. Journal of Phenrat, T., Saleh, N., Sirk, K., Tilton, R.D., Lowry, G.V., 2007. Colloid and Interface Science 334 (1), 96e102. Aggregation and sedimentation of aqueous nanoscale He, F., Zhao, D., 2007. Manipulating the size and dispersibility of zerovalent iron dispersions. Environmental Science and zerovalent iron nanoparticles by use of carboxymethyl Technology 41 (1), 284e290. cellulose stabilizers. Environmental Science and Technology Quevedo, I.R., Tufenkji, N., 2009. Influence of solution 41 (17), 6216e6221. chemistry on the deposition and detachment kinetics of Hu, J.-D., Zevi, Y., Kou, X.-M., Xiao, J., Wang, X.-J., Jin, Y., 2010. a CdTe quantum dot examined using a quartz crystal Effect of dissolved organic matter on the stability of magnetite microbalance. Environmental Science and Technology 43, nanoparticles under different pH and ionic strength 3176e3182. conditions. Science of the Total Environment 408 (16), Rajagopalan, R., Tien, C., 1976. Trajectory analysis of deep-bed 3477e3489. filtration with the sphere-in-cell porous media model. AlChE Iranipour, G.T., Garcia-Rubio, L.H., Harmon, J.P., 2004. Modeling Journal 22 (3), 523e533. and simulation of aggregation processes in colloidal systems. Raychoudhury, T., Naja, G., Ghoshal, S., 2010. Assessment of Journal of Dispersion Science and Technology 25 (4), 555e565. transport of two polyelectrolyte-stabilized zero-valent iron Johnson, P.R., Elimelech, M., 1995. Dynamics of colloid deposition nanoparticles in porous media. Journal of Contaminant in porous media. Blocking based on random sequential Hydrology 118, 143e151. adsorption. Langmuir 11, 801e812. Saleh, N., Kim, H.-J., Phenrat, T., Matyjaszewski, K., Tilton, R.D., Kanel, S., Nepal, D., Manning, B., Choi, H., 2007. Transport of Lowry, G.V., 2008. Ionic strength and composition affect the surface-modified iron nanoparticle in porous media and mobility of surface-modified fe0 nanoparticles in water- application to arsenic(III) remediation. Journal of Nanoparticle saturated sand columns. Environmental Science and Research 9 (5), 725e735. Technology 42 (9), 3349e3355. Ko, C.-H., Bhattacharjee, S., Elimelech, M., 2000. Coupled Saleh, N., Sirk, K., Liu, Y., Phenrat, T., Dufour, B., influence of colloidal and hydrodynamic interactions on the Matyjaszewski, K., Tilton, R.D., Lowry, G.V., 2007. Surface RSA dynamic blocking function for particle deposition onto modifications enhance nanoiron transport and NAPL targeting packed spherical collectors. Journal of Colloid and Interface in saturated porous media. Environmental Engineering Science 229, 554e567. Science 24 (1), 45e57. Kuhnen, F., Barmettler, K., Bhattacharjee, S., Elimelech, M., Saleh, N.B., Pfefferle, L.D., Elimelech, M., 2010. Influence of Kretzschmar, R., 2000. Transport of iron oxide colloids in biomacromolecules and humic acid on the aggregation packed quartz sand media: monolayer and multilayer kinetics of single-walled carbon nanotubes. Environmental deposition. Journal of Colloid and Interface Science 231, 10. Science and Technology 44 (7), 2412e2418. Lambe, T.W., Robert, V.W., 1991. Soil Mechanics. Wiley, ISBN Schrick, B., Hydutsky, B.W., Blough, J.L., Mallouk, T.E., 2004. 978-0-471-51192-2. Delivery vehicles for zerovalent metal nanoparticles in soil Lee, D.G., Bonner, J.S., Garton, L.S., Ernest, A.N.S., Autenrieth, R.L., and groundwater. Chemistry of Materials 16 (11), 2000. Modeling coagulation kinetics incorporating fractal 2187e2193. theories: a fractal rectilinear approach. Water Research 34 (7), Shimabayashi, S.K., Nishino, K., Nakagaki, M., 1992. Aggregation/ 1987e2000. dispersion of amorphous silica particles by simultaneous Li, X., Zhang, P., Lin, C.L., Johnson, W.P., 2005. Role of adsorption of two polymers in an aqueous phase. Colloids and hydrodynamic drag on microsphere deposition and re- Surfaces A: Physicochemical and Engineering Aspects 63, entrainment in porous media under unfavorable conditions. 121e129. Environmental Science and Technology 39, 4012e4020. Song, L., Elimelech, M., 1993. Dynamics of colloid deposition in Lin, M.Y., Lindsay, H.M., Weitz, D.A., Ball, R.C., Klein, R., porous media. Modeling the role of retained particles. Colloids Meakin, P., 1989. Universality in colloid aggregation. Nature and Surfaces A: Physicochemical and Engineering Aspects 73, 339, 360e362. 49e63. 1744 water research 46 (2012) 1735e1744

Tong, M., Johnson, W.P., 2006. Excess colloid retention in porous solutions of cellulose derivatives. Colloid and Polymer Science media as a function of colloid size, fluid velocity, and grain 287 (12), 1391e1404. angularity. Environmental Science and Technology 40, Tufenkji, N., Elimelech, M., 2004. Correlation equation for 7725e7731. predicting single-collector efficiency in physicochemical Torkzaban, S., Bradford, S.A., Walker, S.L., 2007. Resolving the filtration in saturated porous media. Environmental Science coupled effects of hydrodynamics and DLVO forces on colloid and Technology 38 (2), 529e536. attachment in porous media. Langmuir 23 (19), 9652e9660. Yao, K.M., Habibian, M.T., O’Melia, C.R., 1971. Water and waste Tratnyek, P.G., Johnson, R.L., 2006. Nanotechnologies for water filtration: concepts and applications. Environmental environmental cleanup. Nano Today 1, 44e48. Science and Technology 5 (11), 1105e1112. Trinh, L.T.T., Kjniksen, A.-L., Zhu, K., Knudsen, K.D., Volden, S., Zhang, W.-X., 2003. Nanoscale iron particles for environmental Glomm, W.R., Nystrom, B., 2009. Slow salt-induced remediation: an overview. Journal of Nanoparticle Research 5, aggregation of citrate-covered silver particles in aqueous 323e333.