Transport and deposition of stabilized engineered silver nanoparticles in water saturated loamy sand and silty loam

Science of the Total Environment 535 (2015) 102–112

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Transport and deposition of stabilized engineered silver nanoparticles in water saturated loamy sand and silty loam Anika Braun a,⁎, Erwin Klumpp b, Rafig Azzam a, Christoph Neukum a a b

RWTH Aachen University, Department of Engineering Geology and Hydrogeology, Lochnerstr. 4-20, 52064 Aachen, Germany Institute Agrosphere, IBG-3, Institute of Bio- and Geosciences, Forschungszentrum Juelich GmbH, 52425 Juelich, Germany

H I G H L I G H T S • • • •

Transport and deposition of surfactant stabilized AgNPs in glass beads and two natural soils were analyzed. FlFFF was implemented to determine particle size changes after soil passage. Mathematical modeling was used to quantify transport and deposition of the AgNPs. Transport of AgNPs was sensitive to flow velocity, soil composition and ionic strength/valence of cations present.

a r t i c l e

i n f o

Article history: Received 14 July 2014 Received in revised form 21 November 2014 Accepted 7 December 2014 Available online 17 December 2014 Keywords: Engineered nanoparticle Soil Transport and retention Transport modeling

a b s t r a c t It is considered inevitable that the increasing production and application of engineered nanoparticles will lead to their release into the environment. However, the behavior of these materials under environmentally relevant conditions is still only poorly understood. In this study the transport and deposition behavior of engineered surfactant stabilized silver nanoparticles (AgNPs) in water saturated porous media was investigated in transport experiments with glass beads as reference porous medium and in two natural soils under various hydrodynamic and hydrochemical conditions. The transport and retention processes of AgNPs in the porous media were elucidated by inverse modeling and possible particle size changes occurring during the transport through the soil matrix were analyzed with flow field-flow fractionation (FlFFF). A high mobility of AgNPs was observed in loamy sand under low ionic strength (IS) conditions and at high flow rates. The transport was inhibited at low flow rates, at higher IS, in the presence of divalent cations and in a more complex, fine-grained silty loam. The slight decrease of the mean particle size of the AgNPs in almost all experiments indicates size selective filtration processes and enables the exclusion of homoaggregation processes. © 2014 Elsevier B.V. All rights reserved.

1. Introduction As production and application of engineered nanoparticles (ENPs) are growing and the potential of their release into the environment is becoming more evident, the question of ENP mobility and fate in the environment is of increasing significance (Klaine et al., 2012). Regarding the safety of drinking water resources, it is crucial to understand transport and deposition processes of ENP in soils and aquifers. In contaminant transport studies ENPs represent a special case because the behavior of particles in the environment differs from that of conventional contaminants that usually occur in dissolved form. According to particle filtration theory, transport and deposition processes of colloids and nanoparticles in porous media are dominated by physicochemical interactions between the particle surfaces and the porous media (collector)

⁎ Corresponding author. E-mail address: [email protected] (A. Braun).

http://dx.doi.org/10.1016/j.scitotenv.2014.12.023 0048-9697/© 2014 Elsevier B.V. All rights reserved.

surfaces on the one hand, but also to hydrodynamic processes on the other hand (Yao et al., 1971; Torkzaban et al., 2007). Studies with ENP and natural colloids showed deviations of experimental observations from filtration theory, especially in environmentally relevant systems. For instance, under conditions that are generally considered as unfavorable for attachment, deposition may occur due to surface charge heterogeneities providing local favorable attachment sites (Lin et al., 2011). The occupation of these sites is eventually related to hydrodynamic drag forces allowing particles to travel along a collector surface until possibly reaching an attachment site (Bradford et al., 2011; Li et al., 2005). Moreover in natural porous media retention may occur due to mechanical straining in small pore throats and at locations of high surface roughness (Jaisi and Elimelech, 2009; Kasel et al., 2013a; Li et al., 2004; Wang et al., 2012). On the other hand, transport may be enhanced by stabilizing agents causing short range repulsive forces (El-Badawy et al., 2013; Xiao and Wiesner, 2013). The number of available attachment sites is subject to physicochemical factors and may change with the environmental conditions.

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

Silver nanoparticles (AgNPs) are the most commonly used nanomaterial and they are already widely applied in consumer products because of their strong antibacterial activity (WWICS, 2014). Because of the inevitable release of AgNPs into the environment (Geranio et al., 2009; Lorenz et al., 2012) and the high toxicity potential (Reidy et al., 2013) ENP environmental implication research has been particularly focused on AgNPs. Studies with AgNPs in artificial porous media under well-defined conditions have demonstrated that AgNP transport is sensitive to physicochemical factors such as solution ionic strength (IS), presence of divalent cations, grain size and hydrodynamics as is generally suggested by filtration theory (Liang et al., 2013a; Lin et al., 2011; Taghavy et al., 2013; Thio et al., 2011). The behavior of AgNPs has been shown to also strongly depend on the application of stabilizing agents that eventually enhance the mobility due to increased repulsion between AgNPs and AgNPs and porous media surfaces (El-Badawy et al., 2013; Liang et al., 2013a; Thio et al., 2011). Additionally, Liang et al. (2013a) observed concurrent deposition of surfactants to quartz sand surfaces, resulting in nonmonotonic deposition profiles of AgNPs in sand columns. Experimental studies on the transport behavior of AgNPs in natural porous media are scarce. Transport experiments in natural soils (Cornelis et al., 2013; Liang et al, 2013b; Sagee et al., 2012) and sandstones (Neukum et al., 2014a) showed that Ag nanoparticle transport is not only affected by the physicochemical parameters discussed above, but also by texture or grain size of the porous medium and the mineralogical composition. Heteroaggregation of AgNPs with natural colloids (Cornelis et al., 2012, 2013), and subsequent colloid facilitated transport of the AgNPs (Liang et al., 2013b; Neukum et al., 2014b) have been observed, demonstrating an important pathway for AgNP migration in soils and aquifers. In addition to the scarcity of studies with natural porous media, many experimental studies lack the application of adequate analytical techniques, which do not only consider the chemical but also the physical form of the nanomaterials (von der Kammer et al., 2012). The distinction between particulate and dissolved species or the characterization of transport relevant parameters such as particle size distribution and aggregation state are needed in order to understand the processes that occur during the transport of nanoparticles. While in many studies the nanoparticle samples are thoroughly characterized prior to the experiments, their physical state is not monitored during or after the experiments. Exceptions are for example Lin et al. (2011) who monitored the particle size distribution of AgNPs in the effluent of transport experiments by dynamic light scattering and found no change in particle size during transport experiments or Taghavy et al. (2013) who separated particulate Ag from dissolved Ag by centrifugation to monitor the dissolution of AgNPs and found that dissolution did occur during transport through the porous media. The purpose of this study was to gain more insight into the transport and retention processes of AgNPs in natural porous media. Column experiments were carried out with two soils, a loamy sand and a silty loam, and furthermore with glass beads as a reference medium to investigate the effect of different transport velocities and chemical compositions of the background solution on the transport behavior of AgNPs, which was then described with mathematical models. While in comparable studies with saturated porous media AgNP transport is investigated only at relatively high flow velocities that are not necessarily relevant for natural soil or aquifer systems (e.g. Cornelis et al., 2013 (8.14 m d− 1); Sagee et al., 2012 (9.5 m d−1); Taghavy et al., 2013 (7.6 m d−1)) in this study, a wider scope of expected groundwater velocities in comparable natural systems was covered ranging from 0.05 m d−1 to 2.88 m d−1. Another step towards more realistic conditions in soil would be the investigation under unsaturated conditions, which has partly been implemented for AgNPs by Liang et al. (2013b), who applied a water saturation of 90% in undisturbed loamy sand columns. Flow field-flow fractionation (FlFFF) was employed as analytical technique to characterize AgNP concentration and particle size

103

distribution before and after the soil passage. This enabled not only the detection of changes in particle size distribution during transport through the soil but also to account only for particulate Ag in the breakthrough curves because dissolved silver is not detected in the FlFFF measurement. In the case of some selected experiments the depth profile of AgNP retention in the soil columns was also analyzed and modeled providing further insights into the respective transport and deposition processes. 2. Materials and methods 2.1. Porous media and solution chemistry Glass beads with mean diameters between 300 and 400 μm were used as inert reference medium. In order to remove metal oxides and obtain completely hydrophilic surfaces of the glass beads they were cleaned using a procedure that was modified by Han et al. (2006) after Mehra and Jackson (1958) (see supporting information). Two soils were sampled from the upper 30 cm of agricultural field sites in Germany. The soils have already been used for other publications (Unold et al., 2009; Kasel et al., 2013b) and are therefore well characterized. Loamy sand (Gleyic Cambisol) was sampled from the test site Kaldenkirchen-Hülst (KDK), located at 51°18′22″ N 6°12′6″ E and silty loam (Orthic Luvisol) was sampled from the test site Merzenhausen (MZH) with the geographic coordinates 50°55′47‴ N and 6°17′48″ E (Unold et al., 2009). Physicochemical characteristics of the soils are given in Table 1. The clay fraction of the loamy sand soil contains the clay minerals illite, montmorillonite and kaolinite (Liang et al., 2013b). The clay fraction of the silty loam soil contains the clay minerals illite, chloride/vermiculite, kaolinite and swellable illite–smectite minerals (Burkhardt, 2003). The disturbed soil samples were oven dried at 40 °C, homogenized and sieved to b 2 mm. 2.2. AgNPs A stock suspension of stabilized AgNPs with a concentration of 10% w/w (AgPURE W-10) was purchased from ras materials, Regensburg, Germany. The product is identical to the OECD standard material NM300 Silver, which is currently used in nanoparticle research and was already thoroughly characterized by various independent institutions. The AgNPs are stabilized with 4% w/w each of polyoxyethylene glycerol trioleate and polyoxyethylene (20) sorbitan mono-laurat (Tween 20) (Klein et al., 2011), which form sterical repulsion barriers between the AgNPs to stabilize the suspension and minimize dissolution. According to Liang et al. (2013a), the amount of free surfactant in the stock suspension is around 5% w/w. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) indicate that 99% of the AgNPs are within the size range of 15 nm to 20 nm (Liang et al., 2013a). The hydrodynamic diameter of the particles diluted in water was determined with dynamic light scattering (DLS) to be around 50 nm (Klein et al., 2011). Prior to each experiment the stock suspension was diluted to a concentration of approximately 60 mg l−1 with the corresponding electrolyte solution (see Section 2.3).

Table 1 Physicochemical characteristics of the loamy sand (KDK) and silty loam (MZH).

Clay (b2 μm)a Silt (2–63 μm)a Sand (63–2000 μm)a pH (0.01 M CaCl2)a Total organic mattera Cation exchange capacitya Electrophoretic mobilityb a b

Unit

KDK

MZH

% mass % mass % mass

4.9 26.7 68.5 5.9 1.1 7.8 −2.70E−08

15.4 78.7 5.9 6.2 1.3 11.4 −3.20E−08

% mass cmolc kg−1 m2 V−1 s−1

Data adapted from Unold et al. (2009). Data adapted from Kasel et al. (2013b).

104

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

Table 2 Experimental conditions and recovered AgNP mass of transport experiments in columns packed with glass beads (GB), loamy sand (KDK) and silty loam (MZH) with ionic strength IS, Darcy flux q, input concentration C0 and mass recoveries from effluent Meffluent, column material Msoil and total recovery Mtotal. No.

Material

IS

Ion

mM I II III IV V VI VII VIII IX X

GB GB KDK KDK KDK KDK KDK KDK KDK MZH

10 10 10 10 0.01 50 1 10 50 10

NaNO3 NaNO3 NaNO3 NaNO3 H2O NaNO3 Ca(NO3)2 Ca(NO3)2 Ca(NO3)2 NaNO3

q

C0

Input mass

Meffluent

Msoil

Mtotal

cm min−1

mg ml−1

mg

%

%

%

0.1880 0.1701 0.1974 0.0033 0.1851 0.1871 0.1826 0.1896 0.1899 0.0015

0.055 0.056 0.058 0.051 0.041 0.029 0.035 0.016 0.016 0.053

5.365 5.282 6.685 5.026 3.641 2.629 3.071 1.475 1.495 5.479

62.80 79.92 85.54 1.77 100.04 2.15 3.95 0.00 0.00 0.00

7.78 6.70 8.17 57.77 n.d. n.d. n.d. n.d. n.d. 71.80

70.58 86.98 93.72 59.54

71.80

n.d.: Ag concentration in column was not determined.

2.3. Column experiments For the transport experiments the dried soil materials were wet packed into glass columns with an inner diameter of 26 mm and a length of 15 cm. Column in- and outlet were packed with a layer of about 2 mm of quartz wool to prevent fine soil grains from exiting the column and to distribute the flow at the column inlet. The columns were wet packed in the opposite direction of the final flow direction by introducing incremental amounts of soil and water. To allow the soil grains to settle and adjust the columns were carefully tapped with a rubber mallet and weakly compressed with a wooden pestle from time to time. The natural bulk densities of the loamy sand and silty loam, which were 1.5 g cm−3 and 1.6 g cm− 3 respectively (Förster et al., 2008), were used as orientation for the amount of soil packed into the columns. Prior to the transport experiments the hydraulic conductivity of the packed columns was determined using the falling head method. The columns were then connected to a high-pressure multichannel dispenser (ICP, ISMATEC, Germany) in an up flow mode. Tracer experiments with a nonreactive tracer were conducted on each column prior to the AgNP transport experiment. Sodium chloride was used as tracer in glass beads and loamy sand soil and due to massive leaching of the fine soil material when sodium chloride was used a relatively low concentration of sodium nitrate was used as tracer in the silty loam. The columns were first conditioned with several pore volumes of the corresponding electrolyte solution (deionized water for tracer experiment). Subsequently approximately three pore volumes of the solute/AgNP suspension respectively were injected into the column, followed by several pore volumes of deionized water/electrolyte solution. The column effluent was continuously collected in increments of approximately 6 ml, the exact volume of the increments was determined gravimetrically. The sodium chloride and the sodium nitrate concentrations were determined by measuring the electrical conductivity with a calibration using five standard solutions. The AgNP concentration and particle size distribution were determined by FlFFF analysis as explained in Section 2.4. For selected soil columns the material was excavated in 1 cm increments following the transport experiment. Retained Ag was digested with concentrated HNO3 and subsequently analyzed by inductively coupled plasma optical emission spectrometry (ICP-OES, Perkin Elmer, Optima 2000) at a wavelength of 328.068 nm. We selected those columns for the analysis, which were according to the BTCs believed to show the most interesting depth distribution of retained AgNPs, i.e. both glass bead columns, loamy sand columns covering the flow velocity variation and the silty loam column. To examine the effect of flow velocity on AgNP transport through the loamy sand the approach velocity of the experiments was varied between 0.0033 cm min−1 and about 2 cm min−1 as given in Table 2. Furthermore ionic strength was varied in the loamy sand. Concentrations between 10 mM and 50 mM of sodium nitrate (NaNO3) and 1 mM to 50 mM calcium nitrate (Ca(NO3)2) were applied as given in Table 2.

Deionized water was used in a reference experiment with loamy sand. 10 mM NaNO3 was used as background cation in all other experiments. The relatively impermeable silty loam soil allowed only very low flow velocities. One experiment with a flow velocity of 0.0015 cm min− 1 was carried out in this soil. For the very slow experiments No. IV and No. X that were conducted over a period of eight days a fraction collector was used. Generally, for all experiments unfavorable conditions for particle attachment and homoaggregation can be assumed. This is due to the steric stabilization of the AgNPs, their negative zeta potential as well as the negative zeta potential of the glass beads (Tufenkji and Elimelech, 2005) and the two soils under various IS conditions and the related electrostatic repulsion. However, due to the heterogeneity, especially of the natural soils, the occurrence of local favorable attachment sites is possible (Lin et al., 2011). 2.4. Flow field-flow fractionation Asymmetrical flow field-flow fractionation (AsFlFFF) analysis was performed with an AF2000 MT model coupled to a UV/Vis detector and a multi angle laser light scattering detector (MALLS) (Postnova Analytics GmbH, Landsberg/Lech, Germany). The channel of the AsFlFFF was equipped with a 10 kDa regenerated cellulose membrane and a 350 μm spacer providing a channel volume of 1147.125 mm3. A 10% v/v methanol solution was used as carrier liquid. An auto-sampler equipped with a 100 μl injection loop was used to inject the samples. The concentration of AgNPs was measured by UV/Vis detection at an absorption wavelength of 400 nm, which is the main absorption wavelength of AgPURE-W10. The quantitative detection of AgNPs was calibrated using five standard suspensions and the detection limit was 2.95 × 10−4 mg ml−1. Particle size distribution was determined with the residence time–particle size function that was calibrated using latex size standards of 22 nm, 68 nm and 100 nm diameter (Postnova Analytics GmbH, Landsberg/Lech, Germany). The size of the latex standards was verified with additional MALLS detection. The standard deviation of the size measurement is 2 nm as was determined by regularly measuring the AgNP calibration standards over the entire time period of the experiments. 2.5. Mathematical modeling The hydraulic properties of the packed soil columns were characterized by inversely modeling the tracer breakthrough curves (BTCs) using the CXTFIT code (Toride et al., 1999). The code provides an analytical solution to the advection–dispersion equation (ADE). The modeled and measured hydraulic parameters of the repacked glass bead and soil columns are given in the supporting information (Table S1). The HYDRUS1D software package (Simunek and van Genuchten, 2008) was used to model AgNP (BTCs) and, where data was available, the retention profiles (RPs). It was also used to model the column dispersivity using the

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

tracer BTCs. The finite element algorithm uses a Marquardt–Levenberg nonlinear least square optimization routine in order to inversely fit the parameters of the ADE to experimental breakthrough and retention data. The ADE is given as:   ∂θC ∂S ∂ ∂C ∂qC þρ ¼ θD − ∂t ∂t ∂x ∂x ∂x

ð1Þ

where θ [–] is the volumetric water content, t is time [T], ρ is the bulk density of the porous medium [ML−3], S is the solid phase particle concentration [MM−1], x is the spatial coordinate [L], D is the hydrodynamic dispersion coefficient [L2T− 1], C is the particle concentration in the aqueous phase [ML3] and q the flow rate [LT−1]. The solid phase particle mass balance can be expressed as: ρ

∂S ¼ θka ψC−kd ρS ∂t

ð2Þ

where ka is the first-order attachment coefficient [T−1], kd is the firstorder detachment coefficient [T−1] and ψ is a dimensionless colloid retention function that is given as: ψ¼

   S dc þ x −β 1− Smax dc

ð3Þ

where Smax is the maximum solid phase particle concentration [MM−1], dc is the mean grain diameter of the porous medium and β is an empirical variable, which controls the shape of the retention profile. The colloid retention function ψ enables the implementation of different time and depth dependent blocking models, that were also described by Torkzaban et al. (2008), predicting retention profiles that may be exponential, uniform or hyperexponential with depth (Bradford et al., 2006). The first term on the right hand side of Eq. (3) accounts for the time dependent filling and subsequent blocking of attachment sites based on Langmuirian dynamics (Adamczyk et al., 1995). This model considers a time dependent decrease of the attachment rate when attachment sites are filling up and the deposition profile becomes uniform as all attachment sites are filled and S approaches Smax. On the other hand, as Smax is large, the term approaches 1 and time dependent blocking is not considered. The second term on the right hand side of Eq. (3) accounts for depth dependent retention. With β N 0 depth dependent retention is predicted and the higher values of β, the more pronounced the depth dependence, which enables the prediction of hyperexponential deposition profiles. No depth dependent retention is predicted as β = 0, setting this term 1. An exponential distribution of deposited particles with depth, which is in accordance with conventional filtration theory, is predicted as ψ = 1. The parameters of the different model formulations, conventional attachment–detachment (ka, kd), time dependent retention (ka, kd, Smax), depth dependent retention (ka, kd, β) and time and depth dependent retention (ka, kd, Smax, β), were fitted to the experimental BTCs of AgNPs and where RP data was available, the models considering depth dependent retention were simultaneously fitted to BTC and RP. According to the filtration theory the attachment coefficient ka can be calculated after the following equation (Logan et al., 1995): ka ¼

3ð1−θÞ ηav 2dc

ð4Þ

where η (–) is the single collector efficiency, α (–) is the collision efficiency and v (L T−1) is the flow velocity. A measure for the characterization of the ratio of convective transport to diffusive transport is the Peclet number (NPe). It was calculated using the following equation (Tufenkji and Elimelech, 2004): Npe ¼

vdc D∞

ð5Þ

105

where D∞ (L2 T− 1) is the bulk diffusion coefficient described by the Stokes–Einstein equation. 3. Results and discussion 3.1. Characterization of the experimental system The AgNP suspensions were very stable in terms of dissolution and aggregation. Visual examination of AgNPs diluted to a concentration of 60 mg l− 1 in solutions with electrolyte concentrations of up to 500 mM NaNO3 and 100 mM Ca(NO3)2, indicated no signs of aggregation or sedimentation over a time period of weeks. Also the measurement of our standard sample used for the quantitative calibration of FlFFF showed a stable particle size with a mean diameter of 54 nm and a standard deviation of 2 nm and stable particle concentration over the whole experimental time period. The high stability is consistent with measurements by Klein et al. (2011), who observed no aggregation of NM-300 Silver diluted in water for more than 12 months and who estimated the Ag+ ion release to be below 0.01% w/w. Thus, the dissolution of the AgNPs can be neglected under the respective conditions. The zeta potential of the AgNPs used in this study was already measured in a previous study (Neukum et al., 2014a). In all examined electrolyte solutions it was weakly negative at the experimental pH of around 6–7 and it was weakest in 1 mM Ca(NO3)2 and most pronounced for AgNPs diluted in water (− 9 mV in deionized water, −7 mV in 10 mM NaNO3 and −5 mV in 1 mM Ca(NO3)2). Liang et al. (2013a) calculated the interaction energy barriers between the NM-300 Silver nanoparticles and sand surfaces and AgNPs among each other based on extended DLVO theory, which not only takes into account attractive van der Waals and repulsive electrostatic double layer interactions but also repulsive potentials arising from the surfactant, osmotic and elastic interactions. The calculations indicated that van der Waals and repulsive electrostatic double layer energy decreases slightly with increasing IS. Furthermore a shallow secondary minimum was predicted for AgNP–AgNP and AgNP–sand interactions. However, the steric repulsion resulting from the surfactant contributed predominantly to the overall repulsive conditions. Consequently the authors suggested overall unfavorable conditions for homoaggregation and attachment of the stabilized AgNPs in sand. It should be emphasized here that the calculations of the interaction energy are based on global zeta potential measurements for the nanoparticles and soil surfaces. The significance of such calculations for the characterization of heterogeneous porous media, such as natural soils, where local variations of the zeta potential occur due to surface roughness or variations in the mineral phase, has to be questioned (Lin et al., 2011; Pelley & Tufenkji, 2008). 3.2. Transport of AgNPs in glass beads Relative breakthrough concentrations of AgNPs in glass beads are plotted as a function of pore volumes flushed through the column (Fig. 1a). The transport of AgNPs in glass beads was retarded with a late breakthrough and a high degree of asymmetry of both BTCs. The variations between the two BTCs with an earlier breakthrough in experiment No. II in spite of identical experimental conditions occurred probably due to heterogeneities in the packed columns. This is also indicated by the higher hydraulic conductivity in experiment No. II (Table S1, supporting information) The total recovery rates of the experiments conducted in glass beads were relatively poor, with approximately 70% and 87%, which is probably due to the different calibrations used in the methods for silver analysis in effluent samples (FlFFF, internal AgNP standards) and soil samples (ICP-OES, official silver standard) where the absolute silver concentrations determined with FlFFF might be overestimated. The measured data of BTCs are assumed to be reliable because in experiment No. V, where AgNPs showed conservative transport behavior in loamy sand with H2O as background solution, the AgNP recovery was approximately 100%. Hence, only the BTCs were

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

a

b

I fitted 1.0

II fitted NaCl fitted

C/C0

0.8

II observed I observed

0.6 0.4 0.2

Relative particle diameter

106

1.2 1.0 0.8 0.6 0.4

I observed

0.2

II observed

0.0

0.0 0

1

2

3

4

5

0

6

2

0.00E+00 0

1.00E-07

2.00E-07

6

V/V0

V/V0

c

4

3.00E-07

4.00E-07

Transport distance (m)

-0.02 -0.04 -0.06 -0.08

I fitted

-0.1

II fitted I

-0.12

II

-0.14 -0.16

S/C0 (m3 g-1) Fig. 1. Observed and modeled BTCs of AgNPs and NaCl tracer (a), relative particle diameters in the column effluent (b) and observed and modeled RPs (c) of AgNPs in glass beads. Both experiments were conducted with 10 mM NaNO3 as background electrolyte at flow velocities of 0.1880 cm min−1 (I) and 0.1701 cm min−1 (II), respectively. Highlighted data points in the BTCs signify the end of the AgNP pulse and the arrow indicates the end of the tracer pulse. The particle diameters were normalized by the division through the mean particle diameter of the input sample.

considered for the modeling and the RPs were used as a general orientation for a rough qualitative comparison with the modeled RPs (Fig. 1c). The RP is plotted as solid phase concentration S (g g−1) normalized by the input concentration C0 (g m−3) as a function of transport distance. The solid phase concentration of AgNPs in glass beads was increasing with depth for experiment No. I. Even if the measured RP data is not regarded as quantitatively reliable, this distribution of AgNPs with depth was also observed visually. After the experiment a dark yellowish coloring of the glass beads was visible at the column exit, which was fading to a light yellowish coloring towards the column inlet, indicating a higher concentration of AgNPs at the column outlet. The AgNPs showed a relatively late breakthrough with a shallow ascending trend. The late breakthrough of the AgNPs with an asymmetric ascending trend of the BTC could be best fitted with a time dependent blocking

model. As expected due to the poor mass balances, the RPs were not completely captured by the models, but increasing deposition with depth could be modeled for experiment No. I when detachment was considered in the model formulations. Detachment is mostly interpreted by the deposition of AgNPs in the secondary minimum and their subsequent removal due to changes in e.g. hydrochemical conditions (e.g. Liang et al., 2013a). However, detachment would manifest in the BTC as tailing, which was not very pronounced or not at all observable here. To examine if AgNPs attached to glass beads could be remobilized, the second experiment was carried out (No. II) where the flow was started again with the electrolyte solution for another 4.7 pore volumes after a break and the remobilized AgNPs were measured in the effluent (data not shown). In this step, 20.40% of the theoretically deposited mass or 4.09% of the total input mass was recovered in the

Table 3 Parameters of the ADE modeled with HYDRUS-1D, with attachment coefficient ka, detachment coefficient kd, maximum solid phase concentration Smax and parameter of the retention function β. TDR: model with time dependent blocking/filling of retention sites, DDR: model with depth dependent retention, T + DDR: model with time and depth dependent retention, AD: attachment–detachment model. No.

Model

ka −1

I II III III IV IV V VI VII X

TDR TDR TDR T + DDR TDR T + DDR AD TDR TDR TDR

SE ka −1

s

s

1.12E−03 1.78E−03 1.18E−04 2.45E−02 5.70E−05 4.18E−04 2.48E−06 3.05E−03 2.26E−03 1.23E−03

6.53E-05 2.49E−04 1.96E−05 5.78E−03 6.75E−06 4.77E−05 9.00E−06 9.01E−05 1.11E−04 1.23E−06

kd −1

s

1.77E−05

SE kd −1

s

2.31E−05

Smax/C0 ml g

−1

SE Smax/C0 ml g

Beta

R2

−1

2.78E−07 2.23E−07 1.55E−07 1.11E−05 1.13E−06 1.16E−06

6.05E−09 1.04E−08 2.37E−08 3.43E−06 6.12E−08 6.42E−08

1.11E−06 1.40E−06 1.65E−05

1.74E−08 8.66E−08 5.70E−09

1.234 0.302

0.978 0.995 0.987 0.990 0.887 0.899 0.962 0.958 0.905 1.000

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

a

NaCl fitted 1.0

III fitted IV fitted

0.8

C/C0

III observed IV observed

0.6 0.4 0.2

In summary, the most relevant transport mechanisms of AgNPs in glass beads are the time dependent blocking of attachment sites and the concurrent deposition of surfactants and AgNPs to the glass bead surfaces. An additional potential mechanism is the reversible deposition of the AgNPs to the glass beads in the secondary minimum. 3.3. Transport of AgNPs in loamy sand 3.3.1. Flow velocity effect Experiment Nos. III and IV were conducted to examine the effect of hydrodynamics on AgNP transport in loamy sand. While the IS was kept at 10 mM NaNO3 for all experiments, experiment No. III was conducted at a relatively high flow velocity of 0.20 cm min− 1 (2.88 m d− 1) and experiment No. IV at a very low flow velocity of 0.0033 cm min−1 (0.05 m d−1). Those velocities are within or around the scope of typical velocities reported e.g. for shallow sandy aquifers in Germany, a possible real world scenario for the transport in the saturated loamy sand, which are between 0.15 and 1.4 m d–1 (Kunkel and Wendland, 1997). Fig. 2 shows the resulting BTCs and the relative solid phase concentration of AgNPs plotted against the transport distance (RPs) of the experiments. While at the high flow velocity around 85% of the AgNPs were transported through the column and recovered in the effluent, at the low flow velocity transport was strongly inhibited and only 1.79% of AgNPs passed the column. Basically, these observations are in accordance with filtration theory, which predicts increasing attachment with decreasing flow velocity (Yao et al., 1971). Both BTCs are asymmetrical with a late breakthrough, which indicates time dependent blocking behavior. The RPs of the experiment at the high flow velocity showed a hyperexponential shape, while the RP of the slow experiment was distributed rather exponentially with depth. Exponential distribution of deposited particles with transport distance is usually predicted by

b Relative particle diameter

effluent. From this observation we hypothesize that approximately 20% of retained AgNPs were deposited in the secondary minimum, which were probably remobilized due to the hydraulic impulse when the flow was started again (Solovitch et al., 2010). Fitted parameters of the ADE are given in Table 3. The modeled values of ka and Smax are around the same order of magnitude for both experiments in glass beads, although the BTC of experiment No. II showed a slightly steeper ascending trend. The RP of column No. II showed a relatively uniform distribution with depth after the remobilization step as can be seen in Fig. 1c. Liang et al. (2013a) also observed a remobilization of 7% of AgNPs deposited to quartz sand and suggested interactions in the secondary minimum as explanation, which was predicted by their extended DLVO theory calculations. Similar findings were made by Tian et al. (2010). Another explanation for the increasing retention with transport distance would be a concurrent adsorption of free surfactant blocking the attachment sites close to the column inlet due to the high affinity of the surfactants for the smooth glass or quartz surfaces. This effect was also suggested by Liang et al. (2013a), who observed nonmonotonic deposition of NM-300 Silver AgNPs in quartz sand. Surfactants bound to glass beads would further inhibit the deposition of AgNPs due to repulsive effects. Currently there exists no modeling concept where this effect can be considered, which would enhance the understanding of interactions between glass beads, surfactants and AgNPs. The size distribution of the hydrodynamic diameters of the AgNPs in the column effluents did not show any significant trends (Fig. 1b). The mean column effluent particle diameters were slightly below the mean particle diameters of the input samples (Table S2, supporting information). This phenomenon could be explained by selective filtration of the larger particles — increased filtration of larger particles has for example been observed for colloids by Tufenkji and Elimelech (2005). These results indicate that homoaggregation processes can be excluded.

1.4 1.2 1.0 0.8

III

0.6

IV

0.4 0.2 0.0

0.0 0

1

2

3

4

5

0

6

1

0.00E+00 0

5.00E-07

1.00E-06

1.50E-06

2.00E-06

d

Differential

Transport distance (m)

-0.02 -0.04 -0.06 -0.08

5

6

III effluent IV input

0.08

IV effluent

0.06 0.04

IV

0.02

III fitted

0.00

IV fitted

III input

0.10

-0.12

S/C0 (m3 g-1)

4

0.12

III

-0.16

3

0.14

-0.1

-0.14

2

V/V0

V/V0

c

107

1

10

100

Particle radius (nm)

Fig. 2. Observed and modeled BTCs and observed BTC of NaCl tracer (a), RPs (b), effluent mean particle size (c) and particle size distribution of input samples and representative effluent samples (d) of AgNPs in loamy sand with varied flow velocities. The experiments were conducted with 10 mM NaNO3. III: 0.2 cm min−1, IV: 0.0033 cm min−1. Highlighted data points in the BTCs signify the end of the AgNP pulse and the arrow indicates the end of the tracer pulse. Particle diameters and size distributions were only determined when the sample concentration exceeded the detection limit of the FlFFF detection.

108

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

filtration theory. The occurrence of hyperexponential RPs has been explained by Bradford et al. (2011) to depend on the hydrodynamic conditions at the column inlet. As the colloid flux close to the collector surface dominates over the mass transfer rate from the bulk aqueous phase to the near collector surface region, a higher deposition at the column inlet is the result. The impact of this specific flux at the column inlet is increasing with increasing flow velocity, decreasing grain size and increasing particle size under unfavorable attachment conditions (Bradford et al., 2011), which was the case in our experiments at higher flow rate (e.g. experiment No. III). Usually the effect of the grain size would be expected to be vice-versa, with a higher impact of advective transport and decreased deposition with an increasing grain size (Eq. (5)). On the other hand, exponential deposition profiles occur when the mass transfer from the bulk aqueous phase to the region adjacent to the collector surface is the dominating process, which occurs for example at low flow rates (Bradford et al., 2011), as was observed for experiment No. IV. According to filtration theory, diffusion is the dominating mass transfer process for particles b 1 μm (Yao et al., 1971). The dimensionless Peclet number (NPe) was calculated for each experiment using Eq. (5) (Table 2). NPe is strongly decreasing with a decreasing flow rate, from 426 for experiment Nos. III to X for experiment No. IV. These results indicate that at the low flow rate the role of diffusion in relation to advection is much more significant. In comparison to the experiment conducted at the higher flow rate, it can be interpreted that the increasing deposition of AgNPs with a decreasing flow rate is related to the increasing impact of diffusion. The mean particle size of AgNPs in the column effluent samples divided by the mean input particle size is plotted against pore volumes in Fig. 2c and the particle size distribution of the input suspension and selected effluent samples is presented in Fig. 2d. The results indicate that the particle size slightly decreased during transport through the soil matrix for the experiment at the higher flow velocity while it increased slightly at the slow flow rate (see also Table S2, supporting information). While the increase in particle size could be attributed to homoaggregation processes related to the dominating role of diffusion, the decrease of particle size in the other experiments could be explained by the selective filtration of larger particles in the soil matrix as already discussed above for the glass beads. It should be noted that for experiment No. IV because of the limited breakthrough only few samples reached the detection limit of FlFFF enabling the particle size determination. Moreover in experiment No. III there is a decreasing trend of the particle size with time, which cannot be physically explained. Fig. 3 shows different modeled BTCs (a) and RPs (b) for experiment No. III and the fitted parameters are given in Table 3. As can be seen from Fig. 3 the time and depth dependent retention model provides the best description of BTC and RP, while the time dependent retention (blocking) model cannot account for depth dependent retention of

a

Conservative

1.0 0.8

AgNPs and the depth dependent blocking model cannot account for the asymmetrical shape of the BTC. The values of the attachment rate ka modeled with the time and depth dependent retention model were increasing with an increasing flow rate (see Table 3), which is consistent with filtration theory, where ka is proportional to the flow velocity (Eq. (4)). The retention of AgNPs, however, was decreasing with an increasing flow velocity, which is also consistent with filtration theory, because the overall advective transport is proportional to the flow velocity (Bradford et al., 2011). The results are also consistent with the results of Liang et al. (2013a). The parameter β that was fitted to the RPs was increasing with flow velocity, which indicates increasing depth dependence with an increasing flow velocity. Although the models considering depth dependent retention proved to be more powerful in describing the observed hyperexponential RPs, it has to be considered that the parameter is only an empirical variable that cannot account for the underlying physical processes. Moreover, as pointed out above, the hyperexponential distribution of the solid phase concentration occurs due to hydrodynamic processes at the column inlet (Bradford et al., 2011). Possibly this effect is not relevant at all in realworld scenarios. Therefore the time dependent retention model is considered to provide more valuable information about relevant transport and retention processes of nanoparticles in porous media. In summary, the experimental results of the flow velocity variation indicate that AgNPs are mobile in the loamy sand at a high flow velocity and that retention is increased at a low flow velocity, which can be explained by the prevalence of diffusion over advection at low flow velocities. Because the chemical conditions and thus the existence of favorable attachment sites were kept constant during the experiments it can furthermore be concluded that not the availability of attachment sites but the accessibility of these sites according to the hydrodynamic conditions is limiting the retention of AgNPs. At the high flow rate a hyperexponential RP was formed in the columns that can be explained by a higher mass transfer rate of AgNPs to the collector surfaces due to the hydrodynamic conditions at the column inlet (Bradford et al., 2011). 3.3.2. Effect of ionic strength and valence of cation In order to study the influence of the chemical composition of the soil solution on the transport of AgNPs in loamy sand, the ionic strength and the valence of the cation present in the background solution were varied. Fig. 4a and b shows the relative breakthrough concentrations of AgNPs in loamy sand with deionized water (experiment No. V) or different concentrations of NaNO3 (experiment Nos. III and VI) and Ca(NO3)2 (experiment Nos. VII, VIII and IX) as background solutions plotted against flushed pore volumes. Fig. 4c shows the mean diameters of AgNPs in the effluent samples plotted against pore volumes. All experiments were conducted at flow velocities between 0.18 and 0.20 cm min−1. At this flow rate the AgNPs showed the highest mobility

b

TDB

-0.02

DDB

-0.04

Observed

0.4

Depth (m)

C/C0

T + DDB 0.6

0.0E+00 0

5.0E-07

1.0E-06

1.5E-06

-0.06 -0.08

Observed

-0.1

TDB

-0.12

0.2

DDB

-0.14 0.0 0

2

4

V/V0

6

2.0E-06

T + DDB

-0.16

S/C0 (ml g-1)

Fig. 3. Different model formulations for BTC (a) and RP (b) of experiment No. III (10 mM NaNO3, 0.2 cm min−1) considering time dependent retention (blocking) (TDB), depth dependent retention (DDB) and time and depth dependent retention (T + DDB). Highlighted data points in the BTCs signify the end of the AgNP pulse and the arrow indicates the end of the tracer pulse.

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

a

b

III fitted NaCl fitted

VI fitted VII fitted

0.08

VI observed

III observed V observed

0.6

VI observed VII observed

0.4

C/C0

C/C0

0.8

VII observed

0.06 0.04 0.02

0.2

0.00

0 0

1

2

3

4

5

6

0

1

2

3

4

5

6

V/V0

V/V0

Relative particle diameter

0.10

V fitted

1

c

109

1.2 1 0.8 0.6

V III VI VII

0.4 0.2 0 0

1

2

3

4

5

6

V/V0 Fig. 4. Observed and modeled BTCs of AgNPs and observed BTC of NaCl tracer in loamy sand under various IS conditions. a) V: H2O, III: 10 mM NaNO3, VI: 50 mM NaNO3, VII: 1 mM CaNO3, enlarged view of the BTCs of VI and VII (b) and effluent relative mean particle sizes (c). Highlighted data points in the BTCs signify the end of the AgNP pulse and the arrow indicates the end of the tracer pulse. Note that the data of experiment No. III is also shown in Fig. 2.

during the flow rate variation experiments. With deionized water as background solution, transport of AgNPs through the loamy sand was conservative with a recovery of 100.07% in the column effluent (Table 2). The BTC was modeled considering only attachment, resulting in a very small attachment coefficient of 2.48E− 6 s−1. However, the observed breakthrough occurred even earlier than the modeled breakthrough. Early breakthrough behavior of AgNPs in soils has been observed e.g. by Sagee et al. (2012) and Cornelis et al. (2013) and it can be explained by size and charge exclusion effects (Sagee et al., 2012; Bradford et al., 2006). Size or charge exclusion effects occur when particles cannot enter small pores due to their size or due to repulsive forces, restricting their transport to larger pores, where they are able to yield faster velocities than solutes that are traveling also through the smaller pores or that use the whole pore space for traveling (Bradford et al., 2006). The particles are then travelling in fractions of pores with higher transport velocities, resulting in decreased accessibility of solid surfaces for attachment (Bradford et al., 2006). Since the analysis of the particle diameters in the column effluent did not indicate any increase of the particle size or homoaggregation during transport through the soil and because at low IS conditions the electrostatic repulsion is expected to be highest, it is more likely that the charge exclusion effect is relevant here. When 10 mM of NaNO3 was added to the background solution, the breakthrough of AgNPs in loamy sand summed up to only 85% and the BTC had an asymmetric shape (experiment No. III, Fig. 4a, data also shown in Fig. 2). As already discussed above, this phenomenon can be explained by time dependent blocking, where attachment is decreasing as attachment sites are filling up with time. At higher concentrations of NaNO3 (50 mM, experiment No. VI) or in the presence of the divalent cation Ca2+ the breakthrough of AgNPs was strongly limited and it occurred delayed compared to conservative transport, with a recovery in the effluent of 2.15% and 3.95%, respectively (Fig. 4b). The asymmetrical shape of the BTCs due to the delayed

breakthrough indicates the time dependent blocking behavior of AgNPs under the respective conditions. Because no data on the RPs of these experiments was available, the BTCs were fitted only with the time dependent blocking model, which was able to describe the transport behavior, R2 of 0.958 and 0.905, respectively. No breakthrough could be observed at higher concentrations of Ca(NO3)2 (data not shown). The results indicate that the presence of the monovalent cation Na+ inhibits AgNP transport only at higher concentrations, while the presence of the divalent cation Ca2+ already affects AgNP transport at the low concentration of 1 mM. Both effects are well known and have been observed for engineered nanoparticles by other researchers (Liang et al., 2013b; von der Kammer et al., 2010). Increasing deposition with increasing IS is believed to be related to changes in the DLVO interaction potential between particles and collectors through compression of the electrostatic double layer (Tufenkji and Elimelech, 2005; Bradford et al., 2006), which can affect the particle surfaces as well as the collector surfaces. According to the Schulze–Hardy rule, the compression of the electrostatic double layer is increasing with increasing valence of the electrolyte (Lyklema, 2005). Liang et al. (2013b), who observed a similar behavior, already made an attempt to explain the more pronounced effect of the divalent cation Ca2 + on AgNP transport by bridging complexation between functionalized nanoparticles and soil grains. Bridging complexation occurs as functional groups at nanoparticle surfaces become bound to multivalent cations that in turn strongly bind to negatively charged sites e.g. on clay surfaces (Torkzaban et al., 2012). However, there is no direct evidence to prove this hypothesis. In preliminary experiments, where the stability of AgNP suspensions at very high IS (500 mM NaNO3 and 100 mM Ca(NO3)2, respectively) was observed over a time period of several weeks, no visible indications of aggregation or sedimentation processes appeared, which is because of the surfactant stabilization of the AgNPs. This indicates that the enhanced deposition phenomenon at high IS or valence results probably from the physicochemical properties of the soil surface such as the

110

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

characteristics of the electrostatical double layer, while homoaggregation processes can be excluded. This observation is supported by the measured particle size distributions of the effluent samples that indicate no homoaggregation of the AgNPs (Fig. 4c). Generally, in all experiments the particle diameters in the column effluent are slightly decreased compared to the input samples (Fig. 4c, Table S2, supporting information), which can be explained by size selective filtration as already discussed above. Models considering time dependent blocking were established for the four BTCs. A time and depth dependent retention model yielded the best fit for BTC and RP of experiment No. III where information on the RP was available. However, the results of the two different model types are not comparable (Kasel et al., 2013a). So here the time dependent blocking model results are compared for evaluating the effect of IS on AgNP transport. As can be seen from Table 3, the fitted values for ka and Smax increase with increasing IS or valence of the background solution as would also be predicted by filtration theory. In general, the higher mobility of AgNPs in the loamy sand soil compared to the glass beads, especially under low IS conditions, can be explained by more distinct repulsive conditions in the natural soil. This can be attributed to the properties of the soil, like a more pronounced surface charge or the auxiliary stabilization of the AgNPs by natural organic matter present in the soil. It is known that natural organic matter may have a stabilizing effect on nanoparticles or even alter their affinity for deposition to surfaces, which has for example recently been shown by Yang et al. (2014) for AgNPs and humic acids but under certain conditions natural organic matter may also have a destabilizing effect (Huynh and Chen, 2011). However, due to the lack of information on the types and quantity of available humic acids in the investigated soils, their contribution to the stabilization of the AgNPs can hardly be estimated. Additionally, it should be mentioned here that the soil is expected to strongly adsorb the free surfactants in the AgNP suspensions, as discussed already above for the glass beads. Due to the higher sorption capacity of the soil compared to the glass beads this effect occurs only in the first layers of the soil and it was neglected in the existing studies with NM-300 Silver (Liang et al., 2013a,b; Neukum et al., 2014a,b). 3.4. Transport of AgNPs in silty loam In the silty loam soil at a flow velocity of 0.0015 cm min−1 no breakthrough of AgNPs occurred (experiment No. X). The measured RP is strongly hyperexponential (Fig. 5). The applied flow velocity was the maximum flow velocity that could be achieved in the silty loam without generating excessive pressure. It corresponds to only half of the lowest velocity that was applied in the loamy sand (experiment No. IV). Due to the low flow velocity and the smaller mean particle diameter of the soil

Transport distance (m)

0.00E+00 0

5.00E-06

1.00E-05

1.50E-05

the NPe for this experiment was with 0.9 relatively small (Table S1, supporting information), indicating the dominance of diffusive transport versus advective transport, which might explain the increased deposition comparable to experiment No. IV. However, in the loamy sand under low flow velocity conditions the deposited AgNPs were more or less exponentially distributed throughout the whole column, while they were not transported beyond two centimeters in the silty loam. Different hypotheses can be considered to explain the observed high deposition of AgNPs in silty loam, which are related to grain size effects, mineralogical composition of the soil or long-term colloidal stability of AgNPs. An increased mass transfer rate to the collector surface with a decreasing grain size related to an increase in specific surface area is predicted by filtration theory, see also Eq. (4) (Yao et al., 1971). The formation of hyperexponential RPs is, according to Bradford et al. (2011), increased with a decreasing grain size, which would explain the hyperexponential distribution compared to the exponential distribution of AgNPs in the loamy sand under comparable experimental conditions. Furthermore, a decreased pore size can be expected with a decreasing grain size, which might enhance straining effects. In fact, the dp/d50 threshold for straining that was reported by Bradford et al. (2002) is exceeded for AgNPs in the silty loam (0.00224, threshold is 0.0017). Moreover the mineralogical composition of a soil has a major effect on the occurrence of favorable attachment sites. AgNPs were for instance found to be associated with clay minerals such as kaolinite and montmorillonite that were assumed to have positive surface charges at their edges (Cornelis et al., 2012, 2013), to soil colloids containing Fe, Al and Si (probably FeOOH and clay, respectively) (Liang et al., 2013b) or to iron oxides/hydroxides (Neukum et al., 2014b). Since both, kaolinite and iron oxides are also present in the silty loam (Burkhardt, 2003) the occurrence of favorable attachment sites can be assumed. It is worth noting here that also in the silty loam the presence of natural organic matter can have a stabilizing effect on the AgNPs. However, because of the strong restriction of the AgNP mobility in this soil due to the reasons pointed out above, there is no experimental evidence indicating an additional stabilization of the nanoparticles. In summary, the high retention of the AgNPs in silty loam can be explained by enhanced physicochemical filtration related to the high specific surface area due to the fine grain size and maybe the occurrence of favorable attachment sites due to the mineralogical composition of the soil. Moreover, regarding the grain size of the soil mechanical straining is a potential retention mechanism for AgNPs. It is worth noting here that the findings in this study are based on single experiments with repacked natural porous media that were conducted without replications or error quantification. However, the results are considered to be significant since they are consistent between the experiments. Moreover the effect of heterogeneity in the repacked columns was kept low due to the scale of the experiments with a transport length that is significantly larger- than the structural elements of the soils.

2.00E-05

4. Conclusions

-0.02 -0.04 -0.06 -0.08 -0.1 -0.12

X fitted

-0.14

X observed

-0.16

S/C0 (m3 g-1) Fig. 5. RP of AgNPs in the silty loam (experiment No. X). The experiment was conducted with 10 mM NaNO3 at a flow velocity of 0.0015 cm min−1.

The transport and deposition behavior of stabilized AgNPs was examined in glass beads and in two natural soils, a loamy sand and a silty loam. In glass beads the transport of the AgNPs was characterized by time dependent blocking and by the formation of nonmonotonic RPs in the packed columns, which are probably related to concurrent deposition of the stabilizing agents and/or reversible deposition of the AgNPs due to secondary minimum interactions. The AgNPs showed a relatively high mobility in the loamy sand, especially under low IS conditions and at higher flow rates. The transport of the AgNPs in loamy sand was inhibited at a low flow rate, which can be explained by the dominance of diffusion as transport mechanism enabling the access of additional favorable attachment sites, and at high IS or cation valence conditions, which is related to the compression of the electrostatic double layer of AgNP and soil surfaces, resulting in deposition. In loamy sand time and depth dependent retention was indicated by the

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

experimental results. The transport of the AgNPs was completely inhibited in the silty loam, which can be related to enhanced filtration due to the low flow velocity and the high surface area to volume ratio of the fine grained soil, the mechanical straining of AgNPs or to the enhanced availability of favorable attachment sites due to the mineralogical composition of the soil. The particle size of the AgNPs was monitored in the effluent of the column experiments with FlFFF. In almost all experiments the mean particle size of AgNPs slightly decreased after transport through the soil matrix. This indicates on the one hand that no homoaggregation of AgNPs occurred in the free soil solution and on the other hand that the filtration of AgNPs in the investigated media was size selective. In order to verify the observed effect, in future research complementary measurement techniques like SEM should be applied to confirm trends in particle size distribution measurements. It should be noted that especially the minerals that provide potential favorable attachment sites for AgNPs, such as clay minerals and oxides/ hydroxides, are often in the colloidal size scale and relatively mobile themselves. As has been shown by Liang et al. (2013b) for the Kaldenkirchen loamy sand and by Neukum et al. (2014b) for sandstone, there is a potential for the mobilization of colloids from the soil/rock matrix, leading to the colloid facilitated transport of the investigated AgNPs. The mechanisms and implications of colloid facilitated transport of nanoparticles are hardly explored, but the colloid facilitated transport might be a realistic scenario of AgNP remobilization in the environment. This aspect should be considered in future research. Acknowledgments This work was funded by the German Ministry of Education and Research (BMBF) within the WING/NanoNature program under contract 03X0077A. The authors are solely responsible for the content of this publication. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2014.12.023. References Adamczyk, Z., Siwek, B., Szyk, L., 1995. Flow-induced surface blocking effects in adsorption of colloid particles. J. Colloid Interface Sci. 174 (1), 130–141. http://dx.doi.org/ 10.1006/jcis.1995.1374. Bradford, S.A., Yates, S.R., Bettahar, M., Simunek, J., 2002. Physical factors affecting the transport and fate of colloids in saturated porous media. Water Resour. Res. 38 (12), 63-1–63-12. http://dx.doi.org/10.1029/2002WR001340. Bradford, S.A., Simunek, J., Bettahar, M., Van Genuchten, M.T., Yates, S.R., 2006. Significance of straining in colloid deposition: evidence and implications. Water Resour. Res. 42 (12), W12S15. http://dx.doi.org/10.1029/2005WR004791. Bradford, S.A., Torkzaban, S., Simunek, J., 2011. Modeling colloid transport and retention in saturated porous media under unfavorable attachment conditions. Water Resour. Res. 47 (10), W10503. http://dx.doi.org/10.1029/2011WR010812. Burkhardt, M., 2003. Field Studies to Determine the Transport Behavior of Dissolved and Particulate Tracers in a Silty Soil Using a Multi-tracing Technique. (In German). (Ph.D. thesis). Forschungszentrum Jülich, Germany. Cornelis, G., Doolette, C., Thomas, M., McLaughlin, M.J., Kirby, J.K., Beak, D.G., Chittleborough, D., 2012. Retention and dissolution of engineered silver nanoparticles in natural soils. Soil Sci. Soc. Am. J. 76 (3), 891–902. http://dx.doi.org/10.2136/ sssaj2011.0360. Cornelis, G., Pang, L., Doolette, C., Kirby, J.K., McLaughlin, M.J., 2013. Transport of silver nanoparticles in saturated columns of natural soils. Sci. Total Environ. 463, 120–130. http://dx.doi.org/10.1016/j.scitotenv.2013.05.089. El-Badawy, A.M., Aly Hassan, A., Scheckel, K.G., Suidan, M.T., Tolaymat, T.M., 2013. Key factors controlling the transport of silver nanoparticles in porous media. Environ. Sci. Technol. 47 (9), 4039–4045. http://dx.doi.org/10.1021/es304580r. Förster, M., Laabs, V., Lamshöft, M., Pütz, T., Amelung, W., 2008. Analysis of aged sulfadiazine residues in soils using microwave extraction and liquid chromatography tandem mass spectrometry. Anal. Bioanal. Chem. 391 (3), 1029–1038. http://dx.doi. org/10.1007/s00216-008-2081-1. Geranio, L., Heuberger, M., Nowack, B., 2009. The behavior of silver nanotextiles during washing. Environ. Sci. Technol. 43 (21), 8113–8118. http://dx.doi.org/10.1021/ es9018332.

111

Han, J., Jin, Y., Willson, C.S., 2006. Virus retention and transport in chemically heterogeneous porous media under saturated and unsaturated flow conditions. Environ. Sci. Technol. 40, 1547–1555. http://dx.doi.org/10.1021/es051351m. Huynh, K.A., Chen, K.L., 2011. Aggregation kinetics of citrate and polyvinylpyrrolidone coated silver nanoparticles in monovalent and divalent electrolyte solutions. Environ. Sci. Technol. 45, 5564–5571. http://dx.doi.org/10.1021/es200157h. Jaisi, D.P., Elimelech, M., 2009. Single walled carbon nanotubes exhibit limited transport in soil columns. Environ. Sci. Technol. 43, 9161–9166. http://dx.doi.org/10.1021/ es901927y. Kasel, D., Bradford, S.A., Simunek, J., Heggen, M., Vereecken, H., Klumpp, E., 2013a. Transport and retention of multi-walled carbon nanotubes in saturated porous media: effects of input concentration and grain size. Water Res. 47, 933–944. http://dx.doi.org/ 10.1016/j.watres.2012.11.019. Kasel, D., Bradford, S.A., Simunek, J., Pütz, T., Vereecken, H., Klumpp, E., 2013b. Limited transport of functionalized multi-walled carbon nanotubes in two natural soils. Environ. Pollut. 180, 152–158. http://dx.doi.org/10.1016/j.envpol.2013.05.031. Klaine, S.J., Koelmans, A.A., Horne, N., Carley, S., Handy, R.D., Kapustka, L., von der Nowack, B., Kammer, F., 2012. Paradigms to assess the environmental impact of manufactured nanomaterials. Environ. Toxicol. Chem. 31, 3–14. http://dx.doi.org/10.1002/etc.733. Klein, C.L., Comero, S., Stahlmecke, B., Romazanov, J., Kuhlbusch, T.A.J., Van Doren, E., De Temmerman, P.-J., Mast, J., Wick, P., Krug, H., Locoro, G., Hund-Rinke, K., Kördel, W., Friedrichs, S., Maier, G., Werner, J., Linsinger, Th., Gawlik, B.M., 2011. NM-300 Silver—characterisation, stability, homogeneity. NM-Series of Representative Manufactured Nanomaterials, European Commission Joint Research Centre, Institute for Health and Consumer Protection 978-92-79-19068-1. Kunkel, R., Wendland, F., 1997. WEKU—a GIS-supported stochastic model of groundwater residence times in upper aquifers for the supraregional groundwater management. Environ. Geol. 30 (1–2), 1–9. Li, X., Scheibe, T.D., Johnson, W.P., 2004. Apparent decreases in colloid deposition rate coefficients with distance of transport under unfavorable deposition conditions: a general phenomenon. Environ. Sci. Technol. 38 (21), 5616–5625. Li, X., Zhang, P., Lin, C.L., Johnson, W.P., 2005. Role of hydrodynamic drag on microsphere deposition and renetrainment in porous media under unfavorable conditions. Environ. Sci. Technol. 39 (11), 4012–4020. Liang, Y., Bradford, S.A., Simunek, J., Vereecken, H., Klumpp, E., 2013a. Sensitivity of the transport and retention of stabilized silver nanoparticles to physicochemical factors. Water Res. 42 (7), 2572–2582. http://dx.doi.org/10.1016/j.watres.2013.02.025. Liang, Y., Bradford, S.A., Simunek, J., Heggen, M., Vereecken, H., Klumpp, E., 2013b. Retention and remobilization of stabilized silver nanoparticles in an undisturbed loamy sand soil. Environ. Sci. Technol. 47 (21), 12229–12237. http://dx.doi.org/10.1021/ es402046u. Lin, S., Cheng, Y., Bobcombe, Y.L., Jones, K., Liu, J., Wiesner, M.R., 2011. Deposition of silver nanoparticles in geochemically heterogeneous porous media: predicting affinity from surface composition analysis. Environ. Sci. Technol. 45 (12), 5209–5215. http://dx. doi.org/10.1021/es2002327. Logan, B.E., Jewett, D.G., Arnold, R.G., Bouwer, E.J., O'Melia, C.R., 1995. Clarification of clean-bed filtration models. J. Environ. Eng. 121 (12), 869–873. Lorenz, C., Windler, L., Von Goetz, N., Lehmann, R.P., Schuppler, M., Hungerbühler, K., Heuberger, M., Nowack, B., 2012. Characterization of silver release from commercially available functional (nano) textiles. Chemosphere 89 (7), 817–824. http://dx.doi.org/ 10.1016/j.chemosphere.2012.04.063. Lyklema, H., 2005. Introduction to Colloid Science, in Fundamentals of Interface and Colloid Science Volume IV. Particulate Colloids. Academic Press, pp. 1-1–1-17. Mehra, O.P., Jackson, M.L., 1958. Iron oxide removal from soils and clays by a dithionite– citrate system buffered with sodium bicarbonate. Clay Clay Miner. 7, 317–327. Neukum, C., Braun, A., Azzam, R., 2014a. Transport of stabilized engineered silver (Ag) nanoparticles through porous sandstones. J. Contam. Hydrol. 158, 1–13. http://dx. doi.org/10.1016/j.jconhyd.2013.12.002. Neukum, C., Braun, A., Azzam, R., 2014b. Transport of engineered silver (Ag) nanoparticles through partially fractured sandstones. J. Contam. Hydrol. 164, 181–192. http://dx. doi.org/10.1016/j.jconhyd.2014.05.012. Pelley, A.J., Tufenkji, N., 2008. Effect of particle size and natural organic matter on the migration of nano- and microscale latex particles in saturated porous media. J. Colloid Interface Sci. 321, 74–83. http://dx.doi.org/10.1016/j.jcis.2008.01.046. Reidy, B., Haase, A., Luch, A., Dawson, K.A., Lynch, I., 2013. Mechanisms of silver nanoparticle release, transformation and toxicity: a critical review of current knowledge and recommendations for future studies and applications. Materials 6 (6), 2295–2350. http://dx.doi.org/10.3390/ma6062295. Sagee, O., Dror, I., Berkowitz, B., 2012. Transport of silver nanoparticles (AgNPs) in soil. Chemosphere 88 (5), 670–675. http://dx.doi.org/10.1016/j.chemosphere.2012.03.055. Simunek, J., van Genuchten, M.T., 2008. Modeling nonequilibrium flow and transport processes using HYDRUS. Vadose Zone J. 7 (2), 782–797. http://dx.doi.org/10.2136/ wzj2007.0074. Solovitch, N., Labille, J., Rose, J., Chaurand, P., Borschneck, D., Wiesner, M.R., Bottero, J., 2010. Concurrent aggregation and deposition of TiO2 nanoparticles in a sandy porous media. Environ. Sci. Technol. 44, 4897–4902. http://dx.doi.org/10.1021/ es1000819. Taghavy, A., Mittelman, A., Wang, Y., Pennell, K.D., Abriola, L.M., 2013. Mathematical modeling of the transport and dissolution of citrate-stabilized silver nanoparticles in porous media. Environ. Sci. Technol. 47 (15), 8499–8507. http://dx.doi.org/10.1021/es400692r. Thio, B.J.R., Montes, M.O., Mahmoud, M.A., Lee, D.W., Zhou, D., Keller, A.A., 2011. Mobility of capped silver nanoparticles under environmentally relevant conditions. Environ. Sci. Technol. 46 (13), 6985–6991. http://dx.doi.org/10.1021/es203596w. Tian, Y., Gao, B., Silvera-Bastista, C., Ziegler, K.J., 2010. Transport of engineered nanoparticles in saturated porous media. J. Nanoparticle Res. 12, 2371–2380. http://dx.doi.org/ 10.1007/s11051-010-9912-7.

112

A. Braun et al. / Science of the Total Environment 535 (2015) 102–112

Toride, N., Leij, F.J., van Genuchten, M.T., 1999. The CXTFIT code for estimating transport parameters from laboratory or field tracer experiments. Version 2.1. U.S. Salinity LaboratoryResearch Rep. 137USDA-ARS, Riverside, CA. Torkzaban, S., Bradford, S.A., Walker, S.L., 2007. Resolving the coupled effects of hydrodynamics and DLVO forces on colloid attachment in porous media. Langmuir 23 (19), 9652–9660. http://dx.doi.org/10.1021/la700995e. Torkzaban, S., Bradford, S.A., van Genuchten, M.T., Walker, S.L., 2008. Colloid transport in unsaturated porous media: the role of water content and ionic strength on particle straining. J. Contam. Hydrol. 96 (1), 113–127. http://dx.doi.org/10.1016/j.jconhyd.2007. 10.006. Torkzaban, S., Wan, J., Tokunaga, T.K., Bradford, S.A., 2012. Impacts of bridging complexation on the transport of surface-modified nanoparticles in saturated sand. J. Contam. Hydrol. 136, 86–95. http://dx.doi.org/10.1016/j.jconhyd.2012.05.004. Tufenkji, N., Elimelech, M., 2004. Correlation equation for predicting single-collector efficiency in physicochemical filtration in saturated porous media. Environ. Sci. Technol. 38 (2), 529–536. http://dx.doi.org/10.1021/es034049r. Tufenkji, N., Elimelech, M., 2005. Breakdown of colloid filtration theory: role of the secondary energy minimum and surface charge heterogeneities. Langmuir 21 (3), 841–852. http://dx.doi.org/10.1021/la048102g. Unold, M., Kasteel, R., Groeneweg, J., Vereecken, H., 2009. Transport and transformation of sulfadiazine in soil columns packed with a silty loam and a loamy sand. J. Contam. Hydrol. 103 (1), 38–47. http://dx.doi.org/10.1016/j.jconhyd.2008.09.002. Von der Kammer, F., Ottofuelling, S., Hofmann, T., 2010. Assessment of the physicochemical behavior of titanium dioxide nanoparticles in aquatic environments using

multi-dimensional parameter testing. Environ. Pollut. 158, 3472–3481. http://dx. doi.org/10.1016/j.envpol.2010.05.007. Von der Kammer, F., Ferguson, P.L., Holden, P.A., Masion, A., Rogers, K.R., Klaine, S.J., Koelmans, A.A., Horne, N., Unrine, J.M., 2012. Analysis of engineered nanomaterials in complex matrices (environment and biota): general considerations and conceptual case studies. Environ. Toxicol. Chem. 31 (1), 32–49. http://dx.doi.org/10.1002/etc.723. Wang, Y., Kim, J.H., Baek, J.B., Miller, G.W., Pennell, K.D., 2012. Transport behaviour of functionalized multi-wall carbon nanotubes in water-saturated quartz sand as a function of tube length. Water Res. 46 (14), 4521–4531. http://dx.doi.org/10.1016/j. watres.2012.05.036. WWICS (Woodrow Wilson International Center for Scholars), 2014. Project on Emerging Nanotechnology. http://www.nanotechproject.org/ Assessed February 2014. Xiao, Y., Wiesner, M.R., 2013. Transport and retention of selected engineered nanoparticles by porous media in the presence of a biofilm. Environ. Sci. Technol. 47 (5), 2246–2253. http://dx.doi.org/10.1021/es304501n. Yang, X., Lin, S., Wiesner, M.R., 2014. Influence of natural organic matter on transport and retention of polymer coated silver nanoparticles in porous media. J. Hazard. Mater. 264, 161–168. http://dx.doi.org/10.1016/j.jhazmat.2013.11.025. Yao, K.M., Habibian, M.T., O'Melia, C.R., 1971. Water and waste water filtration. Concepts and applications. Environ. Sci. Technol. 5 (11), 1105–1112. http://dx.doi.org/10.1021/ es60058a005.