Investigation of Nanoparticle Transport Inside Coarse-Grained

Nov 18, 2011 - [email protected]. ... Transport and Adsorption of Nano-Colloids in Porous Media ... Transport in Porous Media 2017 119 (2), 403-42...
0 downloads 0 Views 3MB Size
ARTICLE pubs.acs.org/est

Investigation of Nanoparticle Transport Inside Coarse-Grained Geological Media Using Magnetic Resonance Imaging B. Ramanan,†,‡,§ W. M. Holmes,‡ W. T. Sloan,§ and V. R. Phoenix*,† †

School of Geographical and Earth Sciences, Gregory Building, University of Glasgow, Glasgow G12 8QQ, U.K. GEMRIC, Wellcome Surgical Institute, Institute of Neuroscience and Psychology, University of Glasgow, Glasgow G61 1QH, U.K. § School of Engineering, Rankine Building, University of Glasgow, Glasgow G12 8QQ, U.K. ‡

bS Supporting Information ABSTRACT: Quantifying nanoparticle (NP) transport inside saturated porous geological media is imperative for understanding their fate in a range of natural and engineered water systems. While most studies focus upon finer grained systems representative of soils and aquifers, very few examine coarsegrained systems representative of riverbeds and gravel based sustainable urban drainage systems. In this study, we investigated the potential of magnetic resonance imaging (MRI) to image transport behaviors of nanoparticles (NPs) through a saturated coarse-grained system. MRI successfully imaged the transport of superparamagnetic NPs, inside a porous column composed of quartz gravel using T2-weighted images. A calibration protocol was then used to convert T2-weighted images into spatially resolved quantitative concentration maps of NPs at different time intervals. Averaged concentration profiles of NPs clearly illustrates that transport of a positively charged amine-functionalized NP within the column was slower compared to that of a negatively charged carboxyl-functionalized NP, due to electrostatic attraction between positively charged NP and negatively charged quartz grains. Concentration profiles of NPs were then compared with those of a convection-dispersion model to estimate coefficients of dispersivity and retardation. For the amine functionalized NPs (which exhibited inhibited transport), a better model fit was obtained when permanent attachment (deposition) was incorporated into the model as opposed to nonpermanent attachment (retardation). This technology can be used to further explore transport processes of NPs inside coarse-grained porous media, either by using the wide range of commercially available (super)paramagnetically tagged NPs or by using custom-made tagged NPs.

’ INTRODUCTION Nanoparticles (NPs) are utilized in a tremendously diverse array of applications, including cosmetics, optics, medical technology, textiles, and catalysts.13 Problematically, once released into ground and surface waters NPs can have a diverse range of toxic effects.1 Consequently, it is imperative to develop a robust understanding of their transport behavior. From this we can develop the transport models needed in the assessment of nanoparticle contaminant risk. Most NP transport models have been developed using data from columns containing porous media, where the breakthrough of nanoparticles from the outlet is measured.4 While informative, breakthrough curve analysis is limited to the time-dependent profile of nanoparticle concentrations either at the column outlet or at intermediate sampling points and thus provides no or little information on the spatial heterogeneity of the entire system.5 However, many systems of interest (aquifers, soils, riverbeds, and engineered drainage systems) are complex, displaying heterogeneity in structure, hydrodynamics, geochemistry, and microbiology r 2011 American Chemical Society

throughout. A range of novel methods have already been developed to spatially resolve NP transport, including fluorescence imaging5 and scanning optical fiber fluorescence profilers.6 Fluorescence imaging, however, relies on sufficient photon penetration of thin, translucent matrices,5 while fluorescence profilers utilize a number of fiber optic detectors inside the column from which a 2D transport profile must still be inferred. An alternative approach, magnetic resonance imaging (MRI), has significant capacity to noninvasively investigate transport processes inside porous media. While most renowned for its use in medicine and biological sciences, MRI has already been used in hydrogeology research to measure water velocities, characterize porous media properties such as grain size and porosity, determine water and nonaqueous phase liquid distribution, and evaluate Received: April 14, 2011 Accepted: November 18, 2011 Revised: November 8, 2011 Published: November 18, 2011 360

dx.doi.org/10.1021/es2012726 | Environ. Sci. Technol. 2012, 46, 360–366

Environmental Science & Technology

ARTICLE

conditions.13 Commercially available MRI compatible NPs,14 positively charged Molday ION, C6Amine (functionalized with surface amine groups) and negatively charged Molday ION, Carboxyl terminated (functionalized with surface carboxyl groups), were used in this study. Both are 35 nm in diameter and are composed of an iron-oxide core surrounded by an organic polymer coating. As the nanoparticles were iron oxide based, Fe was used as a proxy for nanoparticle concentration (for both nanoparticles 0.1 mM Fe is equivalent to ∼5  1015 particles/liter). Both Molday ION NPs are superparamagnetic, making them detectable by MRI. Transport Experiment. The saturated porous column was vertically positioned at the center of the MRI bore. The porous column was first connected to an 18-MΩ water supply and slowly washed with deionized water at a flow rate of 1 mL/min using a HPLC pump (Agilent, 1100 Series). The system was kept under pressure to avoid air bubble formation inside the column. During the column experiment, an inlet solution of Molday ION, Carboxyl (0.1 mM Fe) was first pumped into the packed column at a flow rate of 1 mL/min for approximately 90 min, and MR imaging was performed every 5 min to record the transport of Carboxyl NPs. The packed column was then flushed with deionized water for nearly 3 h in order to remove all the Carboxyl NPs. Here, the column flushing was done at 3 mL/min (nearly 12 pore volumes) to ensure the maximum removal of Carboxyl NPs. An inlet solution of Molday ION, C6Amine (0.1 mM Fe) was then pumped into the packed column at flow rate of 1 mL/min for approximately 90 min, and MR imaging was performed every 5 min to record the transport of C6Amine NP. The hydraulic residence time of the column was 47 min at a flow rate of 1 mL/min. MRI. The MRI experiments were performed on a Bruker Avance BioSpec system. Full details on the related MRI hardware are provided in the SI. Here, MRI was used to acquire spatially and temporally resolved T2-weighted images of the column, while nanoparticles were transported through the column. In the absence of Molday ION NPs, the T2 value of 1H nuclei of water molecules at any location inside the column is influenced by the spinspin interactions between nearby 1H nuclei, magnetic field inhomogeneities, and surface relaxation effects caused due to the presence of quartz gravel. However, the presence of Molday ION NPs will alter the T2 value, and the change in the T2 values upon the uptake of Molday ION NPs is then known to be solely due to the presence of NPs. Consequently, T2-weighted images can be used to track the transport of Molday ION NPs inside the column. T2weighted images acquired before and after introducing NPs can also be used to quantify concentrations of Molday ION NPs as the change in the transverse relaxation rate has a linear relationship to the NP concentration. Acquisition of T 2 -Weighted Images. The transport of Molday ION NPs inside the column was imaged by acquisition of T2-weighted images using a two-dimensional Rapid Acquisition Relaxation Enhancement (RARE) pulse sequence. The use of fast imaging methods is desirable as images acquired with long scan durations will result in time averaged concentration images, complicating data analysis and estimation of transport properties. Hence, the RARE imaging sequence14 was utilized in this study, being a spinecho based sequence also helps to reduce susceptibility artifacts. Images were obtained along the column (parallel to the flow direction) in 11 adjacent slices (Figure 1). There were no gaps between slices and slices were excited in interleaved mode. Details on imaging parameters are provided in the SI. Determination of the Relaxivity Constant (R) of Nanoparticles. Molday ION nanoparticles are classified as a darkening

transport mechanisms and reactive transport processes of solutes inside porous matrices.7 MRI is also able to quantitatively image the transport of molecules and particles that are labeled with a paramagnetic tag. This approach has been used to observe the transport of paramagnetic colloids (∼1.3 μm) through an MRI compatible silica gel matrix.8 In contrast, Amitay-Rosen et al.9 used MRI to image the deposition of nonparamagnetic colloids (112 μm) in a polystyrene bead pack. As the colloids were nonparamagnetic, the displacement of water by colloid deposition was used to image colloid location. We seek to advance on these colloid transport studies by using MRI to investigate the transport of nanoparticles through natural porous media for the first time. While MRI has been utilized to image colloid transport previously, the application of MRI to NP transport imaging is pertinent as NPs have the capacity to exhibit unique transport properties which differ from colloids.10 For example, NPs are susceptible to greater influence of collector surface heterogeneity, will deposit more in primary energy minima compared to colloids, and experience minimal deposition in secondary energy minima.10 NP transport is also likely to be less influenced by straining compared to colloids. In moving from porous media chosen for its MRI compatibility (e.g., polystyrene beads, silica gel grains) to more natural porous media, both the MR imaging and the NP transport become more challenging and complex. In this paper these are overcome by using a coarse grain system and dilute NP solutions. Use of coarse grains lessens the problems associated with MRI of fine grains, namely strong relaxation and susceptibility effects. It furthermore simplifies NP transport by eliminating any effect of straining and subsequent pore clogging altering the flow field, which can occur in finer grained systems.9 In this way the effect on NP transport of interactions with the natural grain surfaces could be investigated by MRI for the first time. Transport in such a coarse grain system has relevance to riverbeds and the gravel matrices of sustainable urban drainage systems (SUDS). NPs are released into these aquatic environments via either direct urban runoff (e.g., rainstorms)4,11 or poorly treated effluent discharges from wastewater treatment plants.11,12 In this study, a negatively charged Molday ION (Carboxyl) and a positively charged Molday ION (C6Amine), which are commercially available superparamagnetic iron oxide based NPs, 35 nm in diameter, were used to investigate the effect of surface charge upon their transport inside a porous column made up with negatively charged rose quartz grains. The quantitative profiles of time varying concentrations of NPs were compared with those of a transport (convection-dispersion) model developed using a finite element modeling software to not only validate the often used homogeneous assumption to predict transport inside porous columns but also to check the accuracy of this MRI method in quantifying NP transport.

’ MATERIALS AND METHODS Porous Column and Nanoparticles. The heterogeneous porous column employed in this study was made up with randomly packed rose quartz grains with a median size of diameter (d50) 3.5 mm (measured using sieve analysis) inside a cylindrical polypropylene column with the dimensions of 4.5 cm internal diameter and 10 cm height (Figure S1 in the SI). Further details on the flow cell are provided in the Supporting Information (SI). Quartz grains were packed only up to the bottom of the top endcap. Quartz grains are negatively charged at experimental pH 361

dx.doi.org/10.1021/es2012726 |Environ. Sci. Technol. 2012, 46, 360–366

Environmental Science & Technology

ARTICLE

the presence of Molday ION NPs at a particular concentration Ci is given by   S TE ¼ exp ð3Þ S0 T2i where S/S0 is the ratio between the signals measured with and without Molday ION NPs, known as the normalized signal, and T2i is the transverse relaxation time with Molday ION NPs at Ci concentration. Therefore, T2-weighted images acquired before and after introducing the NPs can be used to quantify NP concentration using the calibration eq 4, obtained combining eqs 1 and 3      S 1 ð4Þ =TE  1 =T20 3 C¼ ln S0 R

Figure 1. (a) For the slice along the porous column, the T2-weighted image (b) is acquired using two-dimensional RARE sequence. Acquired T2-weighted images with and without the NPs were used to calibrate the NP concentration map (c) using eq 4. The gray scale indicates the NP concentration in mM Fe. (d) The two-dimensional axis symmetry geometry used to model NP transport in the porous column.

Modeling the Transport of NPs inside the Porous Column. To determine the transport properties of the porous column and NPs such as coefficients of dispersivity and retardation, the NP concentration data obtained using MRI was compared to those simulated by a convection-dispersion mathematical model of NP transport. The shape of the experimented column is variable in space (Figure 1a), and, therefore, it is not possible to represent the NP transport by a one-dimensional convection-dispersion equation. However, the cylindrical shape of the column enables us to use a two-dimensional axis-symmetric model. The finite element model for NP transport inside the porous column was implemented using COMSOL Multiphysics 3.5a. The model domain, Ω, is shown in Figure 1d, which was determined from the actual column geometry (Figure S1 in the SI). By assuming the column as homogeneous with constant porosity, the transport of NP with retardation was defined by17

contrast agent acting through the T2 relaxation process,15 and thus their presence inside the porous column causes a concentration dependent linear increase in the transverse relaxation rates (1/T2) of the surrounding 1H nuclei as described in eq 1   1 1 1 ½C ¼  ð1Þ R T2i T20 where T20 is the relaxation time in the absence of NP, T2i is the relaxation time in the presence of NP, [C] denotes the concentration of the NP, and R is the relaxivity constant of the NP. The relaxivity constants (R) of the NPs in water were measured using five NP solutions at different NP concentrations prepared for both Molday ION Carboxyl and C6Amine NPs. T2 values of each sample were measured using multiple spin echo pulse sequence with the following imaging parameters; echo time (TE) 9.2 ms, repetition time (TR) 15000 ms, and 200 echoes with two signal averages. Then, plots of (1/T2i  1/T20) versus NP concentration for both Carboxyl and C6Amine NPs were plotted separately, and the relaxivity constants were determined by the linear curve fit method. The limit of detection for this method is approximately 1  103 mM Fe. Calibration of Nanoparticle Concentration Using T2-Weighted Images. The signal intensity, S obtained using a RARE pulse sequence is given by16     TR TE S ¼ S0 1  exp exp T1 T2

∂Cðx, yÞ ¼ ∇:½D∇Cðx, yÞ  U:∇Cðx, yÞ; x, y ∈ Ω ∂t ! DT 0 D¼ 0 DL

R

U ¼

u0 0v

ð5Þ ð6Þ

! ð7Þ

where C is the NP concentration, DT is the transverse dispersion coefficient, DL is the longitudinal dispersion coefficient, u is the transverse pore velocity, v is the longitudinal pore velocity, and R is the retardation coefficient. For negligible molecular diffusion (discussion on assumption of negligible diffusion is provided in the SI), lateral and transverse dispersion coefficients can be defined as



ð2Þ

where S0 is the maximum signal intensity, TR denotes the repetition time (the time interval between two successive excitation pulses), TE is the echo time (the time interval between the excitation and signal readout center), T1 is the longitudinal relaxation time, and T2 is the transverse relaxation time. As shown in eq 2, the change in the signal intensity due to the presence of paramagnetic ions is complex and might depend on both T1 and T2 relaxation times. However, the RARE pulse sequence can be used to produce T2-weighted images with carefully chosen image sequence parameters, such as a sufficient recycle time between scans (TR), fairly long echo time (TE) and TR much greater than T1. In a T2-weighted image, the signal decay due to

DL ¼ α L v

ð8Þ

DT ¼ α T u

ð9Þ

where αL is the longitudinal dispersivity, and αT is the transverse dispersivity of the porous column. The discrete change in diameter of the column due to the inlet cap (Figure 1b), meant that the mean flow velocity undergoes a fairly rapid transition (Figure S2 in the SI). We found that the pressure gradients that arise from solving Darcy’s law were not capable of driving nanoparticles into the corners of the changing 362

dx.doi.org/10.1021/es2012726 |Environ. Sci. Technol. 2012, 46, 360–366

Environmental Science & Technology

ARTICLE

geometry interface (Figure 1), which was observed in the experiment. This had the knock on effect of raising the simulated concentrations in the center of the column. Thus the Brinkman extension to Darcy flow for transitional flows was employed, and the modest viscous effects, where the flow velocity drops, ensured that the nanoparticles reached the corners of the changing geometry interface. Hence, the velocity field inside the column was first determined by solving the momentum balance for NP transport using the Brinkman equation18 and the continuity equation η η  ∇2 U þ ∇P ¼  U ϕ k

ð10Þ

∇:U ¼ 0

ð11Þ

where U is the velocity, η is the viscosity of the NP solution, k is the permeability of the porous column, ϕ is the porosity of the packed column, and P is the pressure. The permeability value of 1  107 m2 representative of the column made up of unconsolidated well sorted quartz gravel19 and the average porosity value of 0.4 estimated using the MRI measurements (Figure S3 in the SI) were used in the model as known parameters. Details on porosity estimation using MRI are provided in the SI. The evaluated velocity field was then used in the convectiondispersion equation (eq 5) to determine the transport of NPs. The boundary conditions employed during the simulation are given in Table S4 in the SI. The aim was to capture the movement of the NP front and the longitudinal gradients in concentration. Thus, transverse dispersivity was assumed high, and the longitudinal dispersivity of NPs was calibrated. The concentrations measured by MRI displayed almost no transverse gradients, which bears out this assumption. First the model was used to estimate the longitudinal dispersivity of the porous column using the quantitative concentration data of Carboxyl NPs by assuming there is no retardation (i.e., R = 1) as both rose quartz and Carboxyl NPs are negatively charged. The estimated effective dispersivity of the column was then used in the model to estimate the retardation coefficient for the C6Amine NPs as positively charged C6Amine NPs are expected to be retarded inside the column made up of negatively charged rose quartz. Coefficients of dispersivity and retardation were estimated by calibrating the model using respectively the averaged concentration profiles of Carboxyl and C6Amine NPs along the flow axis (y axis), extracted at seven discrete time points (5, 10, 15, 25, 40, 60, and 80 min). The averaged concentration profiles (averaged over the xz plane) of the NPs along the flow axis were determined by averaging the spatially resolved two-dimensional concentration data of NPs from all eleven slices (Figure 1a) and within the area shown by white dotted lines (Figure 1c). The concentration data were averaged only within the region shown in Figure 1c in order to avoid any edge effect that the change in the geometry of the porous column could cause. Here, the dispersivity and retardation coefficients were estimated using a Golden search algorithm in MATLAB, which called the COMSOL model as a subroutine; where the objective function was the sum of square errors between observed and simulated concentrations. The optimum values of coefficients were estimated at the minimum value of this objective function.

Figure 2. Variation in the change of transverse relaxation rates (1/T2i  1/T20) with respect to Molday ION Carboxyl and C6Amine NP concentration. (Concentration of nanoparticles is shown as concentration of Fe.)

It should be noted here that negatively charged quartz grains may cause permanent attachment (deposition) of positively charged C6Amine NPs. However, the retardation coefficient, R, calculated above assumes nonpermanent attachment. Hence, a permanent attachment (deposition) term8 was explicitly incorporated into the model (eq 12) to investigate whether any observed discrepancies between the experimental data and model may be due to permanent attachment R

∂Cðx, yÞ ¼ ∇:½D∇Cðx, yÞ  U:∇Cðx, yÞ  kC; x, y ∈ Ω ∂t ð12Þ

where k is the deposition rate constant. Deposition rate constant and coefficient of retardation for C6Amine NPs were estimated by calibrating the new model (eq 12) using averaged concentration profiles of C6Amine NPs as described above.

’ RESULTS The Relaxivity Constant (R) of Molday ION NPs. The change in transverse relaxation rates (1/T2i  1/T20) has a linear relationship with the concentration of the NPs as shown in Figure 2. From the slope of Figure 2, the relaxivity constant R, for Carboxyl and C6Amine NPs, are determined as 109 mM1 s1 and 81 mM1 s1, respectively. Transport of Molday ION Nanoparticles Inside Porous Column. The transport of Molday ION NPs into the porous column was recorded by T2-weighted images. The presence of Molday ION NPs is shown as less intense (darker) regions in a T2-weighted image as they reduce the MRI signal, and thus the transport of Carboxyl NPs is shown by expansion of the darker region up into the column (Figure 3a to c). The quantitative concentration maps of Carboxyl NPs calibrated using eq 4 are shown in Figure 3d to f. Here, the expansion of the brighter region up into the column illustrates the transport of NPs. Figure 3g to i shows the convection-dispersion model generated with the estimated dispersivity using Carboxyl NP concentration data. The averaged concentration profiles of both Carboxyl and C6Amine NPs along the flow direction at five different selected time intervals are shown in Figure 4. Mass balance for the Carboxyl NPs comparing the known amount of nanoparticles pumped into the system with the total concentration shown in the concentration profiles (i.e., determined from integration of concentration profiles) showed good agreement (e.g., only 5% 363

dx.doi.org/10.1021/es2012726 |Environ. Sci. Technol. 2012, 46, 360–366

Environmental Science & Technology

ARTICLE

Figure 5. Transport of Molday ION Carboxyl NPs inside the porous column is compared to the model data in order to estimate the coefficient of dispersivity at time intervals of 5, 10, 15, 25, 40, 60, and 80 min. Symbols represent experimental data, and solid lines represent model data. (Concentration of nanoparticles is shown as concentration of Fe.)

from pump rate (e.g., lower by 7% at 25 min). The observation that the total NP concentration from the concentration profiles is lower for C6Amine compared to Carboxyl NPs is consistent with C6Amine NPs attaching to grain surfaces (as NPs removed from the water filled pores are not detected by MRI here). Comparison of Experimental Data with Model Data. Figure 5 shows the comparison of averaged concentration profiles of Carboxyl NPs along the flow direction, with the model data along the transect shown in Figure 3i. A dispersivity coefficient of 3.5  103 m for the column was estimated by calibrating the model using concentration profiles, giving a goodness-of-fit value (R2) of 0.97. Figure 6a and b shows the comparison of averaged concentration profiles of C6Amine NPs along the flow direction, with the model data along the transect shown in Figure 3i. A retardation coefficient (nonpermenant attachment; eq 5) of 1.27 for the C6Amine NP was estimated by calibrating the model using concentration profiles, giving a goodness-of-fit value (R2) of 0.9. When permanent attachment (deposition) was incorporated into the model, a deposition rate constant of 1.09  104 and retardation coefficient of 1.02 for the C6Amine NP was estimated, giving an improved goodness-of fit value (R2) of 0.95.

Figure 3. (a to c) T2-weighted images, (d to f) calibrated concentration maps of Carboxyl NP transport inside the column, and (g to i) convection-dispersion model results at selected time intervals of 5, 25, and 80 min, respectively. The gray scale indicates the NP concentration in mM Fe. NP transport direction is from bottom to top.

Figure 4. Averaged concentration profiles of both Carboxyl and C6Amine NPs, along the flow direction at selected time intervals of 5, 10, 25, 40, and 80 min. (Concentration of nanoparticles is shown as concentration of Fe.)

higher total nanoparticle concentration calculated from concentration profiles compared to that calculated from pump rate after 25 min), thus supporting the MRI data set. In comparison, mass balance for the C6Amine NPs showed that total NP concentration from the concentration profiles was lower than that calculated

’ DISCUSSION In this study, MRI was successfully used to quantitatively measuring the time-varying, spatially distributed concentrations of Molday ION Carboxyl and C6Amine nanoparticles as they were transported into a porous column made up with rose quartz grains. As shown in Figure 4, the averaged concentration profiles for both Molday ION NPs clearly illustrates that transport of C6Amine NP within the porous column was slower compared to that of Carboxyl NP. This is most likely due to the effect of transport inhibition by adsorption of positively charged C6Amine NP onto the negatively charged quartz grains. As illustrated in Figure 5, the concentration profiles of the model match well with the averaged concentration profiles quantified using MRI, illustrating the transport of Carboxyl NP inside the column was without retardation (no retardation was used when fitting the model). The lack of retardation results from electrostatic repulsion between negatively charged Carboxyl NP and negatively charged rose quartz. The estimated dispersivity of 364

dx.doi.org/10.1021/es2012726 |Environ. Sci. Technol. 2012, 46, 360–366

Environmental Science & Technology

ARTICLE

Figure 6. Transport of Molday ION C6Amine NPs inside the porous column is compared to the model data modeled with (a) retardation and (b) retardation and deposition in order to estimate the coefficient of retardation and deposition rate constant at time intervals of 5, 10, 15, 25, 40, 60, and 80 min. Symbols represent experimental data, and solid lines represent model data. (Concentration of nanoparticles is shown as concentration of Fe.)

the experimented porous column was 3.5  103 m, which is similar to the experimental results published elsewhere. Khrapitchev and Callaghan20 pool data from a variety of previous studies to derive a relationship between dispersivity and characteristics of porous media. For a packed spherical bead with diameter equal to the median diameter (d50) of the rose quartz gravels used here their relationship yields a value of 2.5  103 m. This corroborates the accuracy of this MRI approach in quantifying NP concentrations inside coarse-grained porous media. When the C6Amine results were modeled using nonpermanent attachment, an estimated effective retardation coefficient of 1.27 was generated, suggesting that transport of C6Amine NP was moderately retarded by the rose quartz grains. However, there were discrepancies between this model and the experimental data (goodness-of-fit, R2 of 0.9) indicating the transport of C6Amine inside the porous column cannot be explained purely by this model (Figure 6a). Consequently, a permanent attachment (deposition) term was incorporated into the model, generating a much better fit (R2 of 0.95) (Figure 6b). This indicates that electrostatically favorable conditions have caused permanent attachment of C6Amine NPs inside the column. When electrostatic conditions are favorable, permanent attachment of NPs onto mineral grains is known to occur.21 Moreover, when deposition is incorporated into the model, the retardation coefficient (nonpermanent attachment) drops from 1.27 to 1.01, illustrating the importance of deposition. A number of studies have reported unique features of nanoparticle transport which differ from larger colloids. For example, deposition in the secondary energy minima (which can occur when particle and collector have similar charges) is less common for NPs than for colloids, as the secondary energy minima is smaller for smaller particles.10,22 This observation may explain transport characteristics seen in this study. The carboxyl terminated NPs and rose quartz have similar surface charge (both negative), thus some deposition in secondary energy minima could be expected. However, no retardation (and thus no attachment) was observed whatsoever, indicating no deposition in secondary minima. This may well reflect the reduced frequency of NP deposition in secondary energy minima compared to colloids. Deposition in primary energy minima can occur when particle and collector

grains have similar charges, providing the energy barrier against attachment is overcome. Although NPs experience much smaller energy barriers to attachment than colloids,10 the lack of retardation for carboxyl terminated NPs suggests these energy barriers were not overcome here (not to an observable extent). In contrast, an absence of energy barriers when particle and collector exhibit opposite charges ensures deposition in the primary energy minima and attachment is permanent. Certainly here, the C6amine transport model required permanent deposition to generate an acceptable fit (and indeed, permanent deposition accounted for most of the attachment). In this regard the C6amine transport behaved in a manner expected for colloids and nanoparticles. In this study, mass balance calculations on the MRI derived concentration profiles suggest attached NPs are not detected (only those in the pore water are detected). We suggest here that this is due to small concentrations of paramagnetic impurities in natural media, which induce enhanced relaxation rates in 1H nuclei in contact with the mineral surface.23 These enhanced relaxation rates are already similar to the enhanced relaxation rate that our NPs induce. Consequently, NPs in contact with water at the mineral surface will have no detectable impact on relaxation rate (the relaxation rate is already rapid), thus we do not detect attached NPs. This contrasts Baumann and Werth8 where attached paramagnetic colloids were detected. This difference is likely due to the use of artificial porous media (fine grained silica gel) by Baumann and Werth,8 which likely generated much smaller surface relaxation effects due to lack of paramagnetic impurities. Overall, this study illustrates the suitability of this approach in water research to spatially resolve the transport behaviors of different NPs inside saturated coarse-grained systems. The spatially and temporally resolved data was successfully described using reaction transport models, giving goodness-of-fit values of 0.95 or better. This illustrated the robustness of this approach, and thus more complex systems requiring heterogeneous modeling approaches could now be explored in future. Therefore, the ability of MRI to image NP transport inside packed columns will enable us to build more complex and heterogeneous packed columns, and thus unravel how these heterogeneities impact NP transport. Indeed, a wide range of paramagnetic and superparamagnetic NPs are commercially available to explore the 365

dx.doi.org/10.1021/es2012726 |Environ. Sci. Technol. 2012, 46, 360–366

Environmental Science & Technology

ARTICLE

impact of parameters such as molecular mass, shape and surface charge on their transport. While the current setup utilizes a typical vertical packed column, the system could easily be modified into a horizontal arrangement similar to a flume, with NP dosed water flowing over the top of the gravel matrix. This would enable better simulation of NP transport in riverbeds. While the current experimental setup successfully imaged NP transport through coarse grained quartz, as the system becomes increasingly complex, both imaging and interpretation can become more challenging. In particular, MR signal losses and artifacts in MR images are caused respectively by relaxation and susceptibility effects when high levels of paramagnetic impurities occur in the porous media. Natural sands and gravels often contain high enough levels of paramagnetic ions (e.g., Fe and Mn) to induce these effects, thus this imaging approach is generally limited to MRI compatible natural media. Complexities such as biofilms and complex variations in grain size can also make interpretation challenging. In this study a simple concentration calibration based upon NPs in water was used as the image voxels were smaller than the pore-size, and hence NP concentrations were determined from voxels that predominantly contained only water and NPs. In more complex systems, variations in small grain sizes and the presence of biofilms will make deriving concentrations more complex as each voxel will contain a potentially unique ratio of water, grain, and/or biofilm. Thus more advanced calibration approaches will be required.

water-saturated porous media. Environ. Sci. Technol. 2008, 42 (10), 3588–3594. (5) Bridge, J. W.; Banwart, S. A.; Heathwaite, A. L. Noninvasive quantitative measurement of colloid transport in mesoscale porous media using time lapse fluorescence imaging. Environ. Sci. Technol. 2006, 40 (19), 5930–5936. (6) Shang, J. Y.; Liu, C. X.; Wang, Z. M.; Wu, H.; Zhu, K. K.; Li, J. A.; Liu, J. In-Situ Measurements of Engineered Nanoporous Particle Transport in Saturated Porous Media. Environ. Sci. Technol. 2010, 44 (21), 8190–8195. (7) Werth, C. J.; Zhang, C. Y.; Brusseau, M. L.; Oostrom, M.; Baumann, T. A review of non-invasive imaging methods and applications in contaminant hydrogeology research. J. Contam. Hydrol. 2010, 113 (14), 1–24. (8) Baumann, T.; Werth, C. J. Visualization of colloid transport through heterogeneous porous media using magnetic resonance imaging. Colloids Surf., A 2005, 265 (13), 2–10. (9) Amitay-rosen, T.; Cortis, A.; Berkowitz, B. Magnetic resonance imaging and quantitative analysis of particle deposition in porous media. Environ. Sci. Technol. 2005, 39 (18), 7208–7216. (10) Petosa, A. R.; Jaisi, D. P.; Quevedo, I. R.; Elimelech, M.; Tufenkji, N. Aggregation and deposition of engineered nanomaterials in aquatic environments: role of physicochemical interactions. Environ. Sci. Technol. 2010, 44 (17), 6532–8549. (11) Kiser, M. A.; Westerhoff, P.; Benn, T.; Wang, Y.; Perez-Rivera, J.; Hristovski, K. Titanium Nanomaterial Removal and Release from Wastewater Treatment Plants. Environ. Sci. Technol. 2009, 43 (17), 6757–6763. (12) Limbach, L. K.; Bereiter, R.; Mueller, E.; Krebs, R.; Gaelli, R.; Stark, W. J. Removal of oxide nanoparticles in a model wastewater treatment plant: Influence of agglomeration and surfactants on clearing efficiency. Environ. Sci. Technol. 2008, 42 (15), 5828–5833. (13) Jada, A.; Akbour, R. A.; Douch, J. Surface charge and adsorption from water onto quartz sand of humic acid. Chemosphere 2006, 64 (8), 1287–1295. (14) Henning, J.; Naurth, A.; Friedburg, H. RARE Imaging: A fast imaging method for clinical MR. J. Magn. Reson. Med. 1986, 3 (6), 823–833. (15) BioPAL. http://www.biopal.com. (16) Haacke, E. M.; Brown, R. W.; Thompson, M. R.; Venkatesan, R. Magnetic Resonance Imaging: Physical Principles and Sequence Design; Wiley-Liss: 1999. (17) Fetter, C. W. Applied Hydrogeology; Prentice Hall: 1994. (18) Pop, I., Ingham, D. B. Transport Phenomena in Porous Media II; A Pergamon Title 2002. (19) Bear, J. Dynamics of Fluids in Porous Media; Dover, 1972. (20) Khrapitchev, A. A.; Callaghan, P. T. Reversible and irreversible dispersion in a porous medium. Phys. Fluids 2003, 15 (9), 2649–2660. (21) Darlington, T. K.; Neigh, A. M.; Spencer, M. T.; Nguyen, O. T.; Oldenburg, S. J. Nanoparticle characteristics affecting environmental fate and transport through soil. Environ. Toxicol. Chem. 2009, 28 (6), 1191–1199. (22) Hahn, M. W.; Abadzic, D.; Melia, C. R. Aquasols: on the role of secondary minima. Environ. Sci. Technol. 2004, 38 (22), 5915–5924. (23) Kleinberg, R. L.; Kenyon, W. E.; Mitra, P. P. Mechanism of NMR relaxation of fluids in rock. J. Magn. Reson. 1994, 108 (2), 206–214.

’ ASSOCIATED CONTENT

bS

Supporting Information. Details on the flow cell, MRI hardware, imaging parameters, rapid transition in the mean flow velocity, porosity variation along the column, model boundary conditions, and assumption of negligible diffusion in convectiondispersion model. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +44 (0)141 330 5474. Fax: +44 (0)141 330 4894. E-mail: [email protected]. Corresponding author address: School of Geographical and Earth Sciences, Gregory Building, University of Glasgow, Glasgow G12 8QQ, United Kingdom.

’ ACKNOWLEDGMENT This work was funded by a Lord Kelvin and Adam Smith Scholarship, University of Glasgow, and the Natural Environment Research Council [grant number NE/G010269/1]. We thank Jim Mullen for his assistance with the MRI experiments. ’ REFERENCES (1) Handy, R. D.; von der Kammer, F.; Lead, J. R.; Hassellov, M.; Owen, R.; Crane, M. The ecotoxicology and chemistry of manufactured nanoparticles. Ecotoxicology 2008, 17 (4), 287–314. (2) Theron, J.; Walker, J. A.; Cloete, T. E. Nanotechnology and water treatment: Applications and emerging opportunities. Crit. Rev. Microbiol. 2008, 34 (1), 43–69. (3) Dawson, T. L. Nanomaterials for textile processing and photonic applications. Color. Technol. 2008, 124 (5), 261–272. (4) Wang, Y. G.; Li, Y. S.; Fortner, J. D.; Hughes, J. B.; Abriola, L. M.; Pennell, K. D. Transport and retention of nanoscale C-60 aggregates in 366

dx.doi.org/10.1021/es2012726 |Environ. Sci. Technol. 2012, 46, 360–366