TE

JP

h

�

�

�

�

a

ARRAA

KETRSCF

g

(

0h

Colloids and Surfaces A: Physicochem. Eng. Aspects 417 (2013) 89– 98

Contents lists available at SciVerse ScienceDirect

Colloids and Surfaces A: Physicochemical andEngineering Aspects

jo ur nal homep a ge: www.elsev ier .com/ locate /co lsur fa

ransport and retention of engineered nanoporous particles in porous media:ffects of concentration and flow dynamics

ianying Shang ∗, Chongxuan Liu ∗∗, Zheming Wangacific Northwest National Laboratory, K8-96, Richland, WA, 99352, USA

i g h l i g h t s

Dynamic flow condition has a strongeffect on the transport of engineerednanoporous particles in porousmedia.Flow interruption provides a goodway to elucidate the transport mech-anism of nano- or colloidal particles.The transport of engineerednanoporous particles is consistentwith the transport of non-porouscolloidal particles.Increasing input concentrationenhanced relative engineerednanoporous particle transport insaturated porous media.

g r a p h i c a l a b s t r a c t

r t i c l e i n f o

rticle history:eceived 29 August 2012eceived in revised form 13 October 2012ccepted 18 October 2012vailable online 9 November 2012

eywords:ngineered nanoporous silicate particlesransport

a b s t r a c t

Engineered nanoporous particles are an important class of nano-structured materials that can befunctionalized in their internal surfaces for various applications including groundwater contaminantsequestration. This paper reported a study of transport and retention of engineered nanoporous silicateparticles (ENSPs) that are designed for treatment and remediation of contaminants such as uraniumin groundwater and sediments. The transport and retention of ENSPs were investigated under vari-able particle concentrations and dynamic flow conditions in a synthetic groundwater that mimics fieldgroundwater chemical composition. The dynamic flow condition was achieved using a flow-interruption(stop-flow) approach with variable stop-flow durations to explore particle retention and release kinetics

etentiontop flowoncentrationlow dynamics

in saturated porous media. The results showed that the ENSPs transport was strongly affected by the par-ticle concentrations and dynamic flows. The experimental data were used to evaluate the applicabilityof various kinetic models that were developed for colloidal particle retention and release in describ-ing ENSPs transport. Both experimental and modeling results indicated that dynamic groundwater flowcondition is an important parameter to be considered in exploring and modeling engineered particletransport in subsurface porous media.

Abbreviations: ENSPs, engineered nanoporous silicate particles; SGW, syntheticround water; SF, stop flow; FI, flow interruption.∗ Corresponding author. Tel.: +1 509 371 7368; fax: +1 509 371 6354.

∗∗ Corresponding author. Tel.: +1 509 3716350; fax: +1 509 371 6354.E-mail addresses: Jianying.Shang@pnnl.gov (J. Shang), Chongxuan.Liu@pnnl.gov

C. Liu).

927-7757/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.colsurfa.2012.10.030

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

Engineered nanoporous silicate particles (ENSPs) that have alarge surface area, ordered nano-size pore structure, and con-

trolled particle size have been developed for various applicationsin biomolecular sensing and labeling, energy storage and fuel celltechnology, environmental remediation and protection [1]. A func-tionalized ENSPs material with its internal pore surfaces covalently

90 J. Shang et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 417 (2013) 89– 98

erical

bbctgs4tocimhme

mmsntigrlib(pitthagnfltsfputu

E

Fig. 1. Transmission electron micrographs showing sph

onded with salicylamide ligands has been synthesized and haseen demonstrated as a promising material to sequestrate divalentontaminants such as uranium in groundwater [2]. The distribu-ion coefficient (Kd) for uranium between the ENSPs material androundwater reaches as high as 105 ml/g as a result of high internalurface area (∼700 m2/g) and strong ligand complexation. This is

orders of magnitude higher than the Kd value for uranyl adsorp-ion to sediments [3]. The strong uptake capacity and colloidal sizef the material makes it an ideal candidate for building an in situhemical reactive barrier in an aquifer to sequestrate contaminantsn groundwater [4]. The fate and transport of the material in porous

edia that will potentially affect its delivery and stability, however,ave not been well studied, preventing the applicability assess-ent of the material in remediating contaminants in subsurface

nvironments.The colloidal size of the ENSPs suggests that their transport

ay be controlled by two major processes that control com-on colloidal particle transport: attachment–detachment and

training–liberation. Attachment–detachment process reflects theet effect of the adhesive and hydrodynamic forces that act athe liquid–solid interfaces [5–8]. The straining–liberation processs a phenomenon that is still not well understood. This processenerally describes the phenomenon of particle trapping to andelease from narrow pore throats [9–11], or solid–water interfacesike grain–grain contacts or wedging [12–15]. Various factors thatnfluence the transport and retention of colloidal particles haveeen studied including the properties of particles and collectorsshape, surface chemistry, size, and geometry), solution chemistry,article concentration, and flow velocity [16–21]. All these stud-

es, however, were performed under a steady-state flow conditiono facilitate experimentation and data interpretation. Groundwa-er flow is seldom steady-state. Such unsteady flow conditionsave been observed frequently at many contaminated sites suchs at the US Department of Energy (DOE) Hanford site, whereroundwater flow velocity changes hourly as it is influenced byearby river water stage changes [22]. The influence of dynamicow on particle transport and retention is an under-studied areahat may significantly affect particle attachment/detachment andtraining/liberation processes. Limited studies using the sedimentsrom the Hanford site have shown that transient flows can enhancearticle mobilization under unsaturated conditions [23–25]. It isnclear, however, whether the models that describe colloidal par-

icle transport under steady-state flow condition will be applicablender dynamic flow conditions.

In this study, we investigated the transport and retention ofNSPs under dynamic flow conditions in a chemical solution that

morphology and intragrain porous structure of ENSPs.

mimics the groundwater chemical compositions at the US DOEHanford site where groundwater contamination is a major envi-ronmental concern. The major objectives of this study are: (1) tomeasure the transport and retention of ENSPs as a potential reme-diation agent in porous media; and (2) to evaluate whether currentmodels of colloidal particle retention/release can be applied todescribe ENSPs transport under dynamic flow conditions. The effectof particle concentration was also evaluated as it affects the overallcapacity of ENSPs in sequestrating contaminants from groundwa-ter, and may also affect the particle retention/release properties.

2. Materials and methods

2.1. Synthesis and characterization of nanoporous particles

The Bein’s method was used to synthesize colloidal size ENSPswith nanosized internal pores [26] and the Fryxell’s method wasused to functionalize surface areas [27]. A monolayer of salicy-lamide ligands was used as a functional agent, which was covalentlycoated on the internal pore surfaces in the ENSPs to functionalizethe particles for capturing contaminants such as uranyl from aque-ous solutions (Kd ∼ 105 ml/g) and produce fluorescence signals forin situ particle tracking [2].

The particle morphology was examined using transmission elec-tron microscopy (TEM, JEOL JEM-2010). TEM analysis revealed thespherical shape of the ENSPs with a comb-type pore structure intheir intra-grain region (Fig. 1). The hydrodynamic particle sizeand electrophoretic mobility of the ENSPs were measured throughdynamic light scattering (DLS) analysis using a Zetasizer 3000HAS(Malvern Instruments Ltd., Malvern, UK). DLS analysis found thatthe hydrodynamic particle size of the ENSPs ranged from 200 to450 nm and followed a normal probability distribution with a meanof 308 nm and a standard deviation of 15 nm [2]. The internalpore size and volume were estimated by the BJH method [28,29]from Brunauer–Emmett–Teller (BET) N2 adsorption and desorptionmeasurements [30], and the internal pores in the material have anominal pore diameter of 5 nm.

2.2. Porous media

Quartz sand (Accusand 40/60, Unimin Corporation, Le Sueur,MN) with a particle size ranging from 300 to 355 �m (average

330 �m) was used for constructing the porous medium (or collec-tor). The sand was pretreated sequentially with 0.25 M NaOH and0.25 M HCl to remove impurity, and then rinsed with deionizedwater until the rinsed solution pH was equal to that of deionized

Physi

wf

2

scatsTsS

2

cflsraNtcsC

0itncrE2(esuAip4gviiiwtSaw(t

tmYcS1im

J. Shang et al. / Colloids and Surfaces A:

ater. The treated sand was dried in the air and immediately usedor surface characterization and transport experiments.

.3. Solution chemistry

Synthetic ground water (SGW) was used as a backgroundolution in the experiments. The SGW has the same chemicalomposition as the groundwater at the DOE Hanford site (pH 8.2nd ionic strength 6.3 mM) [31]. The electrophoretic mobility ofhe ENSPs and treated sand was measured in the SGW suspen-ions to calculate the zeta potentials of ENSPs and quartz sand.he sand electrophoretic mobility was measured by grinding theand into colloidal powers and suspending the powders in theGW.

.4. Transport experiment

ENSPs transport experiments were conducted in plasticolumns (2.5 cm inner diameter, 15 cm length) under saturatedow condition. The column was equipped with a stainless steelcreen with pore opening of 50 �m at the bottom and top toetain sands. The sand was wet-packed into the columns with anverage bulk solid density of 1.8 g/cm3, and a porosity of 0.32.itrate (1.5 mM), by spiking the SGW with Ca(NO3)2, was used as a

racer to probe the hydrodynamic dispersion in the column. Nitrateoncentrations in the effluents were measured using a UV–vispectrophotometry (UV-2501 PC, Shimadzu Scientific Instruments,olumbia, MD) at a wavelength of 235 nm.

The ENSPs suspensions with three particle concentrations of.01, 0.025, and 0.05 mg/ml were prepared by suspending ENSPs

n the SGW solution. These particle concentrations correspondo 6.7 × 108, 1.7 × 109, and 3.4 × 109 particles/ml. For the conve-ience of description, the columns injected with these particleoncentrations were termed as 0.2Cr, 0.5Cr, and 1Cr column,espectively, where Cr = 3.4 × 109 particles/ml. The dispersion ofNSPs in the SGW was periodically monitored using NanoSightLM0 (NanoSight Ltd., Wiltshire, UK) and dynamic light scatteringDLS) method, and was found stable throughout the transportxperiments. Before injecting ENSPs solution, the ENSPs-free SGWolution was first introduced into the column for at least 5 pore vol-mes (PVs) to chemically equilibrate solids with the SGW solution.

total of 14 pore volumes of ENSPs–SGW were injected upwardsnto the column at a Darcy velocity of 4.9 cm/h using a peristalticump. The flow was interrupted (stopped) for 1, 2, and 3 h after.8, 7.8, and 10.8 pore volumes of ENSPs injection, respectively, toenerate dynamic hydraulic condition in the column. After 14 poreolumes of ENSPs–SGW injection, ENSPs-free SGW was injectednto the column to evaluate particle release kinetics. The flow wasnterrupted twice with two stop-flow durations of 1 and 7 days dur-ng the particle release phase. Extensive preliminary experiments

ere performed to determine the stop-flow (SF) time durations sohat the changes in effluent ENSPs concentrations before and afterF events were large enough for comparison and detection. Twodditional column experiments (ENSPs concentration 0.05 mg/ml)ithout flow interruption (1Cr-WOSF) and with 12 h stop-flow

1Cr-WISF) were also performed under the same conditions to fur-her test transport parameters and models.

The effluent was collected every 10 min by a fraction collec-or. The particle concentration in the effluent was quantified by

easuring fluorescence strength using a Fluorimeter (Horiba Jobinvon Inc., Edison, NJ) at wavelength 425 nm for 0.2Cr and 0.5Cr

olumns and a UV−vis Spectrophotometer (UV-2501 PC, Shimadzu

cientific Instruments, Columbia, MD) at wavelength 301 nm forCr, 1Cr-WOSF, and 1Cr-WISF columns. Two instruments were used

n measuring ENSPs concentrations to take advantage of their opti-al measurement ranges. The detection limit was 0.0002 mg/ml

cochem. Eng. Aspects 417 (2013) 89– 98 91

for the Fluorimeter and 0.002 mg/ml for the UV–vis spectropho-tometer. The detection limits were calculated as three times thestandard deviation of ten replicate samples at concentrations nearthe detection limit.

At the end of each experiment, the columns were dissected into1–1.5 cm increments to measure the spatial distribution of resid-ual ENSPs. The increment sand samples were transferred into 50 mlSGW-containing bottles and the solid suspensions were shaken for4 h to dissociate the attached ENSPs from sands. The ENSPs con-centrations in the supernatants were measured, and the massesof the dry sands were weighted. The measured residual ENSPsconcentrations were normalized to the dry mass of the sands. Inde-pendent experiments found that over 95% of the residual ENSPs canbe extracted by the above procedure. The high recovery of ENSPsimplied that the attachment between ENSPs and sands was underunfavorable conditions.

2.5. Surface coverage rate

Based on the assumption of single layer ENSP coverage on sands,fractional surface coverage was calculated as a function of time toillustrate the rate at which sand surfaces were covered by depositedparticles. The surface coverage � was calculated using the particlebreakthrough curves in the injection phase and using the followingequation by assuming ENSPs and sand particles were spherical [32]:

� =�Urcr2pC0

∫ t0

(1 − C

C0

)dt

3l(1 − ε)(1)

where rc is the averaged sand radius (L), rp is the ENSPs particleradius (L), U is the fluid superficial velocity (LT−1), C0 (ML−3) isthe ENSPs concentration in influent solution, C (ML−3) is the ENSPsconcentration in effluent solution at time t, l is the column length(L), and ε is the porosity (–).

2.6. Modeling theory

The following advection and dispersion equation was used todescribe the effluent ENSPs concentrations:

∂C

∂t= ∂

∂x

(D∂C

∂x

)− v∂C

∂x− R (2)

where D is the dispersion coefficient (L2T−1), v is the pore watervelocity (LT−1), and R is the particle retention/release part asdescribed by the following. The transport equation (Eq. (2)) wassolved using a finite-difference scheme with sequential iterationbetween transport part (advection and dispersion) and parti-cle retention/release part (R). The R term in Eq. (2) is the sum(R = Ratt + Rstr) of the colloid mass transfer terms between the aque-ous and solid phases due to colloid attachment/detachment andstraining/liberation. The dispersion coefficient D was obtained fromtracer effluent data with the Hydrus-1D software [33].

Both models for colloidal attachment/detachment and strain-ing/liberation processes and their combinations were evaluatedas the R term in Eq. (2) to describe ENSPs retention/release. Theattachment/detachment process Ratt is typically described using afirst-order kinetics expression [20,34]:

�Ratt = �∂Satt

∂t= �katt attC − �kdetSatt (3)

where Satt is the attached particle concentration (MM−1), � is thebulk solid density (ML−3), katt and kdet are the first-order parti-cle attachment and detachment coefficients (T−1), respectively, and

9 Physi

a

w(w

�

wsad

wftfd(d

w(

2

etcoamp(bdpcd

ceatcn

fiEdmddptim

2 J. Shang et al. / Colloids and Surfaces A:

att is the dimensionless parameter to account for the decrease inttachment site with increasing number of retained particles [35]:

att = 1 − Satt

Smaxatt

(4)

here Smaxatt is the maximum concentration of attached particles

MM−1). The straining/liberation process Rstr has been describedith the following expression [20]:

Rstr = �∂Sstr

∂t= �kstr strC − �klibSstr (5)

here Sstr is the strained particle concentration (MM−1), kstr is thetraining coefficient (T−1), klib is the liberation coefficient (T−1),nd str is the dimensionless parameter to account for the depth-ependent straining phenomenon [20,34]:

str =(d + z

d

)−ˇ(6)

here d is the diameter of the sand grains (L), z is the distancerom the column inlet (L), and ̌ is a parameter (–) that con-rols the shape of particle distribution. Empirical fitting, however,ound that the following straining function str that couples depth-ependent straining and Langmuirian dynamic blocking functioncalled Langmuirian depth-dependent blocking function) betterescribes concentration-dependent transport [20]:

str =(

1 − Sstr

Smaxstr

)(d + z

d

)−ˇ(7)

here Smaxstr is the maximum concentration of strained particles

MM−1).

.7. Modeling implementation

Five models of different combinations of Eqs. (3)–(7) werevaluated as the R term in Eq. (2) to describe ENSPs concen-rations in effluents and residual ENSPs concentrations in theolumn. Model 1 (M1) was described by Eq. (3) with a conditionf att = 1 that has traditionally been used to describe particlettachment–detachment processes in clean bed porous media;odel 2 (M2) consists of Eqs. (3) and (4) by assuming that attached

articles would affect subsequent particle attachment; model 3M3) assumes that particle retention/release can be describedy straining–liberation processes as described by Eq. (5) with aistance-dependent straining function (Eq. (6)); model 4 (M4) cou-les models 1 and 3 (Eqs. (3), (5), and (6)); and model 5 (M5) alsoouples models 1 and 3, but assumes that the straining functionepends on both distance and site blocking (Eqs. (3), (5) and (7)).

Five models were implemented and developed using Fortranode based on Eqs. (2)–(7). During the stop-flow events in five mod-ls, pore water velocity v in Eq. (2) was zero, and other parametersnd calculating equations were the same as the values and equa-ions used under flow conditions. The models were validated byomparing the tracer results with commonly used analytical andumerical models.

The parameters in the particle retention/release models wererst estimated from the ENSPs breakthrough curve and residualNSPs data from the 0.2Cr column. The goodness of the fit usingifferent particle retention/release models was compared to deter-ine the ability of these models to describe particle transport under

ynamic flow condition. Model 5 was found to best describe theata as shown in the next section. Consequently model 5 with fitted

arameters from the 0.2Cr column was used to predict effluent par-icle concentrations and residual ENSPs in the columns with othernjected particle concentrations (i.e. 0.5Cr, and 1Cr) to evaluate the

odel predictability. Finally, model 5 was refitted to all column

cochem. Eng. Aspects 417 (2013) 89– 98

effluents to derive a column-specific set of parameters for compari-son and evaluation of experimental conditions. The match betweenthe models and experimental data was used as a criterion to eval-uate the applicability of these colloidal models in describing ENSPstransport and retention.

3. Results and discussion

3.1. DLVO interactions

The zeta potentials of ENSPs and quartz sands were −19.0and −27.5 mV, respectively, which were calculated from mea-sured electrophoretic mobilities in the SGW solution. The negativezeta potentials of both ENSPs and quartz sand indicated thatENSPs attachment onto the sand was thermodynamically unfa-vorable in the SGW solution. The results were consistent with theamphoteric electrostatic properties of silica and quartz surfaces atcircum-neutral pH in the SGW solution. The interaction energiescalculated from the Derjaguin–Landau–Verwey–Overbeek (DLVO)theory were 100 and −0.17kT for primary energy barrier (at 1 nm)and secondary energy minimum (at 30.5 nm), respectively [36].The calculations used the zeta potentials for surface potentials,the Hamaker constant of 4.67 × 10−21 J for silica–water–sand sys-tem, and a characteristic length of 100 nm [24,36]. These calculatedenergy values suggest that ENSPs are difficult to release once theyare trapped at the primary energy minimum, and on the otherhand, the particles at the secondary energy minimum can be readilydetached from the sand collector surfaces [37,38].

3.2. Transport and retention of ENSPs

Flow condition and ENSPs concentration significantly affectedENSPs transport (Fig. 2a), and residual ENSPs distributions (Fig. 2b).The effluent ENSPs concentrations became detectable at around0.95 pore volume (PV) in all columns, then rapidly increased to anintermediate concentration ranging from 0.3 to 0.7C/C0 depend-ent on inflow particle concentrations at the 2nd PV, followed by aslow increase toward a plateau concentration at the 3rd or 4th PV(Fig. 2a). The SF events at 4.8, 7.8, and 10.8 PVs resulted in a sharpdecrease in effluent ENSPs concentration once flow was re-started,and then followed by a rapid rebounding to a plateau concentra-tion within one pore volume. The SF-induced decrease in effluentENSPs concentration indicated that the particles deposited ontothe sand collector surfaces during the stop flow durations and thedeposition process was kinetically controlled despite the electro-static repulsion between ENSPs and collector surfaces as indicatedby their negative zeta potentials in the SGW. This decrease was notobserved for tracer Br curves (Fig. S1 of Supporting Information).There was no delay in tracer effluent concentration curves fromthe column after the SF events. No response of effluent Br to the SFevents reflected no or low storage capacity for the tracer Br in theintragrain domains.

After 13.8 PV of ENSPs injection, the influent solution waschanged to ENSPs-free SGW to detach particles from the columns.The effluent ENSPs concentration maintained at a plateau for about1 PV, followed by a quick decrease within 0.3 PV to less than 0.1C0(Fig. 2a), where C0 is the ENSPs concentration in the injected solu-tion during the injection phase. All ENSPs breakthrough curvesshowed a long tailing (Fig. 2a), suggesting that ENSPs release wasalso controlled by rate-limited processes [31,39]. The tail ENSPsconcentration was higher for the column injected with a higher

ENSPs concentration, consistent with the concentration-dependentparticle detachment (Eq. (3)) and/or liberation (Eq. (5)) kinetics.

The SF events with durations of 1 and 7 days led to a grad-ual rebounding in effluent ENSPs concentration once the flow

J. Shang et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 417 (2013) 89– 98 93

Pore V olume

0 5 10 15 20 25

Rela

tive

Co

ncen

trati

on

(C

/C0)

0.0

0.2

0.4

0.6

0.8

1.0

1.20.2 Cr

0.5 Cr

1Cr

1 h 2 h 3 hSGW flus hing

1 da y 7 da ys

(a)

N Ni-1

0 5 10 15 20

Dis

tan

ce f

rom

co

lum

n in

let

(cm

)

0

2

4

6

8

10

12

14

0.2 Cr

0.5 Cr

1Cr

(b)

F s transe lume,

i

rtfrropcSscmccfsih

efcihtciwpsEhcsotwc

3

Etac

cating that attachment/detachment process alone was not enoughto describe ENSPs transport and suggesting that straining pro-cess had to be included in the ENSPs transport model. Model3, which was designed to describe straining process, well fitted

Fra

cti

on

al S

urf

ace C

overa

ge,θ

0.000

0.002

0.004

0.006

0.008

0.0100.2 Cr

0.5 Cr

1 Cr

1h SF

2h SF

3h SF

ig. 2. Measured ENSPs breakthrough curves (a) and retention profiles (b) for ENSPvents. C0 is injected particle concentration, N is residual ENSP mass in one pore von the injected solution.

e-started (Fig. 2a). The rebounding lasted for one PV and thenhe effluent ENSPs concentration decreased as the injected ENSPs-ree SGW continuously removed particles from the columns. Theebounding after SF events confirmed that the release of theetained particles was kinetically controlled. The rebounding peakf ENSPs concentration occurred at the end of the first re-injectedore volume (Fig. 2a), indicating that the attached particles wereontinuously mobilized along flow paths within the first PV afterF events. This result was in contrast to the kinetic desorption ofolutes and contaminants in porous media where effluent con-entrations of desorbing solutes and contaminants reached theaximum immediately after SF events, followed by a monotoni-

al decrease with time [31,40]. While a specific mechanism for thisontrast is still unknown, this phenomenon apparently resultedrom the fact that chemical sorption/desorption is controlled byurface processes that are intrinsically not affected by flow dynam-cs, while particle attachment/detachment is directly affected byydrodynamic force [21,41–43].

The spatial distributions of residual ENSPs at the end of thexperiments were presented in Fig. 2b as a function of distancerom the column inlet. Generally the relative residual ENSPs con-entration was higher with a lower injected ENSPs concentration,ndicating that the relative retention capacity of the sand wasigher for the lower injected particle concentrations. The spa-ial distribution profiles of the residual ENSPs were similar in allolumns, showing higher ENSPs concentrations near the columnnlet and lower concentrations toward the outlet. This phenomenon

as apparently consistent with the colloidal straining process thatreferentially retained particles in the column inlet [20,33,44]. Thetraining mechanism was also supported by the relative residualNSPs concentration near the 0.2Cr column inlet (Fig. 2b) that wasigher than those in other columns with higher injected ENSPs con-entrations. This is because for a system with a fixed number oftraining sites, lower injected ENSPs allowed a higher percentagef particles to be strained from the solution [21]. This also par-ially explained the higher relative breakthrough concentrationsith a higher injected ENSPs concentration in 0.2Cr, 0.5Cr, and 1Cr

olumns (Fig. 2b).

.3. Surface coverage rate

The rate of particle surface coverage increased with increasing

NSPs concentration in the influent solution (Fig. 3). This is qualita-ively consistent with the numerical models for particle attachments Eqs. (3) and (5) indicated that increasing aqueous ENSPs con-entration would increase the rate of ENSPs particle retention.

port experiments with various concentrations (0.2Cr, 0.5Cr, and 1Cr) and stop-flowand Ni is ENSPs mass in one pore volume containing aqueous ENSPs concentration

Quantitatively, however, the ENSPs coverage rate was not con-sistent with the models because the coverage rate nonlinearlyincreased with increasing ENSPs concentration in the injectionsolution, while the models implied otherwise. This nonlinearityfrom 0.2Cr to 1Cr column was likely caused by site blocking effect[20,21]. Blocking effect describes that particle retention decreaseswith time as favorable attachment sites are filled by retained parti-cles [19–21,32,45]. The small jumps in the surface coverage beforeand after SF events resulted from particle deposition/attachmentduring SF durations, which decreased effluent ENSPs concentrationonce flow restarted. Generally, however, the effect of SF events tothe accumulated ENSPs particle concentration was small becauseof the small amount of ENSPs mass in aqueous solutions relative tothe ENSPs attached to solid phase.

3.4. Modeling ENSPs transport and retention

Five particle retention/release models (M1 to M5) had variabledegrees of success in describing the ENSPs transport in 0.2Cr col-umn (Fig. 4) with adjustable parameters, their fitted values, andestimated standard errors of the fitted parameters provided inTable 1. Models M1 and M2 poorly described effluent results atearly time (Fig. 4a) and residual ENSPs concentrations (Fig. 4b), indi-

Pore Volu me

0 2 4 6 8 10 12 14 16

Fig. 3. The temporal evolution of averaged fractional surface coverage of ENSPs ontosand grains in different columns.

94 J. Shang et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 417 (2013) 89– 98

Table 1Column properties (porosity and bulk density) and the parameters used in modeling ENSPs transport.

Model Column Porosity Bulk density (g cm−3) katt (h−1) kdet (h−1) kstr (h−1) klib (h−1) Smaxatt (�g g−1) Smax

str (�g g−1) R2e R2

s

1 0.2Cr 0.322 1.798 0.6(0.017) 0.01(0.002) N/A N/A N/A N/A 0.86 0.612 0.2Cr 0.322 1.798 0.9(0.035) 0.01(0.002) N/A N/A 29.6(3.1) N/A 0.86 0.573 0.2Cr 0.322 1.798 N/A N/A 4.5(0.042) 0.01(0.002) N/A N/A 0.89 0.874 0.2Cr 0.322 1.798 0.4(0.09) 1.0(0.4) 4.7(0.13) 0.01(0.001) N/A N/A 0.92 0.875 0.2Cr 0.322 1.798 0.4(0.1) 1.0(0.4) 5.7(0.24) 0.01(0.001) N/A 49.3(0.18) 0.93 0.865 0.5Cr 0.327 1.783 0.4 1.0 2.5(0.09) 0.01 N/A 49.3 0.97 0.965 1Cr 0.327 1.783 0.4 1.0 2.3(0.16) 0.01 N/A 49.3 0.95 0.565 1Cr-WOSF 0.327 1.783 0.4 1.0 2.3 0.01 N/A 49.3 0.95 N/A5 1C -WISF 0.327 1.783 0.4 1.0 2.3 0.01 N/A 49.3 0.91 N/A

N cient

f ter fitt

tttEmbEewTcp

Fcp

r

/A denotes that the parameters were not used in the models. R2e denotes the coeffi

or spatial distribution data. Bold font denotes the fixed parameters during parame

he residual ENSPs concentration (Fig. 4b), but poorly describedhe front of the effluent breakthrough curve (Fig. 4a), indicatinghat the straining process only was not enough to describe theNSPs transport either. Model M4, which includes both attach-ent/detachment and straining/liberation processes, provided a

etter description of ENSPs breakthrough curves and residualNSPs concentration. Model M5, which is the same as model M4xcept that the Langmuirian depth-dependent blocking functionas used in the straining model, further improved the fitting.

hese results suggested that the transport of ENSPs was affected byoupled colloidal attachment/detachment and straining/librationrocesses. Considering that model M5 provided the best

Pore Volume

0 5 10 15 20 25 30

Rela

tive C

on

cen

trati

on

(C

/C0)

0.0

0.2

0.4

0.6

0.8

1.0

0.2 Cr

Model 1

Model 2

Model 3

Model 4

Model 5

1h 2h 3h

SGW flushi ng

1day 7days

0 10 20 30 40 50 60

Dis

tan

ce f

rom

co

lum

n in

let

(cm

) 0

2

4

6

8

10

12

14

0.2 Cr

Model 1

Model 2

Model 3

Model 4

Model 5

(a)

(b)

g/g sandμ

ig. 4. Measured and simulated breakthrough curves (a) and residual ENSPs con-entration profiles (b) in column 0.2Cr. Five models were used in simulations witharameters provided in Table 1.

of linear regression for effluent data. R2s denotes the coefficient of linear regression

ing.

description of the experimental data, in the subsequent modeling,model M5 was further examined in other columns.

The breakthrough curves and residual ENSPs concentrationssimulated using Model M5 with parameters estimated from the0.2Cr column were qualitatively consistent with, but quantitativelydeviated from the experimental results in 0.5Cr and 1Cr columns(Fig. 5). The predicted results overestimated the residual ENSPsconcentrations and underestimated breakthrough concentrations.Sensitivity analysis found that the match between model resultsand experimental data can be significantly improved by adjustingkstr value (Table 1) for individual columns (solid-line, Fig. 5a, b,c, and d). The decrease in the best fitted kstr value with increasinginjected ENSPs concentration from 0.2Cr to 1Cr (Table 1) apparentlyresulted from the electrostatic repulsion effect of strained parti-cles. Increasing injected ENSPs concentration, which led to a higherENSPs surface coverage (Fig. 3), would increase the repulsion forcefor subsequent particle retention/attachment [17,19,20,46,47]. TheDLVO force between two particles (Fig. S2 of Supporting Informa-tion) also showed a strong repulsive force ranging from 10−12 to10−9 N when particle and particle separation distance was from0.3 to 21 nm.

According to the simulation results of model M5, kstr decreasedfrom 5.7 to 2.3 with increasing input concentration from 0.2Cr

to 1Cr. This implied that the value of the sticking efficiencycaused by straining is decreased with injected ENSPs concentrationincreasing. Similar results were observed [20,21], and this phe-nomenon was attributed to blocking effect [19–21]. Blocking effect,here, describes particle sticking efficiency decreases since avail-able favorable straining sites become less with input concentrationincreasing. The good match between residual ENSPs concentrationprofile and simulation (Fig. 5c and d) further supported that theblocking effect plays a major role in ENSPs transport when increas-ing injected ENSPs concentration.

To further confirm our assumption and the goodness of Model 5,the breakthrough curves of ENSPs concentration 0.05 mg/ml with-out stop-flow and with 12 h flow interruption were predicted usingModel M5 with parameters estimated from the 1Cr column (Fig. 6).The experimental data and predictions were matched well. Thegoodness of fitting showed that the same models can be used todescribe the particle retention/release during flow and SF dura-tions if there is no change of chemical and physical conditionsexcept flow-interruption. Although the mechanism of ENSPs reten-tion during stop-flow events is still unclear, our results showed themodel can generally guess total ENSPs retention during stop-flow,and this can significantly simplify simulating particle transportunder flow-interruption-flow conditions.

3.5. Stop-flow effect

Stop-flow events significantly affected particle transport asrevealed by the response of effluent particle concentrations to the

J. Shang et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 417 (2013) 89– 98 95

Pore Volu me

0 5 10 15 20 25

Rela

tive

Co

ncen

trati

on

(C

/C0)

0.0

0.2

0.4

0.6

0.8

1.0

0.5 Cr

Fitted Kstr

Predicted

1hSF 2hSF 3hSF SGW flushing

1 day7 da ys

μg/ g sa nd

μg/ g sa nd

0 10 20 30 40 50

Dis

tan

ce f

rom

co

lum

n in

let

(cm

) 0

2

4

6

8

10

12

14

0.5 Cr

Fitted Kstr

Predicted

(a) (c)

Pore Volu me

0 5 10 15 20 25

Rela

tive

Co

ncen

trati

on

(C

/C0)

0.0

0.2

0.4

0.6

0.8

1.0

1Cr

Fitted Kstr

Predicted

1hSF 2hSF 3hSF SGW f lushing

1 day7 da ys

0 10 20 30 40 50 60

Dis

tan

ce f

rom

co

lum

n in

let

(cm

)

0

2

4

6

8

10

12

14

1Cr

Fitted Kstr

Predicted

(b) (d)

Fig. 5. Measured and simulated breakthrough curves and residual ENSPs concentration profiles in columns 0.5Cr (plots a and c), and 1Cr (plots b and d). Dash lines are them es arew

SiE(pacbo

Ft

odel predictions using the parameters estimated from column 0.2Cr and solid linith parameters provided in Table 1.

F events (Fig. 2). The calculated amount of deposited ENSPs dur-ng the SF events increased with increasing SF duration in theNSPs injection phase regardless of injected ENSPs concentrationsFig. 7a). The calculations assumed that the amount of depositedarticles during the SF event was proportional to the difference inqueous ENSPs concentration before and after a SF event within the

olumn and that aqueous ENSPs concentration in the column cane approximated by effluent concentration. The increasing amountf ENSPs deposition with increasing SF duration further indicated

Pore Volume

0 2 4 6 8 10 120.0

0.2

0.4

0.6

0.8

1.01Cr -WO SF

Model5

predicted

(a)

Rela

tive

Co

ncen

trati

on

(C

/C0)

ig. 6. Measured and predicted ENSPs breakthrough curves in columns 1Cr without stop-he model predictions using the parameters estimated from Model 5 with column 1Cr.

the fitted results. Both predictions and fitted results were derived using model 5

that the deposition of the ENSPs from aqueous solution during SFevents was a kinetic process.

The kinetic deposition during the SF events can be attributedto various factors including sedimentation driven by gravitationalforce, surface retention/release driven by DLVO force, and Brown-ian motion (diffusion) that brought particles to collector surfaces.

According to Shang et al. [38], the gravitational force was esti-mated to be 2.01 × 10−16 N and buoyancy force to be 1.5 × 10−16 Nbased on the particle density and hydrophilic property.

Pore Vol ume

0 5 10 15 20 25

Rela

tive

Co

ncen

trati

on

(C

/C0)

0.0

0.2

0.4

0.6

0.8

1.0

1Cr -WISF

M5 p redicted

12h SF SGW flushing

12 h

(b)

flow (1Cr -WOSF, plot a) and with 12 h stop-flow (1Cr -WISF, plot b). Solid lines are

96 J. Shang et al. / Colloids and Surfaces A: Physi

Interruptio n time ( h)

0 1 2 3 4

Reta

ined

EN

SP

s (

mg

)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.2Cr

0.5Cr

1Cr

(a)

Interruptio n time ( day)

0 2 4 6 8 10

Rele

ased

EN

SP

s (

mg

)

0.00

0.05

0.10

0.15

0.200.2Cr

0.5Cr

1Cr

(b)

Fig. 7. Retained and released ENSPs mass during stop-flow durations as a functionof stop-flow duration in different columns (0.2C , 0.5C , and 1C ). Plot a showingrd

Td(rdbpa2TttitbaTpw

tMpdcHtIatbtst

r r r

etained ENSPs during the ENSPs injection phase and plot b showing ENSPs releaseuring the ENSPs release phase.

he calculated DLVO force was either attractive or repulsiveepending on the distance between particle and collector surfacesFig. S3 of Supporting Information). It was an attractive forceanging from −10−9 N to 0 when particle and collector separationistance was less than 0.2 nm (i.e., within the primary energyarrier). It became repulsive ranging from 10−12 to 10−9 N whenarticle and collector separation distance was from 0.2 to 23 nm,nd became attractive again as separation distance was larger than3 nm in the order of −10−12 N at the secondary energy minimal.hese calculations indicated that the DLVO force, which was largerhan the gravitational force, controlled particle retention duringhe SF durations. The sedimentation, however, may have facilitatedn bringing particles to collide with collector surfaces in addition tohe Brownian diffusion. A larger decrease in ENSPs concentrationefore and after SF events was shown in the column injected with

larger ENSPs concentration (Fig. S4 of Supporting Information).his result was consistent with the collision explanation that therobability of particle collision with collector surfaces increasedith increasing aqueous ENSPs concentration in the columns.

The deposited ENSPs during the flow interruptions only con-ributed to the straining pool based on the simulations from model

5. In contrast, the simulated ENSPs in the attachment/detachmentool decreased during the SF durations, suggesting that particlesetached from this pool, apparently driven by the decreased con-entration of aqueous ENSPs, which migrated to the straining pool.owever, the decrease in the attachment/detachment pool was less

han the increase in the straining pool (Fig. S4 and S5 of Supportingnformation), leading to the net increase in total retained particlesnd decrease in aqueous particle concentration. The importance ofhe straining process during the flow interruptions was unexpected

ecause there was no hydrodynamic force, which has been showno be an important factor for particle straining [21,42,43]. Diffu-ion apparently played an important role in bringing the particleso the straining locations. However, such an observation was based

cochem. Eng. Aspects 417 (2013) 89– 98

on the assumption that particle retention and release processesthat were applicable under flow conditions were also applica-ble to SF durations. This is a validated assumption for a soluteor dissolved contaminants because intrinsic aqueous and surfacereactions are independent of flow conditions and the same reactionkinetic expression can be applied during flow and no-flow periods.For a particle, however, retention and release models such as mod-els M1–M5 are typically derived under flow conditions, and theirapplicability during flow interruption periods has not been wellunderstood. Nevertheless, this study indicated that flow interrup-tion can be an important factor controlling particle transport, andflow interruption technique may be a useful experimental tech-nique to examine particle transport in porous media.

The SF events also affected ENSPs release (Fig. 2a). The amountof released ENSPs during the SF events (Fig. 7b) reflected the col-lective effect of two factors: time duration and driving force. Alonger stop-flow duration and a larger driving force caused by theimbalance between aqueous and retained phases (Eqs. (3) and (5))will allow more particles to release from their retained locationsand to accumulate in aqueous phase. The driving force for par-ticle release was stronger at the earlier time because of a higherresidual concentration of attached particles. In the column experi-ments, the decreasing of the driving force at the later time stop-flowevents was compensated with increasing time of stop-flow dura-tion (Fig. 2a). Consequently, the particles released during the twostop-flow events for each column were similar (Fig. 7b).

4. Conclusions

Engineered nanoporous silicate particles that have high selectiv-ity, dense population of adsorbing sites, fast sorption kinetics, andstable porous structure are a potentially important class of materi-als for environmental remediation, such as capturing radionuclidesfrom contaminated groundwater. This study demonstrated thatENSPs transport in porous media was strongly affected by con-centration and flow dynamics. Effect of particle concentration onparticle transport has been relatively well investigated and ourresult was generally consistent with those studies in colloidalparticle transport. These results indicated that the transport ofnanoporous particles was controlled by their surface properties,and the internal surfaces and properties within the nanoporousparticles would have minimal effect on particle transport, asexpected.

Groundwater flow dynamics is a common phenomenon in fieldthat is not well studied and understood in its effect on parti-cle transport. The results presented in this study indicated thatparticle transport models incorporating attachment and strainingprocesses can reasonably well describe the particle transport underboth flow and stop-flow conditions, despite the fact that the appliedparticle transport models were derived under steady-state flowconditions. The success in using the same models to describe par-ticle retention/release during the flow and SF durations, however,does not necessarily indicate that the same forces control particletransport. During the SF duration, diffusion and surface interac-tions are expected to be the major mechanisms controlling particleretentions, while under flow condition, hydrodynamic and DLVOforce collectively affect particle transport. Nevertheless, the factthat the same models can describe the particle retention/releaseduring flow and SF durations can significantly simplify simulatingparticle transport under dynamic flow conditions.

Particle concentrations and flow dynamics are two important

considerations in applying ENSPs in remediating contaminants.The particle concentration affects the total capacity of injectedmaterials in removing or sequestrating contaminants from pol-luted groundwater or sediments, while the dynamic flow condition

Physi

ratelcwmoiflirdc

A

OEw(NI

A

f2

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

J. Shang et al. / Colloids and Surfaces A:

eflects the realistic field condition that can affect the deliverynd transport of injected particles. Generally our results showedhat increasing the concentration of ENSPs in the particle deliv-ry solution would increase the particle transport distance andocal ENSPs concentrations, which would subsequently affect theontaminant partitioning between aqueous and solid phases, andould enhance contaminant desorption from contaminated sedi-ents. The optimal particle concentration would therefore depend

n remediation approaches. For in situ sequestration approach,ncreased retention and controlled delivery method such as stop-ow-stop would enhance remediation. On the other hand, particle

njection and extraction approach would require minimal particleetention at the end of remediation. The particle concentrations,elivery method, and field hydraulic conditions need to be carefullyonsidered in designing an optimal remediation approach.

cknowledgements

This research was funded by U.S. Department of Energy (DOE)ffice of Biological & Environmental Research (BER) through a jointPA-NSF-DOE research program. The research and experimentsere performed at Environmental Molecular Science Laboratory

EMSL), a DOE national user facility located at Pacific Northwestational Laboratory (PNNL). PNNL is operated by Battelle Memorial

nstitute under subcontract DE-AC06-76RLO 1830.

ppendix A. Supplementary data

Supplementary data associated with this article can beound, in the online version, at http://dx.doi.org/10.1016/j.colsurfa.012.10.030.

eferences

[1] G.Q. Lu, X.S. Zhao, Nanoporous Materials: Science and Engineering, ImperialCollege Press, London, 2004.

[2] J. Li, K. Zhu, J. Shang, D. Wang, Z. Nie, R. Guo, C. Liu, Z. Wang, X. Li, J. Liu, Flu-orescent functionalized mesoporous silica for radioactive material extraction,Sep. Sci. Technol. 47 (2012) 1507–1513.

[3] W. Dong, W.P. Ball, C. Liu, Z. Wang, A.T. Stone, J. Bai, J.M. Zachara, Influence ofcalcite and dissolved calcium on uranium(VI) sorption to a Hanford subsurfacesediment, Environ. Sci. Technol. 39 (2005) 7949–7955.

[4] D.L. Kaplan, K.J. Cantrell, T.W. Wietsma, M.A. Potter, Retention of zero-valentiron colloids by sand columns: application to chemical barrier formation, J.Environ. Qual. 25 (1996) 1086–1094.

[5] J. Gregory, Interaction of unequal double-layers at constant charge, J. ColloidInterface Sci. 51 (1975) 44–51.

[6] M.M. Sharma, H. Chamoun, D.S.H.S.R. Sarma, R.S. Schehter, Factors controllingthe hydrodynamic detachment of particles from surfaces, J. Colloid InterfaceSci. 149 (1992) 121–134.

[7] J.N. Ryan, M. Elimelech, Colloid mobilization and transport in groundwater,Colloids Surf. Physicochem. Eng. Aspects 107 (1996) 1–56.

[8] S.A. Bradford, S. Torkzaban, Colloid transport and retention in unsaturatedporous media: a review of interface-, collector-, and pore-scale processes andmodels, Vadose Zone J. 7 (2008) 667–681.

[9] R.S. Cushing, D.F. Lawler, Depth filtration: fundamental investigation throughthree dimensional trajectory analysis, J. Environ. Qual. 32 (1998) 3793–3801.

10] S.A. Bradford, S.R. Yates, M. Bettahar, J. Simunek, Physical factors affecting thetransport and fate of colloids in saturated porous media, Water Resour. Res. 38(2002) 1327.

11] S.P. Xu, B. Gao, J.E. Saiers, Straining of colloidal particles in saturated porousmedia, Water Resour. Res. 42 (2006) W12S216, http://dx.doi.org/10.1029/2006WR004948.

12] S.A. Bradford, J. Simunek, M. Bettahar, M.T. van Genuchten, S.R. Yates, Sig-nificance of straining in colloid deposition: evidence and implications, WaterResour. Res. 42 (2006) W12S15, http://dx.doi.org/10.1029/2005WR004791.

13] X.Q. Li, C.L. Lin, J.D. Miller, W.P. Johnson, Pore-scale observation of micro-sphere deposition at grain-to-grain contacts over assemblage-scale porous

media domains using X-ray microtomography, Environ. Sci. Technol. 40 (2006)3762–3768.

14] X.Q. Li, C.L. Lin, J.D. Miller, W.P. Johnson, Role of grain-to-grain contacts onprofiles of retained colloids in porous media in the presence of an energy barrierto deposition, Environ. Sci. Technol. 40 (2006) 3769–3774.

[

cochem. Eng. Aspects 417 (2013) 89– 98 97

15] W.P. Johnson, X.Q. Li, G. Yal, Colloid retention in porous media: mechanisticconfirmation of wedging and retention in zones of flow stagnation, Environ.Sci. Technol. 41 (2007) 1279–1287.

16] R. Lindqvist, J.S. Cho, C.G. Enfield, A kinetic-model for cell-density dependentbacterial transport in porous-media, Water Resour. Res. 30 (1994) 3291–3299.

17] Y. Tan, J.T. Gannon, P. Baveye, M. Alexander, Transport of bacteria in an aquifersand – experiments and model simulations, Water Resour. Res. 30 (1994)3243–3252.

18] P.R. Johnson, M. Elimelech, Dynamics of colloid deposition in porous-media– blocking based on random sequential adsorption, Langmuir 11 (1995)801–812.

19] T.A. Camesano, K.M. Unice, B.E. Logan, Blocking and ripening of colloids inporous media and their implications for bacterial transport, Colloids Surf.Physicochem. Eng. Aspects 160 (1999) 291–308.

20] S.A. Bradford, M. Bettahar, Concentration dependent transport of colloids insaturated porous media, J. Environ. Qual. 82 (2006) 99–117.

21] S.A. Bradford, H.N. Kim, B.Z. Haznedaroglu, S. Torkzaban, S.L. Walker, Coupledfactors influencing concentration-dependent colloid transport and retention insaturated porous media, Environ. Sci. Technol. 43 (2009) 6996–7002.

22] R. Ma, C. Zheng, H. Prommer, J. Greskowiak, C. Liu, J. Zachara, M. Rockhold, Afield-scale reactive transport model for U(VI) migration influenced by coupledmultirate mass transfer and surface complexation reactions, Water Resour. Res.46 (2010) W05509, http://dx.doi.org/10.1029/2009WR008168.

23] J. Zhuang, J.F. McCarthy, E. Perfect, J. Tyner, M. Flury, T. Steenhuis, In-situ colloidmobilization in Hanford sediments under unsaturated transient flow condi-tion: effect of irrigation pattern, Environ. Sci. Technol. 41 (2007) 3199–3204.

24] J. Shang, M. Flury, G. Chen, J. Zhuang, Impact of flow rate, watercontent, and capillary forces on in situ colloid mobilization during infil-tration in unsaturated sediments, Water Resour. Res. 44 (2008) W06411,http://dx.doi.org/10.1029/2007WR006516.

25] J. Zhuang, J.S. Tyner, E. Perfect, Colloid transport and remobilization in porousmedia during infiltration and drainage, J. Hydrol. 377 (2009) 112–119.

26] J. Kobler, K. Moller, T. Bein, Colloidal suspensions of functionalized mesoporoussilica nanoparticles, ACS Nano 2 (2008) 791–799.

27] G.E. Fryxell, H. Wu, Y.H. Lin, W.J. Shaw, J.C. Birnbaum, J.C. Linehan, Z.M. Nie, K.Kemner, S. Kelly, Lanthanide selective sorbents: self-assembled monolayers onmesoporous supports (SAMMS), J. Mater. Chem. 14 (2004) 3356–3363.

28] E.P. Barrett, L.G. Joyner, P.P. Halenda, The determination of pore volumeand area distributions in porous substance I. Computations from nitrogenisotherms, J. Am. Chem. Soc. 73 (1951) 373–380.

29] L.G. Joyner, E.P. Barrett, R. Skold, The determination of pore volume andarea distribution in porous substances. 2. Comparison between nitrogenisotherm and mercury porosimeter methods, J. Am. Chem. Soc. 73 (1951) 3155–3158.

30] S.J. Gregg, K.S.W. Sing, Adsorption, Surface Area and Porosity, Academic Press,Inc., New York, 1982.

31] C. Liu, J.M. Zachara, N.P. Qafoku, Z. Wang, Scale-dependent desorption of ura-nium from contaminated subsurface sediments, Water Resour. Res. 44 (2008)W08413, http://dx.doi.org/10.1029/2007WR006478.

32] C.H. Ko, M. Elimelech, The “shadow effect” in colloid transport and depositiondynamics in granular porous media: measurements and mechanisms, Environ.Sci. Technol. 34 (2000) 3681–3689.

33] J. Simunek, K. Huang, M.T. van Genuchten, The HYDRUS code for simulat-ing the one-dimensional movement of water, heat and multiple solutes invariably-saturated media, in: Research Report No. 144, Salinity U.S. Labora-tory, Agricultural Research Service, US Department of Agriculture, Riverside,Calif., 1998.

34] S.A. Bradford, J. Simunek, M. Bettahar, M.T. Van Genuchten, S.R. Yates, Mod-eling colloid attachment, straining, and exclusion in saturated porous media,Environ. Sci. Technol. 37 (2003) 2242–2250.

35] Z. Adamczyk, B. Siwek, M. Zembala, P. Belouschek, Kinetics of localized adsorp-tion of colloid particles, Adv. Colloid Interface Sci. 48 (1994) 151–280.

36] J. Shang, C. Liu, Z. Wang, K. Zhu, J. Li, J. Liu, J.M. Zachara, In-situ measurements ofengineered nanoporous particle transport in saturated porous media, Environ.Sci. Technol. 44 (2010) 8190–8195.

37] C. Shen, B. Li, Y. Huang, Y. Jin, Kinetics of coupled primary- and secondary-minimum deposition of colloids under unfavorable chemical conditions,Environ. Sci. Technol. 41 (2007) 6976–6982.

38] J. Shang, M. Flury, Y. Deng, Force measurements between particles andthe air–water interface: implications for particle mobilization in unsatu-rated porous media, Water Resour. Res. 45 (2009) W06420, http://dx.doi.org/10.1029/2008WR007384.

39] N.P. Qafoku, J.M. Zachara, C. Liu, P.L. Gassman, O.S. Qafoku, S.C. Smith, Kineticdesorption and sorption of U(VI) during reactive transport in a contaminatedHanford sediment, Environ. Sci. Technol. 39 (2005) 3157–3165.

40] C. Liu, Z. Shi, J.M. Zachara, Kinetics of uranium(VI) desorption from contam-inated sediments: effect of geochemical conditions and model evaluation,Environ. Sci. Technol. 43 (2009) 6560–6566.

41] D.S. Wang, C.P. Gerba, J.C. Lance, Effects of soil permeability on virus removalthough soil columns, Appl. Environ. Microbiol. 42 (1981) 83–88.

42] M.P. Tong, W.P. Johnson, Excess colloid retention in porous media as a function

of colloid size, fluid velocity, and grain angularity, Environ. Sci. Technol. 40(2006) 7725–7731.

43] S. Torkzaban, S.A. Bradford, S.L. Walker, Resolving the coupled effects of hydro-dynamics and DLVO forces on colloid attachment in porous media, Langmuir23 (2007) 9652–9660.

9 Physi

[

[

[46] M. Elimelech, Colloid filtration in fluidized-beds – discussion, J. Environ. Eng.,

8 J. Shang et al. / Colloids and Surfaces A:

44] S.A. Bradford, S. Torkzaban, S.L. Walker, Coupling of physical and chemical

mechanisms of colloid straining in saturated porous media, Water Res. 41(2007) 3012–3024.

45] N. Tufenkji, G. Miller, J. Ryan, R. Harvey, M. Lech, Transport of Cryptosporid-ium oocysts in porous media: role of straining and physicochemical filtration,Environ. Sci. Technol. 38 (2004) 5932–5938.

[

cochem. Eng. Aspects 417 (2013) 89– 98

ASCE 117 (1991) 706–708.47] R. Kretzschmar, M. Borkovec, D. Grolimund, M. Elimelech, Mobile subsur-

face colloids and their role in contaminant transport, Adv. Agron. 66 (1999)121–193.