ARTICLE pubs.acs.org/Langmuir
Transport and Retention of TiO2 Rutile Nanoparticles in Saturated Porous Media under Low-Ionic-Strength Conditions: Measurements and Mechanisms Gexin Chen,*,† Xuyang Liu,† and Chunming Su‡ †
National Research Council Resident Research Associate at the U.S. Environmental Protection Agency, 919 Kerr Research Drive, Ada, Oklahoma 74820, United States ‡ Ground Water and Ecosystems Restoration Division, National Risk Management Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency, 919 Kerr Research Drive, Ada, Oklahoma 74820, United States
bS Supporting Information ABSTRACT: The mechanisms governing the transport and retention kinetics of titanium dioxide (TiO2, rutile) nanoparticle (NP) aggregates were investigated in saturated porous media. Experiments were carried out under a range of well-controlled ionic strength (from DI water up to 1 mM) and ion valence (NaCl vs CaCl2) comparable to the low end of environmentally relevant solution chemistry conditions. Solution chemistry was found to have a marked effect on the electrokinetic properties of NP aggregates and the sand and on the resulting extent of NP aggregate transport and retention in the porous media. Comparable transport and retention patterns were observed for NP aggregates in both NaCl and CaCl2 solutions but at much lower ionic strength with CaCl2. Transport experimental results showed temporal and spatial variations of NP aggregate deposition in the column. Specifically, the breakthrough curves displayed a transition from blocking to ripening shapes, and the NP retention profiles exhibited a shift of the maximum NP retention segment from the end toward the entrance of the column gradually with increasing ionic strength. Additionally, the deposition rates of the NP aggregates in both KCl and CaCl2 solutions increased with ionic strength, a trend consistent with traditional DerjaguinLandauVerwey Overbeek (DLVO) theory. Upon close examination of the results, it was found that the characteristics of the obtained transport breakthrough curves closely followed the general trends predicted by the DLVO interaction-energy calculations. However, the obtained NP retention profiles were found to deviate severely from the theory. We propose that a NP aggregate reconformation through collision between NP aggregates and sand grains reduced the repulsive interaction energies of NPNP and NPsand surfaces, consequently accelerating NP deposition with transport distance and facilitating approaching NP deposition onto NPs that had already been deposited. It is further suggested that TiO2 NP transport and retention are determined by the combined influence of NP aggregate reconformation associated with solution chemistry, travel distance, and DLVO interactions of the system.
1. INTRODUCTION Titanium dioxide (TiO2) nanomaterials are widely used in a variety of applications and commercial products including photocatalysts, photovoltaics, sunscreens, cosmetics, coatings, paints, and pigments.1,2 The annul production of TiO2 nanomaterials is expected to expand exponentially from a current 40 000 metric tons to an upper bound of ∼2.5 million metric tons by 2025 in the United States alone.3 Given such an expansive production and use of these materials, some of the TiO2 nanomaterials utilized in our household and industrial commodities will be released into the natural aquatic environment.4 As a matter of fact, TiO2 nanomaterials have recently been reported in actual municipal wastewater treatment plant effluents at concentrations of 10100 μg/L Ti5; these effluents are discharged into surface waters. There may be unintended environmental consequences for such release and accumulation of engineered TiO2 nanomaterials in the environment because accumulated evidence has shown the adverse exposure effects on aquatic organisms r 2011 American Chemical Society
including microbes, algae, invertebrates, and fish.7 Hence, it is critical to understand the mechanisms governing the transport and deposition of these nanomaterials in porous media so as to thoroughly evaluate the mobility and persistence of these materials in aquatic environments because their bioavailability and toxicity will inevitably be governed by their fate and transport in the environment.6 A limited number of transport and deposition studies of engineered TiO2 NPs have been conducted by utilizing welldefined porous media packed in columns.813 These studies showed that TiO2 NP transport and retention in saturated porous media differed and depended on the background solution chemistry (pH and ionic strength), flow velocity, presence of surfactants, TiO2 NP characteristics and concentration, and Received: January 19, 2011 Revised: March 9, 2011 Published: March 29, 2011 5393
dx.doi.org/10.1021/la200251v | Langmuir 2011, 27, 5393–5402
Langmuir porous media properties. For instance, high TiO2 NP retention was observed in packed sand columns regardless of NP concentration and flow velocity.8 In addition, the observed nonmonotonic NP retention profiles suggest spatially dependent NP deposition in the column, which indicates that distinct processes likely took place at different column depths.14 Existing kinetic models based on classical colloid filtration theory failed to predict the NP transport and retention.8 The limited transport of TiO2 NPs without surface modification was also reported in clean porous media, especially as the pH of NP suspensions approached the point of zero charge of TiO2 NPs, which facilitated NP aggregation.12,13 In these studies, the concurrence of aggregation and the deposition of NPs in the column was often noticed because prior to transport experiments, the prepared NP suspensions were not stable under the conditions tested and thus the influence of aggregation was coupled to that of deposition. As a consequence, the aggregation process increased TiO2 NP retention.9,13 In contrast, recent studies illustrated varying extents of TiO2 NP retention in sandy porous media relying on background solution pH, ionic strength (IS), and the presence of surfactant.9,13 Moreover, the stability and mobility of TiO2 NPs were dramatically enhanced under conditions of evaluated pH or in the presence of surfactants that are likely more environmentally relevant. Indeed, a TiO2 NP aggregation study suggests that NP dispersions were often stable under environmentally relevant conditions of pH, ionic strength, and the presence of natural organic matter.15 It is important to obtain insight into how stable NP aggregates interact with each other and with porous media while being transported. Therefore, a mechanistic understanding of the TiO2 NP transport and deposition behavior in porous media under conditions of a reaction-limited deposition regime and the minimization of NP aggregation will allow us to predict the mobility and behavior of this material in an aquatic environment, which have not been explored systematically. NP aggregation studies have demonstrated that bare TiO2 NPs form aggregates to reduce their free interfacial energy resulting from their high surface area once dispersed in an aqueous solution.16,17 TiO2 NP aggregates in suspension tend to exhibit a size distribution and are polydisperse in solution.10,11 The NP aggregates may be considered to be porous fractals with a lower packing density and a much higher surface area than a solid particle with a similar hydrodynamic diameter.2,6 These complex characteristics of TiO2 NPs likely render much more complicated interactions between NPs and the surrounding surfaces. Hence, the transport of TiO2 NP aggregates in porous media may be complicated and different from that of solid particles of a similar size. Interestingly, an examination of TiO2 NP deposition behavior onto silica using the quartz crystal microbalance (QCM) technique18 revealed that the interactions between the NPs and silica surface were in qualitative agreement with the traditional DerjaguinLandauVerweyOverbeek (DLVO) theory, which is extensively applied to predict colloidal stability.19 Furthermore, a micromodel flow cell study showed that TiO2 NPNP and NPcollector interactions were largely dominated by the NP surface potential.2 Both studies suggest that the physicochemical properties of TiO2 NP aggregates and collector surfaces governed the interfacial interactions and NP deposition. As part of a comprehensive study on thoroughly examining the influence of a variety of environmentally relevant conditions such as solution chemistry, the presence of natural organic matter, and flow velocity on TiO2 NP aggregate fate and transport in the
ARTICLE
subsurface environment, it is the purpose of this study to achieve a better understanding of the mechanisms governing TiO2 NP aggregate transport and retention in porous media under simplified solution chemistry conditions (without the presence of natural organic matter) while minimizing NP aggregation. Specifically, the transport and retention behavior of TiO2 NP aggregates was investigated using a well-defined column packed with silica sand. Experiments were carried out under a range of well-controlled ionic strength and ion valence (NaCl vs CaCl2) comparable to the low end of environmentally relevant solution chemistry conditions (i.e., ISs ranged from DI water to 1.0 mM for NaCl solutions and to 0.06 mM for CaCl2 solutions at nearly neutral pH). Under such conditions, the NP suspensions remained stable and the sensitivity of NP aggregate transport and retention characteristics was allowed to be examined. Finally, combined with extensive complementary TiO2 NP characterization techniques, the detailed mechanisms governing NP transport and deposition were explored.
2. MATERIALS AND METHODS 2.1. TiO2 Suspension Preparation and Characterization. TiO2 NPs (98þ%; Nanostructured & Amorphous Materials, Inc., Houston, TX) with a nominal size of 10 40 nm were used in all experiments. The crystalline composition of the NPs was determined to be a pure rutile phase using X-ray diffraction (XRD) analysis (MiniFlex diffractometer, Rigaku, Inc., Danvers, MA, Supporting Information (SI) Figure S1). A stable aqueous suspension of TiO2 NP aggregates for each experiment was prepared by adding 10 mg of TiO2 nanopowder as received to 500 mL of deionized (DI) water (18.2 MΩ/cm, Nanopure Diamond model D11911, Barnstead International, Dubuque, IA) or a prepared 500 mL electrolyte solution (DI water containing reagentgrade sodium chloride (NaCl, Fisher Scientific, Fair Lawn, NJ) or calcium chloride (CaCl2, Fisher Scientific, Fair Lawn, NJ) to achieve the desired ionic strength). This was followed by stirring the suspension on a magnetic plate for 30 s and immediate bath sonication (Branson 3510R-DTH sonicator, 100 W, 42 kHz, Danbury, CT) for 30 min at room temperature. The suspension was then stirred again for 30 s to ensure a homogeneous suspension. The resulting suspension contained a final TiO2 concentration of 20 mg/L and gave a solution pH of 6.0 ( 0.1. Samples were imaged with a transmission electron microscope (TEM, Philips/FEI CM20 FE TEM/STEM, Hillsboro, OR) after wetting a carbon-supported grid with the suspension and then drying under ambient conditions. TEM imagery confirmed the needlelike morphology of a single TiO2 NP as reported by the manufacturer and aggregate formation in suspensions (SI Figure S2a), similar to what has been reported.17 To understand better the mechanisms governing NP aggregate transport and deposition in porous media, extensive complementary characterization of the TiO2 aggregates was conducted under identical conditions to those tested in the column experiments. The electrophoretic mobility of the TiO2 aggregates in aqueous suspensions was obtained using a ZetaSizer Nano-ZS ZEN3600 analyzer (Malvern Instrument Inc., U.K.) that utilizes laser Doppler velocimetry. To a first approximation, experimentally determined electrophoretic mobility values were converted to zeta (ξ) potentials for each condition using the Smoluchowski equation.19 The latter was used for the electrostatic interaction energy calculations for the TiO2 NPNP and NPsand grain, which are described in detail later. The average TiO2 aggregate size and the intrinsic size distributions in various suspensions were measured using dynamic light scattering (DLS) (ZetaSizer Nano-ZS ZEN3600 analyzer, Malvern Instrument Inc., U.K.). The CONTIN algorithm was used to convert intensity 5394
dx.doi.org/10.1021/la200251v |Langmuir 2011, 27, 5393–5402
Langmuir
Figure 1. (a) Average hydrodynamic diameter of TiO2 NP aggregates in the effluent samples from the column test in DI water and the corresponding relative NP concentrations. (b) Intensity-weighted hydrodynamic diameter distributions of TiO2 NP aggregates of the influent and the two effluent samples as indicated in plot a. autocorrelation functions to intensity-weighted TiO2 aggregate hydrodynamic diameter distributions based on the EinsteinStokes relationship for spherical particles. 20,21 TiO2 NP aggregates in all suspensions with different solution chemistries tested in this study had similar hydrodynamic diameter distributions ranging from 59 to 459 nm, with a mean diameter of 149.6 ( 1.5 nm (Figure 1b). This value was comparable to the previously reported hydrodynamic diameter of rutile NP aggregates (i.e., 137 nm in DI water).17 These intensity-weighted distributions were further converted to numberweighted size distributions, which produced an NP aggregate size ranging from 51 to 220 nm with an average diameter of 84.5 ( 2.4 nm (SI Figure S2b). The latter number-weighted sizes were used for the interaction-energy calculations for the TiO2 NPNP and NPsand grain as described later in this article. All characterization measurements were repeated at least three times using freshly prepared TiO2 NP suspensions for each condition. 2.2. Column Transport Experiments. Ottawa sand (U.S. Silica, Berkeley Springs, WV) was utilized as the packing material for column
ARTICLE
transport experiments. The Ottawa sand was sifted through 250 and 300 μm sieves (USA Standard Testing Sieves, ATM Corp., New Berlin, WI). The fraction of the sand between 250 and 300 μm in size was collected. The resulting average sand diameter was approximately 275 μm. Prior to use, the sand was cleaned thoroughly by the procedure described elsewhere22 to remove any metal oxide and organic contaminants. Briefly, the sand was soaked in 12 N HCl acid (Fisher Scientific) for at least 24 h, washed in DI water, and baked at 800 C for 8 h. The ξ potentials of the clean sand for the range of solution conditions studied in the transport experiments were determined using the method detailed by Tufenkji23 and reported in Table 1. Glass chromatography columns (15 cm long and 2.5 cm inner diameter, Kontes Company, Vineland, NJ) were wet packed uniformly with the clean sand. The porous media were supported on a nylon spectra/mesh filter (pore size, 70 μm; thickness, 70 μm; Spectrum Laboratories, Inc., Rancho Dominguez, CA). The resulting porosity of the porous media was gravimetrically determined to be ca. 0.37. Once packed, the column was operated in a downward direction using a syringe pump (Harvard Apparatus, Holliston, MA) and equilibrated by sequentially pumping approximately 6 pore volumes (PVs) of DI water followed by 10 PVs of background electrolyte solution through the column at a constant Darcy velocity of 5.1 103 cm/s, which corresponds to a calculated Peclet number of 5.4 106 for the NP aggregates in the system.19 A pulse of stable TiO2 NP suspension (20 mg/L) with the same background electrolyte composition was then introduced into the column for 4 PVs, followed by an NP-free background electrolyte solution (ca. 4 PVs) injection. DI water injection was followed for selected experiments to examine the release feature of the retained NPs from the column so as to confirm the role of the potential mechanism of secondary energy minima with the ionic strength of DI water (ca. 105.5 M)24 being 2 orders of magnitude lower than that of the test solutions. Experiments were conducted at ionic strengths ranging from DI water to 1.0 mM for NaCl solutions and to 0.06 mM for CaCl2 solutions until clear trends in NP aggregate transport and retention characteristics were obtained. All column experiments were conducted at least in duplicate. Prior to column experiments with NP suspensions, nonreactive tracer tests (100 mM KNO3) were conducted to assess water flow characteristics and column performance. Column effluent samples were collected in 20 mL liquid scintillation vials using a fraction collector (Retriever 500, Teledyne Isco, Inc., Lincoln, NE). The influent (C0) and effluent concentrations (C) of NPs were determined with a UVvisNIR scanning spectrophotometer (Shimadzu Corporation, Columbia, MD) at a wavelength of 286 nm. Wavelength scan and calibration curves were determined for TiO2 NPs (SI Figure S4). The NP deposition rate coefficient, k, was then experimentally determined for each condition25,26 U C k ¼ ln εL C0 where U is the superficial flow velocity, ε is the packed-bed porosity, L is the length of the packed bed, and C/C0 is the normalized NP concentration exiting the column. To determine the initial NP deposition kinetics, the average value of C/C0 was obtained between 1.86 and 2.00 PVs for each experient, which represents “clean-bed” conditions.25 Following the completion of each transport experiment, the spatial distribution of TiO2 NPs retained in the column was determined. The end fitting was removed, and the porous media were carefully excavated in 1.5 cm increments and transferred into ten 40 mL vials. Excess 1 mM NaOH solution was added to fill the vials. In this low-ionic-strength, high-pH (∼11) solution, highly negatively charged surfaces of both TiO2 NPs and silica sand grains caused the release of retained NPs from the sand surface. After 1 h, the vials containing the sandNP solution mixture were gently shaken to obtain a homogeneous concentration of TiO2 NPs in the supernatant.27 The concentration of NPs in the excess 5395
dx.doi.org/10.1021/la200251v |Langmuir 2011, 27, 5393–5402
Langmuir
ARTICLE
Table 1. Electrokinetic Properties of TiO2 NPs and Sand Grains, Column Mass Balance,a and Energy Barrier Heights as Calculated by DLVO Theoryb for NPSand and NPNP Interactions ionic strengt
TiO2 NP
sand grain
salt type
(mM)
ξ potential (mV)
ξ potential (mV)
Meff (%)
DI water NaCl
0.003 c 0.01
47.53 ( 0.68 43.07 ( 0.38
55.40 ( 0.81 53.11 ( 0.84
99.6 ( 0.1 99.6 ( 0.1
CaCl2
0.10
35.17 ( 0.21
54.11 ( 0.73
99.6 ( 0.2
0.56
31.60 ( 0.10
53.89 ( 0.60
97.6 ( 0.6
Mret (%)
Mtot (%) >99.6 >99.6
3.6 ( 0.9
NPsand
NPNP
primary max (kBT)
primary max (kBT)
153.5 127.5
126.2 97.7
>99.6
88.1
50.3
101.2 ( 0.3
64.7
24.8
0.75
29.77 ( 0.21
53.40 ( 0.47
76.0 ( 1.2
19.0 ( 2.9
95.0 ( 1.3
56.2
17.9
0.80
29.23 ( 0.15
52.87 ( 2.34
26.4 ( 4.6
71.5 ( 3.4
97.8 ( 1.2
53.7
16.2
1.00
28.50 ( 0.26
52.11 ( 0.21
3.0 ( 0.6
93.6 ( 1.5
96.7 ( 2.1
49.2
12.8
0.015
29.93 ( 0.15
28.97 ( 0.21
95.3 ( 0.9
3.8 ( 2.2
99.1 ( 1.4
46.8
43.1
0.0225 0.030
28.53 ( 0.31 25.53 ( 0.57
27.77 ( 0.15 26.20 ( 1.41
63.6 ( 3.5 39.0 ( 2.0
34.4 ( 6.4 59.5 ( 5.9
98.0 ( 2.8 98.5 ( 4.0
41.5 33.9
37.3 27.9
0.036
23.53 ( 0.12
25.33 ( 0.57
15.0 ( 0.6
84.4 ( 1.3
99.4 ( 0.7
29.4
22.4
0.060
20.27 ( 0.29
24.80 ( 1.04
1.1 ( 0.1
96.8 ( 0.8
97.8 ( 0.6
22.7
14.1
a
Here, Meff, Mret, and Mtot refer to the effluent, retained, and total percentages of TiO2 NPs recovered from column experiments, respectively. Meff was determined by integrating beneath the breakthrough curves in Figure 2. Mret was calculated from the experimentally determined retention profiles in Figure 2. b Calculations made by assuming Hamaker constants of 4.5 1020 J for the silicawaterrutile system and 26 1020 J for the rutilewaterrutile system.14 c The ionic strength of DI water was ca. 105.5 M.24
Figure 2. Representative (a, b) breakthrough curves and (c, d) retention profiles for TiO2 NPs at different solution IS values in (a, c) NaCl and (b, d) CaCl2 solutions. aqueous solution was measured with the UV spectrophotometer. The sand samples were then oven dried at 80 C overnight to obtain the dry weight of the solid. The total percentage of recovered NPs (Mtot) was then calculated from percentages in the effluent (Meff) and retained in the column (Mret) and was reported in Table 1.28 Mass recoveries were all between 95.0 and 101.2%. The excellent mass balance suggests that most of the retained TiO2 nanoparticles detached into the aqueous
phase, validating the experimental methodology used to generate the effluent and retention data for the simplified, well-defined system. 2.3. Dynamic Batch Experiments. To assess the attachment affinity of TiO2 NP aggregates to the sand surface in the absence of the influence of pore structures existing in the packed-bed porous media, dynamic batch experiments were conducted by placing 20 g of dry Ottawa sand and 10 mL of a TiO2 NP suspension (20 mg/L) into a 5396
dx.doi.org/10.1021/la200251v |Langmuir 2011, 27, 5393–5402
Langmuir
ARTICLE
40 mL glass scintillation vial under selected background electrolyte conditions. The suspension and sand were allowed to equilibrate for 3 h (transport experiment time span) by slowly rotating the vial at 8 rpm using a rotator (Thermo Scientific, Waltham, MA). The initial and final concentrations of TiO2 NPs in the suspension were determined using the UV spectrophotometer. All experiments were conducted at least in triplicate. Control experiments were performed to quantify the background optical absorbance originating from the sand.
3. RESULTS AND DISCUSSION 3.1. Electrokinetic Properties of TiO2 NPs and Sand. The zeta potentials of the TiO2 NPs and sand grains at different electrolyte concentrations are presented in Table 1 for both NaCl and CaCl2 electrolytes. The results indicated that both the NPs and sand used in this study were negatively charged over the range of IS and pH conditions tested. The values were consistent with the previously reported zeta potential of rutile NPs from the same company (e.g., 43 mV in DI water without pH adjustment vs 47.53 ( 0.68 mV in DI water from our measurements).17 The absolute magnitude of the zeta potentials decreased with an increase in salt concentration (in either NaCl or CaCl2) as expected from the electrostatic double layer compression that occurs in the presence of electrolytes. Table 1 also shows that both the NPs and sand in NaCl solutions possessed more negative zeta potentials than those in CaCl2 solutions even though the ionic strengths in CaCl2 solutions were much lower than those in NaCl solutions. This observation suggested that Ca2þ ions were more effective at screening the surface charge of solid surfaces than Naþ ions.29,30 These zeta potential values were used to calculate the DLVO interaction-energy profiles between the NPNP and the NPsand surfaces, which are presented later. 3.2. TiO2 NP Breakthrough Curves and Retention Profiles. TiO2 NP transport experiments were conducted by passing the NP suspensions downward through vertical columns packed with silica sand. The representative results of transport experiments based on different electrolyte concentrations and ion valences are presented in Figure 2a,b. The experimental data were displayed as NP breakthrough curves (i.e., the fraction of the influent NP concentration leaving the packed bed, C/C0, as a function of PVs). Solution chemistry was found to have a marked effect on the extent of TiO2 NP aggregate transport. As shown in Figure 2a, TiO2 NPs were almost completely mobile in NaCl solutions at ionic strength lower than 0.75 mM. In contrast, 93.6% was retained in the porous media as IS gradually increased to 1 mM, which suggested that the NP retention rate increased with increasing IS. A comparison of panels a and b in Figure 2 revealed that comparable NP transport and retention trends were observed in CaCl2 solutions but at much lower ISs. The mass balance results shown in Table 1 indicated that over 97% of the NPs exited from the packed column when the IS was lower than 0.75 mM in NaCl solutions. The NP breakthrough was observed to decrease to 76.0, 26.4, and 3.0% when the IS was increased to 0.75, 0.8, and 1 mM, respectively. Similarly, 95.3, 63.6, 39.0, 15.0, and 1.1% of the injected NPs came out of the column when the IS was gradually increased from 0.015 to 0.06 mM in CaCl2 solutions. Interestingly, Figure 2a,b shows that the reduction in breakthrough NP concentrations associated with increasing IS exhibited a transition from blocking to ripening behavior, indicating temporal variations of the NP deposition rate. The
Figure 3. Deposition rate coefficient (k) as a function of IS for TiO2 rutile NPs. The solid lines are drawn to provide guides to the eye to indicate trends in changing k. Error bars indicate one standard deviation.
ripening shapes were observed to be more pronounced in NaCl solutions than in CaCl2 solutions. TiO2 NP retention percentages and profiles were determined following the completion of transport experiments. Representative retention profiles are shown in Figure 2c,d with respect to the influence of electrolyte IS and ion valence. Consistent with the observations from the breakthrough curves, Table 1 shows that the amount of NPs retained in the column increased with increasing IS for both NaCl and CaCl2 electrolytes. Interestingly, the NP retention profiles in Figure 2c,d showed that the maximum NP retention segment shifted from the end toward the entrance of the column gradually with increasing IS under both electrolyte conditions, suggesting spatial variations in the NP aggregate deposition rate along the column. Figure 3 presents k as a function of the IS and ion valence. As expected, the k values followed the same trend as NP retention in the column experiments. Namely, the k values increased with increasing IS for IS values greater than 0.1 mM in NaCl solutions and for all CaCl2 solutions tested; whereas, the k values were essentially independent of IS for IS less than 0.1 mM in NaCl solutions. Interestingly, comparable increasing k trends were observed in both NaCl and CaCl2 but at lower CaCl2 concentrations relative to those of NaCl, suggesting that the electrostatic double layer interaction played a marked role in the NP aggregate transport and deposition process. 3.3. DLVO Interaction-Energy Calculations. To qualitatively understand the transport and retention trends of the TiO2 NP aggregates in saturated porous media, to a first approximation, DLVO theory was used to calculate the total NPNP and NPsand interaction energies as the sum of van der Waals and electrostatic double layer interactions as the two surfaces approach one other. The NPNP interaction energy was determined by assuming the NPNP system to be formed from a spheresphere interaction; whereas, the NPsand grain interaction energy was calculated by assuming the NPsand system to be formed from a sphereplate interaction. The retarded van der Waals attractive interaction energies for the NPNP (φVDWNN) and NPsand (φVDWNS) systems 5397
dx.doi.org/10.1021/la200251v |Langmuir 2011, 27, 5393–5402
Langmuir
ARTICLE
were then calculated using eqs 1 and 2, respectively31 1 A131 ap1 ap2 5:32D λ ln 1 þ 1 φVDW NN ¼ λ 5:32 6Dðap1 þ ap2 Þ ð1Þ φVDW NS ¼
A132 ap 14D 1 1þ λ 6D
ð2Þ
where ap1 and ap2 in eq 1 refer to the radii of two interacting NP aggregates and ap in eq 2 refers to the radius of the NP aggregate. D is the separation distance, λ is the characteristic wavelength of interaction, usually taken to be 100 nm, and A131 and A132 are Hamaker constants for NPwaterNP (26 1020 J) and NPwatersand (4.5 1020 J), respectively.14 The NP aggregate average diameter of 84.5 nm was used for calculations. The electrostatic double layer interaction energies for the NPNP (φEDLNN) and NPsand (φEDLNS) systems were determined using eqs 3 and 4, respectively32 ( 2ψp1 ψp2 2πap1 ap2 n¥ kB T 2 2 φEDL NN ¼ ðψp1 þ ψp2 Þ ln 2 ψp1 2 þ ψp2 2 ðap1 þ ap2 Þk 1 þ expðkDÞ þ ln½1 expð2kDÞ ð3Þ 1 expðkDÞ ( φEDL NS ¼ πε0 εr ap ðξp þ ξc Þ 2
2
2ξp ξc
ðξp 2 þ ξc 2 Þ 1 þ expðkDÞ þ ln½1 expð2kDÞ 1 expðkDÞ
k¼
e2 ∑ni0 zi 2 ε0 εr k B T
ð4Þ
!1=2 ð5Þ
where n¥ and ni0 are the number concentrations of bulk ions and ion i in the bulk solution, kB is the Boltzmann constant, T is the absolute temperature of the system, ψp1 and ψp2 are the reduced potentials (ψ = zeξ/kBT) of two interacting NP aggregates, ε0 is the permittivity of vacuum, εr is the relative dielectric constant of the medium, ξp and ξc are the electrical potentials of the NP aggregate and the sand, zi is the valence of ion i in bulk solution, e is the electron charge, and κ is the DebyeH€uckel reciprocal length. It is important to be aware that the above calculations have limitations even though they may provide insight into understanding the general trend in NP transport and retention behavior. For instance, the above calculation approaches utilize the Derjaguin approximation, which tends to overestimate the magnitude of van der Waals and electrostatic double layer interactions.33 In addition, the expressions assume perfectly spherical particles interacting with each other or with a flat plate, which is not valid for complex systems. The results for the NPNP and NPsand interaction primary energy barrier calculations are shown in Table 1 and indicate that energy barriers existed between the NPNP and NPsand interactions under all conditions tested, with decreases in the interaction energy barriers concomitant with IS increasing for both electrolyte types. The NPNP interaction-energy barriers were less than those of the NPsand interactions under the same solution chemistry conditions, presumably because of the much
Figure 4. NPsand grain interaction-energy profiles generated with DLVO theory for TiO2 rutile NP aggregates of different sizes as a function of the separation distance. Surface potentials of NP and sand dispersed in DI water under the conditions given in Table 1 were used for the calculations.
greater Hamaker constant of NPwaterNP (26 1020 J) compared to that of NPwatersand (4.5 1020 J). It became even more considerable with increasing IS, suggesting that increased solution IS favored the NPNP interaction over the NPsand interaction. Furthermore, the differences in the energy barriers between NPNP and NPsand under the same conditions in NaCl solutions were more sensitive to the changes in solution IS relative to CaCl2 solutions. Additionally, the DLVO interaction-energy calculations predicated secondary energy minimum depths of (0.020.05)kBT between the NP aggregates and sand grains upon close approach for an IS of 0.561.0 mM in NaCl solutions and no secondary energy minimum existing in NaCl solutions when the IS was less than 0.56 mM and when CaCl2 solutions were used (data not shown). 3.4. Effect of NP Aggregate Size Distribution. Figure 1b and SI Figure S2b present the inherent size distribution of TiO2 NP aggregates in DI water, clearly indicating the polydisperse nature of TiO2 NP aggregates in suspensions.10,11 To examine the effect of the NP size distribution on the NP transport and retention in saturated porous media, selected effluent samples from the column experiment in DI water were analyzed by DLS. The NP average hydrodynamic diameter and relative concentration (C/C0) of the samples are plotted as a function of the pore volumes and are shown in Figure 1a. The diameter distributions of aggregates in the NP influent suspension and the two effluent samples as indicated in Figure 1a are presented in Figure 1b. Apparently, the fraction of NPs with smaller diameter in the suspension was removed from the aqueous phase and rendered elevated NP average diameters for the effluent in the early stage of the column test (Figure 1a,b). With the depletion of favorable deposition sites in the porous media likely induced by sand surface charge heterogeneity and/or surface roughness under this condition, the NP average diameter in the effluent samples gradually decreased as shown in Figure 1a and eventually reached the same NP average diameter and size distribution as the influent sample, which suggested that all NP aggregates injected passed through the column. This experimental observation demonstrated 5398
dx.doi.org/10.1021/la200251v |Langmuir 2011, 27, 5393–5402
Langmuir greater removal efficiency for NP aggregates with a small diameter from the packed-bed column relative to that of large aggregates under the same physicochemical conditions. This phenomenon can be attributed to two causes. First, the NP transport-dominant mechanism was diffusion as suggested by the small Peclet number. Indeed, a simple calculation of the collector efficiency predicted by the correlation equations proposed by Tufenkji and Elimelech27 suggested that 99.7% of NP mass transfer to the sand surface in these experiments was attributed to diffusion. Hence, smaller NPs would have greater diffusion coefficients on the basis of the EinsteinStokes relationship, which then led to greater mass transfer to the collector surface relative to larger NPs. Second, DLVO interaction energy profiles shown in Figure 4 indicated that for deposition to occur on the collector surface for smaller NPs, a lower energy barrier would have to be overcome. Notably, our observations contrasted strongly with those previously reported for nanoscale fullerene aggregates (nC60)34 and TiO2 NP aggregates.9 Wang et al.34 demonstrated that the nC60 aggregate size remained constant during transport through their columns even when there was significant retention of nC60 and concluded that the retention of nC60 aggregates was not a sizeselective process. This inconsistency may be due to the nature of NP size distributions (a monodispersion of nC60 aggregates10,11 vs a polydispersion of TiO2 aggregates produced in this study). Additionally, a much larger average size of TiO2 NP aggregates that eluted from the column was noticed by Solovitch et al.9 compared to the size of aggregates in the influent solution. The observation strongly indicated aggregate formation in the column under the condition in which they tested, which was confirmed by the aggregation kinetics study conducted under the same condition. 3.5. Evidence of the Dominance of DLVO Interactions on TiO2 NP Aggregate Transport Kinetics. Comparable breakthrough curves of TiO2 NPs at different solution IS values in the presence of NaCl and CaCl2 solutions (Figure 2a,b, respectively) were observed, and they exhibited a transition from blocking behavior at low IS to ripening behavior at elevated IS (0.751.0 mM in NaCl and 0.02250.06 mM in CaCl2). At low IS, the NP deposition dynamics is characterized by the blocking of sand surfaces and declining deposition rates with increasing PVs as displayed in Figures 1a and 2a,b. A similar effect has been reported for the deposition and transport of nC60 NPs in quartz sands.35 These observations may be explained by the limited availability of favorable sites on sand surfaces likely induced by sand surface roughness and heterogeneity for NP aggregate deposition as both the sand and NP aggregates were highly negatively charged under these conditions, suggesting considerable repulsive interactions among NPNP and NPsand as shown in Table 1. The retained NPs then blocked the subsequent deposition of NP aggregates, resulting in a larger fraction of NPs being eluted from the porous media as indicated by increasing effluent concentration.19 This observed blocking effect was not likely to cause an increase in NP effluent concentration through NP detachment as implied by the absence of tailing on all NPs and the nonreactive KNO3 tracer breakthrough curves (data not shown for the tracer breakthrough curves).36,37 A significant increase in TiO2 NP removal with time was noticed at high IS and became even more pronounced with increasing IS as shown in Figure 2a,b. This phenomenon was also observed in traditional colloid filtration studies and is commonly referred to as filter ripening.38,39 The enhancement of NP
ARTICLE
Figure 5. Removal percentage of TiO2 rutile NPs to silica sand as a function of IS in NaCl solutions. Error bars indicate one standard deviation.
removal with time may indicate two possible mechanisms: (1) the NPs deposited on the sand surface altered the surface charge distribution of the sand and rendered more favorable NP deposition conditions as the NP and sand surfaces had different charge properties in the same electrolyte solutions, as demonstrated by zeta potential values in Table 1 and (2) the interaction between approaching NPs and NPs that had been already deposited was more favorable for deposition than the interaction of the approaching NPs with the bare sand surfaces.28 To examine the possibility of the first mechanism hypothesized, dynamic batch experiments were conducted by mixing 10 mL of a TiO2 NP suspension and 20 g of dry sand in NaCl solutions with IS values of 0.56, 0.75, and 1.00 mM. The percentages of NPs removed from the aqueous phase as a result of NP attachment to the sand surface were observed to be 11.2, 18.8, and 30.2% with increasing IS (Figure 5). In contrast, the retention of NPs in the column was much more sensitive to IS because 3.6, 19.0, and 93.6% of TiO2 NPs were retained in the porous medium at the corresponding ISs. The dissimilar sensitivity of NP attachment to the sand surface to IS in dynamic batch experiments and column transport tests then ruled out the first hypothesized mechanism. Additionally, the mechanism of straining was likely not responsible for the dissimilar sensitivity of NP attachment to the sand surface associated with IS between two systems. Empirically, straining has been found to be an important retention mechanism when the critical ratio of particles to median sand grain diameter is greater than 0.001640 or 0.0017.41 Our estimated ratio of the average TiO2 NP aggregate diameter (84.5 nm) to the median sand grain diameter (275 μm) is 0.00031, which suggests that straining did not play a marked role in TiO2 retention in this study. Therefore, the potential NPNP interactions were most likely responsible for the ripening process as observed in Figure 2a,b. Indeed, the interactionenergy barriers calculated between NPs were less than those between NPs and sand under all conditions tested as shown in Table 1, which is likely a common characteristics of TiO2 NP transport and deposition in porous media as the much greater Hamaker constant of NPwaterNP compared to that of NPwatersand, thereby facilitating approaching NP 5399
dx.doi.org/10.1021/la200251v |Langmuir 2011, 27, 5393–5402
Langmuir deposition onto NPs that had already been deposited. A similar ripening process has recently been reported by Solovitch et al. and Godinez et al. for TiO2 NP transport in porous media.9,13 Moreover, the interaction-energy barriers calculated between NPs decreased faster than those between NP and sand with increasing background electrolyte IS (Table 1), which implies that this effect became more significant with increasing IS. This is the case in which NP retention enhancement was accelerated with IS as displayed in Figure 2a,b. Additionally, the difference in magnitude of the two types of interaction energy barriers (NPNP vs NPsand) under the same conditions was more substantial in NaCl solutions than in CaCl2 solutions (Table 1). This would predict that ripening is more pronounced in NaCl solutions than in CaCl2 solutions at high ISs, which is consistent with the shapes of NP breakthrough curves shown in Figure 2a,b. Therefore, TiO2 NP aggregate transport kinetics in porous media followed the general trends in the interaction energy calculations among NPNP and NPsand based on the DLVO theory. 3.6. Deviation of TiO2 NP Aggregate Transport and Retention Behavior from DLVO Interaction Energy Calculations. It is evident that solution chemistry (i.e., IS and ion valence) dominated the TiO2 NP aggregate transport kinetics in this study because the general trend for all column experiments indicated an increase in NP retention with IS. Overall, this retention trend is qualitatively consistent with the DLVO interaction energy calculations shown in Table 1 in that increasing IS reduced the primary energy barrier preventing NP attachment to the sand grain surface in the primary minimum because of electrostatic repulsion. As stated above, the NP transport kinetics also followed the general trends in the interaction-energy calculations for NPNP and NPsand based on the DLVO theory. However, DLVO interaction-energy calculations shown in Table 1 also displayed sizable energy barriers that predict unfavorable conditions for NP attachment to sand. Indeed, a close inspection of the NP retention profiles presented in Figure 2c,d revealed a negligible amount of NPs retained in the inlet segment of the column, even when over 93% of the NPs were retained in the column at the highest ISs tested in both NaCl and CaCl2 solutions (Figure 2c,d). Therefore, other mechanisms besides the diminishing of this primary energy barrier were responsible for the enhancement of the NP retention with increasing solution IS and are discussed below in great detail. The shape of the NP retention profiles presented in Figure 2c, d illustrated that a variety of spatially dependent deposition processes were also occurring during these experiments. Drastically contrasting with the prediction of the monotonic decrease with distance from the top of the column by traditional colloid filtration theory,23 the amount of NP retention monotonically increased with distance at low IS (i.e., below 1.00 mM in NaCl solutions and 0.030 mM in CaCl2 solutions), strongly suggesting that the NP deposition rate was enhanced with increasing transport distance in the column. Moreover, the maximum NP retention segment gradually shifted from the end toward the entrance of the column with IS in both electrolyte solutions at the high IS tested (i.e., 1.00 mM in NaCl solution and greater than 0.03 mM in CaCl2 solutions, respectively) and generated these unusual nonmonotonic profiles of the retained NPs with increasing transport distance (increased and then decreased the retention). Similar distance-dependent NP retention profiles were reported for nC60 NPs transport and deposition in quartz porous media under varying flow conditions.35 Nonmonotonic profiles of the retained TiO2 NPs with increasing transport distance were
ARTICLE
also noticed in sand columns by Choy et al.8 It is therefore likely that the observed spatial variations in the NP deposition rate are due to processes fundamental to NP transport and retention in porous media. The governing mechanisms resulting in these observed spatially dependent NP retention profiles in sand columns are still largely unknown. Porous media factors such as the surface charge heterogeneity42 and surface roughness43 cannot interpret the NP retention profiles mentioned here because these characteristics would have been homogeneously distributed within the packedbed sand columns.44 Furthermore, the detachment and reentrapment of NPs is unlikely because of the absence of tailing on all NPs and the nonreactive KNO3 tracer breakthrough curves as shown in Figure 2a,b.45 In the previous study of TiO2 NP transport in sand columns, the authors proposed that the observed nonmonotonic retention profiles were due to a balance between the migration downgradient of NPs associated with secondary energy minima and the trapping of NPs between the grain surfaces.8 However, the DLVO interaction-energy calculations for the interactions between NPs and sand grains suggested that this mechanism was not likely responsible for the observed spatially dependent retention profiles in this study because no secondary energy minima existed in the CaCl2 solutions or when the ISs were less than 0.56 mM in NaCl solutions; whereas, the magnitude of the calculated secondary energy minima was only (0.020.05)kBT in NaCl solutions when ISs were greater than 0.56 mM, which is more than 1 order of magnitude less than the driving force of NP transport by diffusion (related to kBT).46 Moreover, upon the completion of the transport experiment at 1 mM NaCl, DI water injection was followed to examine the release feature of the retained NPs from the column. A negligible amount of retained TiO2 NPs were released back into the aqueous phase as shown in Figure 2a, which further corroborates the minor contribution of this mechanism. A similar observation was also reported for TiO2 NP transport in porous media without the presence of surfactant in the system.13 An identical conclusion with respect to the mechanism—secondary energy minima—was drawn for the transport and deposition of nC60 NPs in quartz sands by Li et al.35 Figure S2a in the SI shows the porous fractal-like structure of the TiO2 NP aggregates. This complex characteristic of TiO2 NP aggregates likely renders complicated interactions between NPNP aggregates and NPsand surface.2 For instance, the NP aggregate conformation might be rigid at low IS because individual NPs possessed an electrostatic double layer of considerable thickness. With increasing IS, a greater number of salt ions in solution would compress the single NP double layer thickness as suggested by a decrease in the NP electrostatic double layer characteristic length (1/κ),47 which likely allows the individual NPs in the aggregate to reconform through the collisions between NP aggregates and the sand surface when the aggregate is transported in porous media. Figure 1a shows that the NP average diameter in the effluent samples decreased slightly but statistically after reaching the same NP average diameter and diameter distribution as for the influent sample. This noticeable size decrease in the effluent samples is likely due to the NP aggregates’ reconfirmation and formation of a more compact structure. On the basis of the observation that the amount of TiO2 NP retention monotonically increased with increasing transport distance at low IS in both electrolytes in the experiments discussed herein, we postulate that a subtle reconformation of 5400
dx.doi.org/10.1021/la200251v |Langmuir 2011, 27, 5393–5402
Langmuir the NP aggregates occurred by collisions between NP aggregates and sand grains when the aggregates were transported in the column. It altered the NP aggregate surface potentials2 and therefore reduced the interaction-energy barriers for NPNP aggregates and NPsand surface interactions. Consequently, NP aggregates became more “sticky” to each other and/or to the sand grains, which led to an enhanced NP deposition rate with increasing transport distance at low ISs as shown in Figure 2c,d and to the ripening processes at high ISs as observed in Figure 2a, b as well.9 Furthermore, an increase in solution IS favored such a process as greater NP retention was observed with increasing IS under low IS conditions at column depth greater than 5.25 cm. In contrast, negligible NP was retained in the column at a depth of less than 3.75 cm under these conditions, as shown in Figure 2c,d. As IS increased further, the depletion of NPs in the aqueous phase resulted in a decreased NP deposition rate after reaching the maximum retention segment. However, the relationship among NP aggregate reconformation, NP aggregate surface potential alteration, travel distance, and solution IS is unknown and likely complex. Thus, future work is needed to address the relationship among the variables so that a mechanistic model on TiO2 NP transport and deposition in porous media can be developed.
4. CONCLUSIONS The mechanisms controlling the transport and retention kinetics of TiO2 NP aggregates were investigated in saturated porous media. The results presented here indicate that the transport and retention of TiO2 NPs are markedly sensitive to solution IS and ion valence. In addition, temporal and spatial variations of TiO2 NP aggregate deposition along the sand column were observed. The transport and retention characteristics suggest that the NP transport is governed by the combination of NP aggregate reconformation associated with solution chemistry, travel distance, and DLVO interactions. Unlike monodisperse solid individual colloids for which a constant particle deposition rate is usually assumed,23 these temporal and spatial variations in TiO2 NP aggregate deposition induced by NP aggregate reconformation imply that the transport and deposition of TiO2 NPs in porous media is not only a kinetic process but also a dynamic process. This must be taken into consideration when developing reliable NP transport mechanistic models for determining potential exposure to NPs for impact studies.2 It highlights that the traditional practice of utilizing the loglinear extrapolation of discrete measurements of colloid attenuation44 cannot provide the actual NP concentration upon removal of NPs with distance from the source. ’ ASSOCIATED CONTENT
bS
Supporting Information. Additional materials and methods, additional results, and data pertaining to the crystalline phase and morphology of TiO2 NPs, the size distribution, the stability test, the wavelength scan, and the calibration curve of TiO2 NP suspensions. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone (580) 436-8651. Fax (580) 436-8703. E-mail: chen.gexin@ epa.gov.
ARTICLE
’ ACKNOWLEDGMENT This research was funded by the National Nanotechnology Initiative through the U.S. Environmental Protection Agency (EPA). This article has not been subjected to an internal policy review by the U.S. EPA. Therefore, the research results do not necessarily reflect the views of the agency or its policy. We acknowledge Stephanie Burrage for laboratory assistance through the U.S. EPA Environmental Research Apprenticeship Program (ERAP) at the Robert S. Kerr Environmental Research Center. ’ REFERENCES (1) Chen, X.; Mao, S. S. Chem. Rev. 2007, 107, 2891–2959. (2) Guzman, K. A. D.; Finnegan, M. P.; Banfield, J. F. Environ. Sci. Technol. 2006, 40, 7688–7693. (3) Robichaud, C. O.; Uyar, A. E.; Darby, M. R.; Zucker, L. G.; Wiesner, M. R. Environ. Sci. Technol. 2009, 43, 4227–4233. (4) Brar, S. K.; Verma, M.; Tyagi, R. D.; Surampalli, R. Y. Waste Manage. 2010, 30, 504–520. (5) Kiser, M. A.; Westerhoff, P.; Benn, T.; Wang, Y.; Perez-Rivera, J.; Hristovski, K. Environ. Sci. Technol. 2009, 43, 6757–6763. (6) Wiesner, M. R.; Lowry, G. V.; Alvarez, P.; Dionysiou, D.; Biswas, P. Environ. Sci. Technol. 2006, 40, 4336–4345. (7) Scown, T. M.; van Aerle, R.; Tyler, C. R. Crit. Rev. Toxicol. 2010, 40, 653–670. (8) Choy, C. C.; Wazne, M.; Meng, X. G. Chemosphere 2008, 71, 1794–1801. (9) Solovitch, N.; Labille, J.; Rose, J.; Chaurand, P.; Borschneck, D.; Wiesner, M. R.; Bottero, J. Y. Environ. Sci. Technol. 2010, 44, 4897–4902. (10) Lecoanet, H. F.; Wiesner, M. R. Environ. Sci. Technol. 2004, 38, 4377–4382. (11) Lecoanet, H. F.; Bottero, J. Y.; Wiesner, M. R. Environ. Sci. Technol. 2004, 38, 5164–5169. (12) Joo, S. H.; Al-Abed, S. R.; Luxton, T. Environ. Sci. Technol. 2009, 43, 4954–4959. (13) Godinez, I. G.; Darnault, C. J. G. Water Res. 2011, 45, 839–851. (14) Petosa, A. R.; Jaisi, D. P.; Quevedo, I. R.; Elimelech, M.; Tufenkji, N. Environ. Sci. Technol. 2010, 44, 6532–6549. (15) Domingos, R. F.; Tufenkji, N.; Wilkinson, K. J. Environ. Sci. Technol. 2009, 43, 1282–1286. (16) Yang, K.; Lin, D. H.; Xing, B. S. Langmuir 2009, 25, 3571–3576. (17) Mukherjee, B.; Weaver, J. W. Environ. Sci. Technol. 2010, 44, 3332–3338. (18) Fatisson, J.; Domingos, R. F.; Wilkinson, K. J.; Tufenkji, N. Langmuir 2009, 25, 6062–6069. (19) Elimelech, M.; Gregory, J.; Jia, X.; Williams, R. A. Particle Deposition and Aggregation: Measurement, Modeling and Simulation; Butterworth-Heinemann: Oxford, England, 1995; p 441. (20) French, R. A.; Jacobson, A. R.; Kim, B.; Isley, S. L.; Penn, R. L.; Baveye, P. C. Environ. Sci. Technol. 2009, 43, 1354–1359. (21) Phenrat, T.; Kim, H. J.; Fagerlund, F.; Illangasekare, T.; Tilton, R. D.; Lowry, G. V. Environ. Sci. Technol. 2009, 43, 5079–5085. (22) Redman, J. A.; Walker, S. L.; Elimelech, M. Environ. Sci. Technol. 2004, 38, 1777–1785. (23) Tufenkji, N.; Elimelech, M. Langmuir 2004, 20, 10818–10828. (24) Elimelech, M. J. Colloid Interface Sci. 1991, 146, 337–352. (25) Walker, S. L.; Redman, J. A.; Elimelech, M. Langmuir 2004, 20, 7736–7746. (26) Haznedaroglu, B. Z.; Kim, H. N.; Bradford, S. A.; Walker, S. L. Environ. Sci. Technol. 2009, 43, 1838–1844. (27) Tufenkji, N.; Elimelech, M. Environ. Sci. Technol. 2004, 38, 529–536. (28) Kim, H. N.; Bradford, S. A.; Walker, S. L. Environ. Sci. Technol. 2009, 43, 4340–4347. (29) Chen, K. L.; Elimelech, M. Langmuir 2006, 22, 10994–11001. 5401
dx.doi.org/10.1021/la200251v |Langmuir 2011, 27, 5393–5402
Langmuir
ARTICLE
(30) Jiang, X. J.; Tong, M. P.; Li, H. Y.; Yang, K. J. Colloid Interface Sci. 2010, 350, 427–434. (31) Gregory, J. J. Colloid Interface Sci. 1981, 83, 138–145. (32) Hogg, R.; Healy, T. W.; Fuerstenau, D. W. Trans. Faraday Soc. 1966, 62, 1638–1651. (33) Bhattacharjee, S.; Elimelech, M. J. Colloid Interface Sci. 1997, 193, 273–285. (34) Wang, Y. G.; Li, Y. S.; Fortner, J. D.; Hughes, J. B.; Abriola, L. M.; Pennell, K. D. Environ. Sci. Technol. 2008, 42, 3588–3594. (35) Li, Y. S.; Wang, Y. G.; Pennell, K. D.; Abriola, L. M. Environ. Sci. Technol. 2008, 42, 7174–7180. (36) Yan, Y. D. Langmuir 1996, 12, 3383–3388. (37) Johnson, W. P.; Li, X. Q.; Assemi, S. Adv. Water Resour. 2007, 30, 1432–1454. (38) Yao, K. M.; Habibian, M. T.; O’Melia, C. R. Environ. Sci. Technol. 1971, 5, 1105–1112. (39) Elimelech, M.; Omelia, C. R. Environ. Sci. Technol. 1990, 24, 1528–1536. (40) Shen, C. Y.; Huang, Y. F.; Li, B. G.; Jin, Y.Water Resour. Res. 2008, 44, doi: 10.1029/2007WR006580. (41) Bradford, S. A.; Yates, S. R.; Bettahar, M.; Simunek, J.Water Resour. Res. 2002, 38, doi: 10.1029/2002WR001340. (42) Johnson, P. R.; Sun, N.; Elimelech, M. Environ. Sci. Technol. 1996, 30, 3284–3293. (43) Shellenberger, K.; Logan, B. E. Environ. Sci. Technol. 2002, 36, 184–189. (44) Li, X. Q.; Johnson, W. P. Environ. Sci. Technol. 2005, 39, 1658–1665. (45) Bradford, S. A.; Simunek, J.; Walker, S. L. Water Resour. Res. 2006, 42; doi:10.1029/2005WR004805. (46) Omelia, C. R. J. Environ. Eng. 1985, 111, 874–890. (47) Gregory, J. Particles in Water: Properties and Processes; CRC Press: Boca Raton, FL, 2006; p 173.
5402
dx.doi.org/10.1021/la200251v |Langmuir 2011, 27, 5393–5402
