Pore-Scale Quantification of ColloidTransport in Saturated PorousMediaJ E N N I F E R S M I T H , † B I N G A O , ‡

H I S A K A G E F U N A B A S H I , † T H U A N . T R A N , †

D A N L U O , † B E T H A . A H N E R , †

T A M M O S . S T E E N H U I S , †

A N T H O N Y G . H A Y , § A N DM . T O D D W A L T E R * , †

Department of Biological and Environmental Engineering,Cornell University, Ithaca, New York 14853, Department ofAgricultural & Biological Engineering, University of Florida,Gainesville, Florida 32611, and Department of Microbiology,Cornell University, Ithaca, New York 14853

Received March 28, 2007. Revised manuscript received No-vember 6, 2007. Accepted November 12, 2007.

It is currently not clear how to quantifiably relate pore-scaleobservations of colloid transport to larger scales, so, we proposeda geometric theory showing that pore-scale-derived rateconstants may be appropriate to model a larger scale system.This study considered three different types of colloids: latexmicrospheres, Escherichia coli, and microspheres made of polylactic acid (PLA). Colloid attachment and detachment rateconstants were calculated using digital microscope images,taken in rapid (1 s) sequences, from which rates of attachingand detaching colloids were readily observed. Average rateconstants from >1000 images per colloid-type were used tomodel Darcy-scale colloid transport breakthrough curves. Themodeled and observed breakthrough curves agreed well forall threetypesofcolloids.However, for latexandPLAmicrospheres,the model systematically under predicted the breakthroughcurves’ rising limb, which may indicate that the rate “constants”are actually dependent on the amount of attached colloids.Insights into these sorts of complexities are best addressed byresearch that considers both pore-scale phenomena and larger-scale transport responses.

IntroductionThere is increasing concern about subsurface transport ofcolloidal-sized materials, diameter of 1 nm to 10 µm,especially pathogenic biocolloids and abiotic materials towhich normally immobile contaminants may attach and betransported (1). Colloid transport mechanisms are alsoimportant for bioaugmentation and biofouling during biore-mediation. The ability to predict colloid transport is still notfully realized, especially with respect to the quantification ofcolloid reaction rates during transport through porous media.Most previous studies have been based on mass balance andbreakthrough curves in packed sand columns (2–5); as such,predictions of tracer movement have been limited to “black

box” modeling. Several researchers have used pore-scalevisualization techniques to gain insight on mechanisms thatwould be difficult to observe in the environment (6–10). Thesevisualization studies have provided valuable qualitativeinformation about features of transport mechanisms, e.g.,straining, grain retention, retention in immobile pore spaces,and air/water/surface interface interactions. Only a fewstudies have used pore-scale visualizations to describe colloidtransport quantitatively (6). There remains a critical knowl-edge gap in quantifiably linking the observed pore-scalemechanisms to large-scale colloid transport behavior.

The convective-diffusion equation for colloids in saturatedmedia can be expressed as

∂C∂t

)D∂

2C

∂x2- v

∂C∂x

-Fb

n∂S∂t

(1)

where C is the concentration of suspended colloids (mg L-1),D is the diffusion or hydrodynamic dispersion coefficient(m2 h-1), ν is the average linear water velocity (m h-1), x isthe travel distance in the direction of flow (m), Fb is the mediabulk density (g L-1), n is the porosity, and S is the deposited/retained colloid concentration (mg colloid per g sand). Thelast term includes both the rate of colloid retention andremobilization. Colloid transport studies have generally onlyconsidered bulk attachment, modeling the process with firstorder irreversible kinetics. This is due in part from the classicfiltration theory first developed by Yao et al. (11), whichprovides an easy way to calculate colloid transport rate inporous media (12). The clean bed filtration theory works finefor the idealized experimental systems (i.e., columns packedwith uniform sized glass beads), however, several studiesindicate that the theory needs to be modified to apply to thereal porous medium such as quartz sand (e.g., refs 13 and14). More recent studies have shown that both attachmentand detachment processes are significant in colloid transport(3, 4). The colloid deposition term can be expressed as

∂S∂t

) nFb

kaC- kdS (2)

where ka is the first-order attachment rate constant (h-1)and kd is the first-order detachment rate constant (h-1). Inprevious studies, ka and kd have been calibrated to best-fitthese equations to breakthrough-curve data (3, 4). However,with any optimized modeling investigation involving severalparameters, similarly good fits can be achieved with morethan one combination of variables.

In this study, we determined colloid breakthrough curvesfrom a flow chamber and also visualized colloid transport atthe pore scale via rapid (1 s) digital image sequences. Ourmain objective was to mechanistically model Darcy-scalecolloid transport using independently estimated, pore-scaleattachment and detachment rate constants. Another objectivewas to compare the transport behavior of three types ofcolloids: carboxylated latex microspheres, bacteria (Escheri-chia coli), and poly(L-lactide) (PLA) microspheres. The lattercolloid type, PLA microsphere, was chosen because we believeit may serve as an ideal colloidal tracer at landscape scalessince it can encapsulate unique labels (e.g., DNA or nanobar-codes) to uniquely distinguish one sphere from another, thusallowing researchers to simultaneously distinguish amongseveral interacting tracer signals. Unlike commonly used latexmicrospheres, PLA is degradable (nonpolluting), biocom-patible, and nontoxic both as a polymer and degradationproducts (15). In fact, PLA has been widely investigated formedical applications, e.g., dissolvable sutures and slow-

* Corresponding author phone: 607-255-2488; fax: 607-255-4080;e-mail: mtw5@cornell.edu.

† Department of Biological and Environmental Engineering,Cornell University.

‡ University of Florida.§ Department of Microbiology, Cornell University.

Environ. Sci. Technol. 2008, 42, 517–523

10.1021/es070736x CCC: $40.75 2008 American Chemical Society VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 517

Published on Web 12/15/2007

release drugs and other chemicals, and as such, has beenapproved by the FDA for human clinical applications. BecausePLA has been so widely used in medicine, we can build oncopious experience in fabricating PLA microspheres withspecific properties (16, 17). The only environmental ap-plication of degradable microspheres to date is the work byQuaglia et al. (18), who used microspheres composed ofmono-, di-, and triglycerides encapsulating a pesticide as anovel method for application of agrochemicals.

Theory. It is currently unclear how to quantifiably relatepore-scale observations to larger scale systems. As a firstapproximation, we provide the following logic to suggestthat attachment and detachment rates observed at the pore-scale may be appropriate for describing Darcy-scale behavior.At the pore-scale, we view the system through a microscopeas a plane in which we can distinguish among pore, soilgrain, and colloid areas, Ap, Ag, and Ac, respectively. Since weare focusing on saturated soils, all pore area is filled withwater and colloids. We can translate Ap and Ag into volumesby multiplying by the focal depth (d). Assuming the colloidsare spherical with radius, r, we can translate Ac into a volumeby multiplying by 4 r/3. Now, C, S, and n in eqs 1 and 2 canbe written as

S )FcAcg4/3r

FgAgd(3)

C )FcAcp4/3r

Apd(4)

n )Apd

ATd(5)

where Fc is the density a colloid (mg m-3), Fg is the densityof the soil grain (mg m-3), Acp is the area of colloids suspendedin the pore space as seen in the microscope images (m2), andAcg is the area of colloids attached on the grain in themicroscope images (m2). We can convert bulk density intoa two-dimensional parameter via Fb ) (1 - n)Fg and notingthat 1 - n is the fraction of porous media occupied by sandgrains, which in the microscope images can be expressed asAg/AT (AT is the total image area (m2), (Ap + Ag + Ac)).Substituting eqs 3-5 and Fb )Fg/(Ag/AT) into eq 2 and divingthrough by the mass of an individual colloid, reduces it to

∆Ng

∆t) kaNp - kdNg (6)

where “Ng” and “Np” are a “colloid counts” for the parts ofan image associated with the soil grain surface and the porespace, respectively and ∆t is the time between two observa-tions (microscope images). Note, the rate constants ka andkd for the two-dimensional system are, at least theoretically,the same as ka and kd for the three-dimensional system,respectively.

Materials and MethodsExperimental Apparatus. Colloids investigated were car-boxylated latex fluorescent microspheres (Magsphere, Pasa-dena, CA, emission ) 488 nm, excitation) 528), geneticallymodified Escherichia coli K-12 PHL 628 that constitutivelyexpressed a green fluorescent protein (emission ) 488 nm,excitation ) 528 nm) (20, 21), and PLA microspheres(16, 22–24). Colloid properties are summarized in Table 1and additional details on colloid preparation can be foundin the Supporting Information. All experiments used ahorizontal flow cell (Figure 1) packed with silica sand (300–600µm, Unimin Co., NJ) acid washed in 50% nitric acid, rinsedin deionized water, and subsequently dried at 105 °C for 24 h(19). A ceramic porous plate (RH 1000-course pores, R&HFilter Co.) was placed on each end of the column to evenly

distribute flow. A syringe pump provided steady influent flowrates (0.15 mL min-1, v ) 116 cm h-1) and a peristaltic pumpoperating at a higher rate was used to remove effluent fromthe downstream end of the flow chamber. To stabilize pHin the system, all colloids were suspended in phosphate buffersolution (PBS: 0.26 g KH2PO4, 1.156 g Na2HPO4 per L Milli Q,pH 7.4, ionic strength)23 mM). PBS and sand were sterilizedvia autoclaving for bacteria and PLA experiments. A con-servative tracer (0.1 M NaCl) was used to estimate thedispersion coefficient, and verify the calculated average flowvelocity and sand porosity.

Column Breakthrough Curve and Mass Balance. Break-through curves were separately triplicated for NaCl (0.1M),latex microspheres (∼5 × 108 spheres mlL-1), bacteria (∼5× 108 cells ml-1), and PLA microspheres (∼5 × 108 spheresmL-1). Preceding bacteria and PLA microsphere experiments,the system was rinsed with 15% (v/v) bleach followed byautoclaved PBS to ensure sterile conditions. The flowchamber was repacked with sand before each experiment.For each experiment, flow was allowed to equilibrate for 10min (∼4 pore volumes) before a pulse of NaCl or colloidswas injected into the chamber followed by several minutes

TABLE 1. Physical and Chemical Properties of Sand and Col-loids. Parentheses Show Standard Deviations on Four Sam-ples for Zeta Potential and the Octanol–Water PartitioningCoefficient (Kow)a

diameter(µm)

zeta potential(mV) Kow

sand 300–600 -58.9 (4)latex microspheres 1.1 -65.9 (3.6) 0.64 (0.10)bacteria 1.0 × 1.5 -40.4 (5.0)PLA microspheres 1.04 (0.4) -36.7 (4.3) 3.84 (0.68)

a Parentheses for PLA microspheres diameter show stan-dard deviation of 1000 spheres analyzed.

FIGURE 1. Schematic of flow chamber; dimensions ) 4.68 ×1.15 × 0.3 cm; sand porosity, n, ) 0.34; flow rate ) 0.15 mLmin-1.

FIGURE 2. Schematic showing how attachment and de-tachment rates were calculated based on images taken onceper second. The dark circles indicate colloids. The first rowrepresents the original data. The second row shows how manycolloids were attached between two consecutive images. Asindicated in the third row, the change in the number ofattached colloids between images was used to determine theattachment and detachment rates constant. This method issimilar to that presented in ref 29.

518 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 2, 2008

of NaCl or colloid-free influent to flush the system. Sampleswere collected every minute. The NaCl pulse was 15 min (6.3pore volumes); the colloid pulse was 27 min (11.3 porevolumes). Cl- was measured with a Buchler Digital Chlo-rimeter model 442–5000 (limit of detection (LOD) ) 1% ofthe influent concentration (Co)). Concentrations of latexmicrospheres, bacteria, and PLA microspheres were mea-sured with a fluorometer (LOD were, respectively, 2, 3, and3% of Co). Special care was taken to preserve fluorophoreintegrity in PLA micrspheres by minimizing exposure to lightduring experimentation. To remove fluorescent taggers

released during PLA microsphere degradation, effluentsamples were centrifuged (4000g, 4 °C, 2 min), 100 µLdecanted from the supernatant, and resuspended in freshautoclaved PBS. To account for colloids sorbed to sand grains,the sand with attached colloids was placed in 500 µL PBSand sonicated for 10 min, except bacteria which weresonicated 1 min, and the aqueous phase analyzed. Colloidretention within parts of the apparatus other than the porousmedia (e.g., tubing, syringe pump, etc.) was determined byrunning (in triplicate) empty flow-cell experiments for latexand PLA microspheres. We subtracted-out these apparatus-retained colloids in subsequent analyses, so that mass balanceand breakthrough curves represented only the colloids inthe effluent or sand fractions. Thus Co was calculated as thepercent recoverable in empty flow-cell experiments multi-plied by the concentration placed in the syringe pump at thebeginning of each experiment.

Pore Scale Analysis. Colloids on sand grain surfaces andin the aqueous phase were observed via microscopy (LeicaTCS SP2 spectral confocal microscope, 10×, 0.4 UV objective),as described in Zevi et al. (9). Rhodamine B (0.0004%) wasmixed with the colloid solution to visually distinguish waterfrom air in the pores. The Leica system allowed for visual-ization of three separate channels: transmitted light showinggrains, green fluorescence showing colloids (argon laser,excitation ) 488 nm, emission ) 543 nm) and red fluores-cence showing water (HeNe laser excitation ) 543 nm,emission ) 638 nm), which could be superimposed onto asingle image using the integrated Leica confocal software.Random locations in the column were observed to accountfor sand heterogeneity. At each location, a series of images(1 s time steps), collected over 60 s, were compiled using theLeica software;>1000 images were captured for each colloid.

A threshold function in image J (25) was used on the greenfluorescence image (showing colloid positions) based onintensity and converted to a binary image. The thresholdfunction was used as a filter to eliminate colloids that werenot in the focal plane but could still emit damped fluores-cence. A threshold value was chosen for each series sincefluorescent intensity could have some variation between porelocations due to the water depth above the focal plane andother condition variations. To quantify attached colloids onthe grain, the in-focus portion of the grain surfaces, delineatedby hand from the transmitted light image, was overlain withthat from the green fluorescent images (colloids), to isolatecolloids potentially attached to the grain surface and storedin a separate image. Image J was used to count the numberof colloids that were at the same location between every pairof consecutive images; these colloids were consideredattached, i.e., Ng. A particle was considered attached whenthe colloid-pixels from two separate consecutive imagesoverlapped by at least 2 pixels. A single colloid (area ∼ 0.78µm2) was approximately 3–5 pixels in images that were 512

FIGURE 3. Four consecutive images of one of our pore-scale ex-periments in which we point out two colloids that are attachedthroughout the sequence (1) and (2) and another that attachesand then detaches during this period (3). t ) 0 is a bright fieldimage to provide context for the dashed lines, which show theapproximate area of the in-focus sand grain surfaces. Althoughwe have only highlighted three colloids here, we recognizethat there are other attached and detaching colloids in theseimages.

TABLE 2. Attachment and Detachment Rate Constant ValuesUsed to Model Column-Scale Transport. Upper and Lower 95%Confidence Interval (CI) Limits Were Used for SensitivityAnalysis on the Model

colloidattachment rateconstant (h-1)

detchment rateconstant (h-1)

ka

upperCI

lowerCI kd

upperCI

lowerCI

latexmicrospheres 616 703 530 460 516 406

bacteria(E. coli) 465 531 400 1193 1292 1096

PLAmicrospheres 808 892 724 302 328 276

VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 519

× 256 pixels. The focal depth, d, was 3–5 µm. Colloids in thepore space were similarly quantified by overlaying the porearea in the red fluorescence image with colloids from thegreen fluorescence image. Since the chamber was saturated,the grain surface in the transmitted light image overlappedwith the pore area in the red fluorescent image. The porearea did not include grain surface areas.

The sequential composite images of attached colloids werecompared to determine positive and negative changes in thenumber of attached colloids, ∆Ng/∆t (eq 6) and, subsequentlyto calculate the attachment and detachment rate constants,separately (Figure 2). Equation 6 was rearranged to solve forthe attachment and detachment rate constants assumingthat kd)0 if ∆Ng/∆t>0 between a consecutive pair of imagesand visa versa for ka:

ka )1

Np

∆Ng

∆t(7)

kd )- 1Ng

∆Ng

∆t(8)

This simplification results in underestimated rate con-stants when both attachment and detachment occur simul-taneously within a 1 s time interval. All rate constants werecalculated during Darcy-scale steady state conditions. Finalrate constants were estimated by taking averages over 27, 29,and 39 image series for latex microspheres, bacteria and PLAmicrospheres, respectively (>50 images per series). Note,incidents when ∆Ng ) 0 were used in both attachment anddetachment calculations. Figure 3 illustrates examples ofattaching and detaching particles for a sequence of fourconsecutive images taken at 1 s intervals.

Results and DiscussionPore Scale Attachment and Detachment. Average attach-ment rate constants (eq 7), ranged from 465 to 808 h-1 forbacteria and PLA microspheres, respectively (Table 2). Therange of detachment rates (eq 8) was nearly twice as wide,from 302 to 1193 h-1 for PLA microspheres and bacteria,respectively. To evaluate whether a sufficient number ofimages was used to achieve representative rate constants,the constants were calculated 1000 times by randomlyselecting 50% of the data. Rate constants calculated withrandomly selected data were normally distributed, and fellwithin the 95% confidence limits (CI) of the respective dataset for each colloid type (Table 2). Differences between theaverage rate constants and the confidence limits werebetween 9 and 14%.

Darcy-Scale Breakthrough Curves. The dispersion coef-ficient (D) was determined by fitting the one-dimensionalconvective-dispersion equation for a conservative, nonre-active substance (i.e., eq 1 with no storage term) to the Cl-

breakthrough data (Figure 4a): D) 54 cm2 h-1. Breakthroughcurves for each type of colloid were modeled (eqs 1 and 2)using these values for D and v and attachment and detach-ment rate constants from the pore scale analysis (Figure 4).Although the attachment and detachment rate constants weredetermined from observations at a much smaller scale, themodel worked well for all the microspheres, especiallybacteria, which had standard error of the estimate (SEE) <0.05 and R2)0.99 (Figure 4). Indeed, although we can achievebetter fits by optimizing the rate constants, we feel theseuncalibrated results using independently determined pa-rameters are more meaningful.

Although our results suggest that converting betweenpore-scale and Darcy-scale colloid behavior can work well

FIGURE 4. Breakthrough curves for (a) chloride (standard error of the estimate (SEE) ) 0.084, R 2 ) 0.95), (b) bacteria (SEE ) 0.049,R 2 ) 0.99), (c) latex microspheres (SEE ) 0.126, R 2 ) 0.91), and (d) PLA microspheres (SEE ) 0.136, R 2 ) 0.85). Solid lines are themodels using visually derived rate constants. Error bars are standard deviations on triplicate samples. Where error bars are notshown, they are smaller than the symbol size.

520 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 2, 2008

and can be physically consistent between the scales, we mayhave been fortunate in that Ag/AT in our images was similarto (1-n) at the Darcy scale. It is likely that we achieved goodresults in part because we used so many images to get averagerate constants and, thus, probably inadvertently sampled arepresentative cross-section of the media heterogeneities.

One possible reason the model overestimated attachmentin the breakthrough curve data during transient conditionsfor both latex and PLA microspheres (Figure 4) is that themicroscopic images represented either too much poor spaceor too much sand grain relative to the bulk porosity, n.Another explanation for this discrepancy may be that ourrate constants were determined from images of steady stateflow and they may not capture the mechanisms during therising and falling parts of the break-through curve. Both hada higher attachment than detachment rate constants and, assuch, had deposition as the dominant mechanism, whichmay vary between transient and steady-state conditions. Forexample, the first colloids to attach may have been slowerto do so due to differences in deposition between grainsdevoid of colloids and when other attached colloids arepresent (26). Indeed, as can be seen in Figure 4, the initialPLA microspheres broke-through earlier than predicted bythe model, suggesting that the attachment rate constant wastoo high for this early period. Variable attachment/detach-ment rates have been seen by other researchers as well, e.g.,bond aging (6, 27, 28). Another potential disconnect between

our pore-scale rate constants and Darcy-scale model isunintentional bias in choosing observation sites with themicroscope, i.e., colloid-free pores, if any existed, were neverselected.

As mentioned earlier, previous colloid transport studieshave often used optimized parameters to model rateconstants. Bradford et al. (3) found modeled attachment anddetachment rate constants to be 3.4 × 105 and 7.8 × 104 h-1,respectively, for carboxylated latex microspheres in saturatedsand. Although our study also found a higher attachmentrate constant than detachment, the absolute differencesbetween our constants and Bradford et al. (3) are huge.However, this could be due to differences in systems (ionicstrengths, porous media, and flow rates). Tong et al. (5)reported an adhesion deficient bacterial (Comamonas DA001)deposition and re-entrainment rate coefficients for distrib-uted deposition as 2.86 and 1.57 h-1, where ionic strengths(50 mM) and average linear velocity varied significantly fromthis study (33 cm h-1 compared with 116 cm h-1 in this study).Bradford et al. (4) reported modeled attachment anddetachment rate constants for Cryptosporidium oocysts in1 mM NaBr of 6.6 and 0.78 h-1, respectively. Although it isdifficult to compare rates in this study with previous workbecause of experimental differences, studies that calibratetheir system constants are always susceptible to the problemof getting the “right answer” for the “wrong reasons.” Indeed,in our study the shapes of latex microsphere and bacteria

FIGURE 5. Sensitivity analysis of attachment and detachment rate constants for colloids using ka and kd associated with the upperand lower 95% confidence intervals (dashed lines); solid lines show model with original rate constants. (a), (c), and (e) showsensitivity to the attachment rate constant and (b), (d), and (f) for the detachment constant. (a) and (b) ) latex microspheres, (c) and(d) ) bacteria (E. coli), and (e) and (f) ) PLA microspheres.

VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 521

breakthrough curves were similar but their rate constantswere very different. Latex microspheres had a higher pore-scale attachment than detachment rate constant, whereasbacteria had a higher detachment rate constant. Thesemechanisms are different despite similar bulk breakthroughcurve data and it is possible that rate constants calculatedwith the breakthrough curve best-fit could misrepresenttransport mechanisms. For our experiments, best-fit attach-ment and detachment rate constants were as much as>100%higher than observed at the pore scale. Whether optimizedor based on pore-scale observations, our detachment rateconstants were high compared to our attachment rateconstants, which is why we observe very little “tailing” at theend of our experiments as one would expect to see in systemswhere the colloids are irreversibly attached to the media. Wesuspect that much of our attachment was associated with asecondary energy minima, but additional experiments areneeded to confirm this.

We tested how sensitive our model was to our attachmentand detachment rate constants by rerunning the model usingthe pore-scale derived rate constants associated with theupper and lower 95% confidence intervals (Table 2) for allthree colloid types (Figure 5). In all cases the model wasinsensitive to these changes in the rate constants with modelvariations much smaller than the variability in the experi-mental data. Thus, the disagreements between the modeland measurements are either due to missing mechanisms inthe model, e.g., straining or nonlinear attachment/detach-ment processes (e.g., ref 6), or problems in how we deter-mined the rate constants.

There are a few yet unmentioned potential artifacts in thepore scale rate constant that may cause problems whenapplied to Darcy-scales. For example, fluorescent colloidsjust above or below the focal plane may still emit fluorescencein the images and, although a threshold value was used to“filter” the image during image analysis to minimize theseeffects, our fluorescent-based counts may introduce someerror. Special care was taken to observe particle movementin each individual series before choosing a threshold value.Due to instrument limitations, we used a 1 s time durationto validate attachment of a single colloid. We found similarattachment and detachment values using time steps of both1 and 5 s but we have no way of checking if a shorter timestep would give different results. Unfortunately, it is unclearfrom the literature, how long a colloid needs to remainstationary on a grain surface to be considered “attached,”note the large range of published attachment coefficientsnoted earlier. It is also possible that our assumption that wecan use the depth of the focal plane, d, used to transferbetween 3-dimensional and 2-dimensional systems mightintroduce errors, especially at the edge of the pores whereparticle curvature into the image plane may introducedifferences in areas at the top and bottom of the focal plane.

PLA microspheres had a significantly higher attachmentrate constant than detachment and the attachment rateconstant was also higher compared with other colloids.Breakthrough curve data for PLA microspheres was moreattenuated, as compared to other colloids. PLA microsphereshave been studied extensively for human therapeutic pur-poses, where engineering their surface properties has rangedfrom influencing hydrophobicity and size to surface charge.This information will be used to design future studies tobetter simulate natural colloids in the environment with PLAmicrospheres.

This study demonstrates that complexities in colloidtransport are best addressed by research that considers bothpore-scale phenomena and larger-scale transport responses.Indeed, this approach may uniquely provide additionalinsights needed to help answer a number of questions raisedin this study, such as “how long should a colloid remain

stationary before it is considered attached” or “can we linkattachment times to specific attachment processes?”

AcknowledgmentsThis work was funded by the United States Department ofAgriculture (sponsor ID no. 2005–35603–15306) for NanoscaleScience and Engineering for Agriculture and Food Systemsand by fellowship from the National Water Research Institute.Special thanks to Dr. Annette Dathe, Yuniati Zevi, andRosemarie Fehrman who also contributed substantially todesigning the experimental apparatus and developing theimage analysis techniques for visualization experiments. Wealso thank the three anonymous reviewers for their helpfulcomments and suggestions.

Supporting Information AvailableColloid preparation techniques and methods for determiningvarious, relevant properties are described in detail. Thismaterials is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited(1) McCarthy, J. F.; McKay, L. D. Colliod Transport in the subsurface:

past, present, and future challenges. Vadose Zone J. 2004, 3,326–337.

(2) Lenhart, J. J.; Saiers, J. E. Transport of silica colloids throughunsaturated porous media: experimental results and modelcomparisons. Environ. Sci. Technol. 2002, 36, 769–777.

(3) Bradford, S. A.; Simunek, J.; Bettahar, M.; Van Genuchten, M.;Yates, S. R. Modeling colloid attachment, straining, and exclusionin saturated porous media. Environ. Sci. Technol. 2003, 37, 2242–2250.

(4) Bradford, S. A.; Bettahar, M. Straining, attachment and detach-ment of Cryptosporidium oocysts in saturated porous media.J. Environ Qual. 2005, 34, 469–478.

(5) Tong, M.; Camesano, T. A.; Johnson, W. P. Spatial variation indeposition rate coefficients of an adhesion-deficient bacterialstrain in quartz sand. Environ. Sci. Technol. 2005, 39, 2679–3687.

(6) Meinders, J. M.; Noordmans, J.; Busscher, H. J. Simultaneousmonitoring of the adsorption and desorption of colloid particlesduring deposition in a parallel plate flow chamber. J. ColloidInterface Sci. 1992, 152 (1), 265–280.

(7) Wan, J.; Wilson, J. L. Colliod transport in unsaturated porousmedia. Water Resour. Res. 1994, 30, 857–864.

(8) Crist, J. T.; McCarthy, J. F.; Zevi, Y.; Baveye, P.; Throop, J. A.;Steenhuis, T. S. Pore-scale visualization of colloid transport andretention in partially saturated porous media. Vadose Zone J.2004, 3, 444–450.

(9) Zevi, Y.; Dathe, A.; McCarthy, J. F.; Richards, B. K.; Steenhuis,T. S. Distribution of colloid particles onto interfaces in partiallysaturated sand. Environ. Sci. Technol. 2005, 39, 7055–7064.

(10) Gao, B.; Saiers, J. E.; Ryan, J. Pore-scale mechanisms of colloiddeposition and mobilization during steady and transient flowthrough unsaturated granular media. Water. Resour. Res. 2006.42, W01410, dio: 10.1029/2005WR004233.

(11) Yao, K. M.; Habibian, M. T.; O’Melia, C. R. Water and wastewaterfiltration: concepts and applications. Environ. Sci. Technol. 1971,5, 1105–1112.

(12) Rajagopalan, R.; Tien, C. Trajectory analysis of deep-bed filtrationwith the sphere-in-cell porous media model. AIChE J. 1976, 22,523–533.

(13) Martin, M. J.; Logan, B. E.; Johnson, W. P.; Jewett, D. J.; Arnold,R. G. Scaling bacterial filtration rates in different sized porousmedia. J. Environ. Eng. 1996, 122, 407–415.

(14) Burks, G. A.; Velegol, S. B.; Paramonova, E.; Lindenmuth, B. E.;Feick, J. D.; Logan, B. E. Macroscopic and nanoscale measure-ments of the adhesion of bacteria with varying outer layer surfacecomposition. Langmuir 2003, 19, 2366.

(15) Lee, W.; Iwata, T.; Gardella, J. A., Jr. Hydrolytic behavior ofenantiomeric poly(lactide) mixed monolayer films at the air/water interface: stereocomplexation effects. Langmuir 2005,21, 11180–11184.

(16) Luo, D.; Woodrow-Mumford, K.; Belcheva, N.; Saltzman, W. M.Controlled DNA delivery systems. Pharm. Res. 1999, 16, 1300–1308.

522 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 2, 2008

(17) Ouchi, T.; Toyohara, M.; Arimura, H.; Ohya, Y. Preparation ofPoly(L-lactide) based microspheres having a cationic or anionicsurface using biodegradable surfactants. Biomacromol. 2002,3 (5), 885–888.

(18) Quaglia, F.; Barbato, F.; De Rosa, G.; Granata, E.; Miro, A.; LaRotonda, M. I. Reduction of the environmental impact ofpesticides: waxy microspheres encapsulating the insecticidecarbaryl. J. Agric. Food Chem. 2001, 49, 4808–4812.

(19) Litton, G. M.; Olsen, T. M. Colliod deposition rates on silica bedmedia and artifacts related to collector surface preparationmethods. Environ. Sci. Technol. 1993, 27, 185–193.

(20) Vidal, O.; Longin, R.; Prigent-Combaret, C.; Dorel, C.; Hooreman,M.; Lejeune, P. Isolation of an Esherichia coli mutant strain ableto form biofilms on inert surfaces: involvement of a new OmpRallele that increases curli expression. J. Bacteriol. 1998, 180 (9),2442–2449.

(21) Evans, D. G.; Evans, D. J., Jr.; Tjoa, W. Hemagglutination ofhuman group A erythrocytes by enterotoxic Escherichia coliisolated from adults with diarrhea: correlation with colonizationfactor. Infect. Immunol. 1977, 18 (2), 330–337.

(22) Dai, X.; Hozalski, R. M. Evaluation of microspheres as surrogatesfor Cryptosporidium parvum oocysts in filtration experiments.Environ. Sci. Technol. 2003, 37, 1037–1042.

(23) Kiss, E.; Bertóti, I.; Butler, B. XPS and wettability characterizationof modified poly(lactic acid) and poly(lactic/glycolic acid) films.J. Colloid Sci. 2002, 245, 91–98.

(24) Van Loosdrecht, M.; Lyklema, J.; Norde, W.; Schraa, G.; Zehnder,A. The role of bacterial cell wall hydrophobicity in adhesion.Appl. Environ. Microbiol. 1987, 53 (8), 1893–1897.

(25) Rasband, W. S. , U. S. ImageJ; U. S. National Institutes of Health:Bethesda, MD, 1997–2005; http://rsb.info.nih.gov/ij/.

(26) Johnson, P. R.; Elimelech, M. Dynamics of colloid depositionin porous media: blocking base don random sequentialadsorption. Langmuir 1995, 11, 801–812.

(27) Meinders, J. M.; Vandermei, H. C.; Busscher, H. J. Physico-chemical aspects of deposition of Streptococcus-themophius-Bto hydrophobic and hydrophilic substrata in a parallel-plateflor chamber. J. Colloid Interface Sci. 1994, 164 (2), 355–363.

(28) Vadillo-Rodriguez, V.; Busscher, H. J.; Norde, W.; de Vries, J.;van der Mei, H. C. Atomic force microscopic corroboration ofbond aging for adhesion of Streptococcus thermophilus to solidsubstrata. J. Colloid Interface Sci. 2004, 278 (1), 251–254.

(29) Abdel-Fattah, A. I.; El-Genk, M. S.; Reimus, P. W. On visualizationof sub-micron particles with dark-field light microscopy. J.Colloid Interface Sci. 2002, 246 (2), 410–412.

ES070736X

VOL. 42, NO. 2, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 523