Heteroaggregation of Titanium Dioxide Nanoparticles with Natural

Apr 25, 2015 - To better understand and predict the fate of engineered nanoparticles in the water column, we assessed the heteroaggregation of TiO2 na...
0 downloads 10 Views 1MB Size
Download PDF
Article pubs.acs.org/est

Heteroaggregation of Titanium Dioxide Nanoparticles with Natural Clay Colloids Jérôme Labille,*,†,§ Carrie Harns,‡ Jean-Yves Bottero,†,§ and Jonathan Brant‡ †

Aix-Marseille Université, CNRS, IRD, CEREGE UM34, Aix en Provence 13545, France Department of Civil and Architectural Engineering, University of Wyoming, 1000 E. University Avenue, Laramie, Wyoming 82071, United States § iCEINT, CEREGE, Aix en Provence 13545, France ‡

S Supporting Information *

ABSTRACT: To better understand and predict the fate of engineered nanoparticles in the water column, we assessed the heteroaggregation of TiO2 nanoparticles with a smectite clay as analogues for natural colloids. Heteroaggregation was evaluated as a function of water salinity (10−3 and 10−1 M NaCl), pH (5 and 8), and selected nanoparticle concentration (0−4 mg/L). Time-resolved laser diffraction was used, coupled to an aggregation model, to identify the key mechanisms and variables that drive the heteroaggregation of the nanoparticles with colloids. Our data show that, at a relevant concentration, nanoparticle behavior is mainly driven by heteroaggregation with colloids, while homoaggregation remains negligible. The affinity of TiO2 nanoparticles for clay is driven by electrostatic interactions. Opposite surface charges and/or high ionic strength favored the formation of primary heteroaggregates via the attachment of nanoparticles to the clay. The initial shape and dispersion state of the clay as well as the nanoparticle/clay concentration ratio also affected the nature of the heteroaggregation mechanism. With dispersed clay platelets (10−3 M NaCl), secondary heteroaggregation driven by bridging nanoparticles occurred at a nanoparticle/clay number ratio of greater than 0.5. In 10−1 M NaCl, the clay was preaggregated into larger and more spherical units. This favored secondary heteroaggregation at lower nanoparticle concentration that correlated to the nanoparticle/clay surface area ratio. In this latter case, a nanoparticle to clay sticking efficiency could be determined.

■

milligrams per liter in a river of low discharge12 to tens of grams per liter during a flood.13 Because of the high specific surface area (m2/g) and the reactivity of colloids, they may act as a carrier phase in the water column and strongly affect the fate and transport of nanoparticles.11,14,15 Nanoparticle heteroaggregation with inorganic colloids has been assessed using different techniques. Dynamic light scattering (DLS) can be used to quantify aggregation rates for nanoparticles and colloids in diffusion limited scenarios and to estimate global sticking efficiencies, α global , in the heterogeneous systems.16−18 However, DLS is inappropriate for measuring large and polydisperse aggregates thereby making interpretation of nanoparticle−colloid heteroaggregation data challenging. Some studies have evaluated nanoparticle−colloid heteroaggregation by measuring the sedimentation rate.19−21 The αglobal values returned by both of these experimental approaches are system specific and thus cannot be applied to

INTRODUCTION Nanotechnology-based products, or nanoproducts so-called because they are constituted of some fraction of engineered nanoparticles, have moved well beyond the laboratory and into the commercial marketplace.1 This has garnered them growing attention from different regulatory agencies interested in minimizing the environmental impact of these novel materials. To assess this risk, it is necessary to consider the likelihood of exposure to engineered nanoparticles through aqueous media as it is a receptacle to these materials during their life-cycle.2−5 Many factors influence the balance between nanoparticle dispersion/transport and aggregation/deposition in aqueous systems, both environmental and intrinsic to the nanoparticles themselves.6 The predicted concentrations of engineered nanoparticles in surface water systems are expected in the microgram per liter level for nano-TiO2 and even lower for many other types such as ZnO and nAg.7−9 Consequently, in real aquatic systems, the probability that nanoparticles interact with each other may be lower relative to their collision frequency with naturally occurring colloids that are present at substantially higher concentrations.10−12 Suspended particulate matter (SPM) in surface water indeed ranges from a few © 2015 American Chemical Society

Received: Revised: Accepted: Published: 6608

January 21, 2015 April 22, 2015 April 25, 2015 April 25, 2015 DOI: 10.1021/acs.est.5b00357 Environ. Sci. Technol. 2015, 49, 6608−6616

Environmental Science & Technology

Article

■

THEORETICAL APPROACH The mechanism for particle aggregation is generally viewed as the product of the interparticle collision frequency βij and the attachment efficiency αij, as first proposed by Von Smoluchowski.24,25 The population balance of a given aggregate size as a function of time can be modeled according to eq 1:

broader scenarios. A nanoparticle−colloid sticking efficiency that is independent of nanoparticle or colloid concentrations and only on the solution chemistry, here after called αNP‑C, is better suited as an input to nanoparticle fate models that account for heteroaggregation.10,11,15 However, αNP‑C remains a challenge to determine experimentally. Barton et al. could estimate αNP‑C values for different engineered nanoparticles in activated sludge, from a relationship to the time-resolved distribution coefficients,22 by measuring the nanoparticle concentrations in the sludge versus in the dispersing medium. However, this holistic approach does not allow elucidation of the underlying heteroaggregation mechanisms. In a previous work,14 we developed a mechanistic experimental approach based on laser diffraction that enabled us to measure heteroaggregation kinetics for TiO2 nanoparticles and silica microspheres (diameter = 0.5 μm) under relevant hydrochemical conditions. We demonstrated that αNP‑C could be determined from these data, using the surface area ratio of nanoparticles attached to the colloids because this ratio drives the probability that nanoparticles bridge the colliding colloids. Natural colloids are rarely monodisperse suspensions of homogeneous composition and surface charge like the silica microspheres used in our previous work. The validity of these advances remains uncertain with natural colloids of more complex characteristics and calls for more investigation. Smectite clay is a common material that is analogous to inorganic colloids occurring in surface waters. These phyllosilicates display a lamellar shape and a heterogeneous surface charge.23 The heteroaggregation phenomenon of nanoparticles with such colloids has been scarcely studied,16,19 and the driving mechanisms need to be identified. An open question is whether the surface area ratio of nanoparticles to clay drives the overall heteroaggregation kinetics as was previously observed for TiO2 and silica microspheres or whether there is a minimum number of nanoparticles per clay sheet required for heteroaggregation. In this work, the heteroaggregation kinetics of smectite colloids with TiO2 nanoparticles was characterized using laser diffraction. Our objective was to elucidate the fundamental mechanisms and to identify the key variables governing the evolution of homo- and heteroaggregation kinetics for the clay and nanoparticles. This was done at low nanoparticle concentrations (0.1−4 mg/L) with regard to the clay occurrence (100 mg/L) to develop realistic fate scenarios for surface water systems. The effect of the initial aggregation state of the clay was also studied by adjusting the solution ionic strength to 10−3 M or 10−1 M NaCl. A concentration of 10−3 M NaCl led to dispersed colloids with lamellar shape and high specific surface area, while 10−1 M NaCl led to clay homoaggregated in large and rather spherical units with low surface area. Finally, the impact that pH and natural organic matter have on nanoparticle heteroaggregation was also explored in order to develop a comprehensive assessment of the behavior of nanoparticles in surface waters. It is worthwhile to note that both the nanoparticle and clay concentrations used in this approach are in a range 100 times higher than the values expected in surface water. SPM typically ranges in the milligram per liter range during low discharge,12 while the predicted environmental concentration for nano-TiO2 is in the microgram per liter level. Thus, the nanoparticle/ colloid mass ratio studied in this work, for example, 1/1000, and the related heteroaggregation mechanism identified remain environmentally relevant.

dnk 1 = dt 2

∞

∑ i + j =k

αijβijninj − nk ∑ αikβik ni i =1

(1)

where ni and nj are the respective concentrations of particles of sizes i and j. When the aggregation process involves multiple types of particles assembling together, the heteroaggregation scenario becomes more complex to model because of the multiplicity of interacting components.26 In the present case of the heteroaggregation of colloids with nanoparticles, it can be viewed as a two steps mechanism: the nanoparticles first attach to the suspended colloids, forming the so-called primary heteroaggregates; then, heteroaggregation further occurs when the colliding heteroaggregates attach to each other and grow as secondary heteroaggregates. This latter step depends on the affinity of the meeting surfaces, that is, on the composition and orientation of the colliding heteroaggregates. Three different types of collisions can be expected: nanoparticle−nanoparticle, colloid−colloid, or nanoparticle−colloid, each leading to a respective sticking efficiency αNP‑NP, αC‑C, and αNP‑C. The overall heteroaggregation kinetics, that is, the global sticking efficiency αglobal, may thus result from the sum of the three complementary collision scenarios. The following expression for αglobal was proposed by Therezien et al.27 for such a two components system: αglobal(fi , f j ) = fi f j αNP‐NP + (1 − fi )(1 − f j )αC‐C + ((1 − fi )f j + (1 − f j )fi )αNP‐C

(2)

where f i and f j are the ratios of nanoparticles to colloids in the heteroaggregates i and j, respectively. The authors used f in terms of volume ratio, assuming that the respective probability for any of the three colliding scenarios depends on the volume fraction of nanoparticles to colloids. In the primary heteroaggregation step, the probability that discrete nanoparticles and colloids collide with each other is indeed a function of this volume ratio. However, during secondary heteroaggregation, it is not clear which dimension (volume, surface area, or number) of the nanoparticle/colloid ratio, f, will likely drive the collision scenario. The surface area ratio could reasonably be expected to be driving as previously shown for nano-TiO2 with silica microspheres.14 Meanwhile, the platelet shape of the clay colloid used here may also be determining and give more weight to the number ratio. Each of these three dimensions was tested in the present work in the consideration of f in order to study which one drives the heteroaggregation mechanism. Here, we considered that the fraction of incorporated nanoparticles in any two colliding heteroaggregates is the same due to the mechanically homogenized system. In turn, f i = f j, = f, and αNP‑C can be extracted from eq 2 as follows:14 αNP‐C = 6609

αglobal − f 2 αNP‐NP − (1 − f )2 αC‐C 2(1 − f )f

(3) DOI: 10.1021/acs.est.5b00357 Environ. Sci. Technol. 2015, 49, 6608−6616

Article

Environmental Science & Technology

the Cumulant algorithm are presented here in term of the average hydrodynamic diameter as a function of time. Clay Homo- and Heteroaggregation Measurement. Aggregation of the clay was measured by time-resolved laser diffraction (Mastersizer S, Malvern Instruments Ltd., Worcestershire, UK) as a function of solution chemistry. The montmorillonite suspension was prepared at 100 mg/L and at two different electrolyte concentrations (10−1 or 10−3 M NaCl). The pH of the suspensions was adjusted to 5 or 8 using HCl or NaOH. The suspension was permanently stirred in a baffled beaker with a motor mixer at 90 rpm, which corresponds to a mean shear rate of ∼100 s−1. The suspension was made to circulate in recycle mode through the measuring cell of the instrument with a pump at a constant flow rate of 30 mL/min. The clay was first let to equilibrate under the given conditions for 140 s so as to reach a stable state. During this equilibration time, the time-resolved size distribution was measured (a sequence of 0.5 s processing followed by 0−4.5 s pause) and is presented here in terms of the clay homoaggregation induced by the salt NaCl. In some cases, the effect of the natural organic matter humic acid, HA NOM, (Sigma−Aldrich, St. Louis, MO, USA) on heteroaggregation kinetics was also evaluated. The HA NOM was supplied as a dry powder, which was reported by the manufacturer to be constituted of 50−60% humic acid. In the associated experiments, the HA NOM was introduced into the clay suspension to achieve a HA NOM concentration of 1 mg/L. The size was then allowed to stabilize again for 140 s, while maintaining the pH at 5 or 8. Then, the heteroaggregation experiment started. At a new time 0, an appropriate aliquot of the TiO2 stock suspension was added to the montmorillonite suspensions to achieve the desired TiO2 concentration (0.1−4 mg/L), and the clay size distribution was measured over time. The laser diffraction kept blind to the presence of the nanoparticles dispersed in the reactor, as their low concentration and size with regard to the clay implied that the totality of the measured signal originated from the clay. Only the heteroaggregation of the clay induced by the nanoparticle addition was tracked in the present conditions.

While primary heteroaggregation does not lead to any measurable size increase due to the negligible size of the relevant nanoparticle (30 nm average hydrodynamic diameter) with regard to that of the clay particles (600 nm average lateral extension), secondary heteroaggregation does and could be measured using time-resolved laser diffraction. The heteroaggregation data measured were fitted with the SmoluCalc model28 (see model details in the Supporting Information). An overall sticking efficiency was returned by the model, corresponding to αglobal. The homoaggregation rates for the clay and nanoparticles in the same solution chemistries were measured separately so as to determine αC‑C and αNP‑NP, respectively. Then, eq 3 could be tested according to the dimension of f (see the detailed calculation of f in the Supporting Information), and the returned αNP‑C value could be examined to elucidate which of these dimensions drives the heteroaggregation kinetics.

■

MATERIALS AND METHODS Nanoparticles. TiO2 nanoparticles (lot no.: 7012-060410) were purchased from Nanostructured & Amorphous Materials, Inc. (Houston, TX, USA). As reported by the manufacturer, the TiO2 nanoparticles were of the mineral form anatase (purity = 99%). The stock nanoparticle dispersion was supplied as a 15% (w/w) water solution at pH 2. The discrete TiO2 nanoparticles had an average hydrodynamic diameter measured by DLS of 30 nm. They were characterized by an isoelectric point, IEPTiO2, of 6.5 (Figure S1, Supporting Information).14,29 The TiO2 nanoparticles were used without further purification and were stored in a refrigerator (T = 4 °C) until used. Clay. The smectite clay used was a sodium montmorillonite (SWy-2, Na-rich Montmorillonite, Crook County, WY, USA) purchased from the Clay Minerals Society (Chantilly, VA, USA). It was purified to ensure homogeneous association with sodium cations and removal of impurities (see the detailed procedure in the Supporting Information). A stock suspension of montmorillonite was made at a concentration of 500 mg/L by overnight magnetic agitation in ultrapure water (18.2 MΩ· cm). It was then stored in the dark and used within a week. The ζ potential of the montmorillonite in stock suspension was −47.8 ± 4.7 mV and remained weakly altered by changing the pH or NaCl concentration due to its permanent structural charge (see the Supporting Information). The size, aspect ratio, and stacking structure of the clay platelets are expected to play an important role in the heteroaggregation process with nanoparticles, since they affect the specific surface area of the clay. Each of the aforementioned properties were measured using an atomic force microscope. It returned an average platelet thickness of 2.2 nm and a lateral radius of 290 nm, giving a specific surface area of 35.2 m2/100 mg for the dispersed clay (see the Supporting Information and Figure S3). Nanoparticle Homoaggregation Measurement. Homoaggregation of the TiO2 nanoparticles induced by NaCl salt was measured by DLS at pH 5. Upon NaCl addition into the TiO2 suspension (10 mg/L), the size measurements were taken continuously for 10 min. Each measurement lasted for 15 s, and there was no delay between measurements. Size measurements were performed using the Malvern Zetasizer Nano ZS instrument (Malvern Instruments Ltd., Worcestershire, UK). After control of the satisfying fit residual, the data returned by

■

RESULTS AND DISCUSSION Salt-Induced Homoaggregation of the Nanoparticles. The evolution of the average hydrodynamic diameter of the TiO2 nanoparticles over time and as a function of solution ionic strength is given in Figure 1. As expected, the increasing ionic strength induces increasing aggregation kinetics due to progressive screening of the repulsive electrostatic interactions between the positively charged nanoparticles (pH = 5). The critical coagulation concentration (CCC) can be determined when the maximum aggregation rate is reached, corresponding to the diffusion limited aggregation mode, that is, the sticking efficiency (αNP‑NP) is 1. The obtained CCC of NaCl was 2.5 × 10−2 M. This value is close to the CCC 1 × 10−2 M measured at pH 8.14 Using the SmoluCalc model to fit these experimental aggregation plots, αNP‑NP could be determined for the lower salt concentrations, giving 0.15 and 0.03 for 10−2 and 7.5 × 10−3 M NaCl, respectively (Figure 1 and Table S5, Supporting Information). In typical seawater characterized by a total dissolved solids (TDS) concentration of 35 000 mg/L, the sodium cation represents about 1.08% of the total ionic composition, corresponding to 4.7 × 10−1 M. In surface water, the ionic concentration is approximately 2 orders of magnitude lower at 6610

DOI: 10.1021/acs.est.5b00357 Environ. Sci. Technol. 2015, 49, 6608−6616

Article

Environmental Science & Technology

mentally relevant concentrations. One consequence is that overestimated sedimentation rate may be predicted if extrapolation to a lower concentration using αNP‑NP is not made. In turn, in toxicology assays, this suggests that screening the nanoparticle concentration in a wide range could likely imply an altered exposure in the highest values due to the higher aggregation and sedimentation rates. Here at 0.1 mg/L, the homoaggregation is so slow that the induced aggregates remain in the micrometric size domain, that is, stable in suspension, for hours (e.g., 6 h is needed to reach 1 μm size). In surface water, whether the nanoparticles will undergo homoaggregation rather than heteroaggregation with inorganic colloids depends on the respective kinetics of these competing scenarios. High nanoparticle concentration will increase the homoaggregation rate while high colloid concentration will likely favor heteroaggregation. Thus, the respective concentration ratio is a key factor to study to determine the balance. Salt-Induced Homoaggregation of the Clay. The homoaggregation of the clay suspension induced by NaCl salt is shown on Figure S2 (Supporting Information) in terms of the median of the volumic size distribution, Dv50, as a function of time. In the electrolyte solution (10−3 M NaCl), the clay dispersion remains stable, characterized by a Dv50 of 0.5 μm at pHs 5 and 8, in good agreement with the lateral extension of the platelets. By increasing the ionic strength to 10−1 M NaCl, the CCC is surpassed, and the clay undergoes homoaggregation as already observed elsewhere.16,23 The resulting aggregation kinetics at 10−1 M NaCl were fitted with the SmoluCalc model, which returned αC‑C values of 0.13 and 0.027 for pHs 5 and 8, respectively. The lower sticking efficiency for homoaggregation at pH 8 is certainly related to the deprotonation of Al−OH groups located on the sheet edge, which favors interparticle electrostatic repulsions and thus higher dispersion stability.16,33 Conversely, at the acidic pH value, the Al−OH groups are positively charged and capable of favorably interacting with the negatively charged faces. The steady-state size obtained after a few minutes around 15 μm in 10−1 M NaCl corresponds to the initial size encountered in the following heteroaggregation measurements where the clay was first let to equilibrate with this electrolyte before introducing the nanoparticles.

Figure 1. Homoaggregation kinetics of TiO2 nanoparticles as a function of NaCl concentration measured by dynamic light scattering and modeled with SmoluCalc. pH = 5; TiO2 concentration = 10 mg/ L.

10−3−10−2 M mostly composed of calcium salts.12,26,30 This means that the CCC of Na+ to the TiO2 is likely to be reached and surpassed when the TiO2 nanoparticles are transported by a river through the salinity gradient in the estuary area, in agreement with previous findings.4,30−32 Nevertheless, nanoparticle homoaggregation may not be significant in such areas, since the collision frequency is a direct function of the nanoparticle concentration. The predicted environmental concentration of TiO2 nanoparticles is in the microgram per liter range,7−9 while for optimal measurement with DLS, the working concentration used here was 10 mg/L. To characterize the homoaggregation kinetics at more relevant nanoparticle concentrations, the SmoluCalc model was used to extrapolate the time-resolved aggregation plots at nanoparticle concentrations of 1 and 0.1 mg/L, while keeping αNP‑NP = 1. The resulting plots are shown in Figure 1. They reveal that the time required to reach a given aggregate size increases proportionally when the nanoparticle concentration decreases. Thus, the experimental investigations working at nanoparticle concentrations in the tens of milligrams per liter range measure an overestimated homoaggregation kinetics with regard to what would be experienced by the nanoparticles at more environ-

Figure 2. Clay/nanoparticle heteroaggregation kinetics measured at pH 5, as a function of NaCl concentration (a) 10−3 M or (b) 10−1 M and of the nanoparticle concentration (0.1−4 mg/L). Clay concentration = 100 mg/L. Numerical fits obtained from the SmoluCalc model are shown in a continuous line and labeled in the legend with the respective global sticking efficiency αglobal used. Dv50 is the median of the volume size distribution. 6611

DOI: 10.1021/acs.est.5b00357 Environ. Sci. Technol. 2015, 49, 6608−6616

Article

Environmental Science & Technology

Fitting the heteroaggregation curves with the SmoluCalc model returned a range of sticking efficiencies αglobal from 0.0075 to 1, respectively, to the nanoparticle dosage from 0.8 to 4 mg/L (Figure 2a and Table S6, Supporting Information). At the exact critical concentration of 0.8 mg/L, it is worthwhile to note that the size measurement returned a delay of 185 s in the onset of heteroaggregation. This delay could not be fitted numerically by using a lower sticking efficiency. It is likely to be due to the time required for the primary heteroaggregates to meet each other in an optimal orientation for nanoparticle bridging so that the required critical number of bridges is reached to bind neighboring particles. A shorter delay of 28 s was also observed at the higher nanoparticle dosage of 1 mg/L, while heteroaggregation started instantaneously at the highest dosages of 2.5 and 4 mg/L. At the critical concentration of 0.8 mg/L, the total projected area of the nanoparticles is 0.01 m2/L, based on a spherical shape with a 30 nm diameter (see the Supporting Information). On the other hand, the face and edge surface areas of the clay are about 35 and 0.27 m2/L, respectively (see the Supporting Information for NaCl 10−3 M). At this dosage, the total nanoparticle projected area thus corresponds to 0.029% of the face surface and to 3.9% of the edge surface of the clay. For comparison, a critical surface ratio of 5%, very close to the 3.9% calculated here, was evidenced previously for spherical microsilica heteroaggregating with the same nano-TiO2.14 This comforts that, in the present heteroaggregation mechanism, nanoparticle bridging might be located preferentially on the more energetic clay sheet edges. A hypothetical αNP‑C value was calculated from αglobal using eq 3 (the respective sticking efficiencies for homoaggregation αC‑C and αNP‑NP are zero). However, while a constant αNP‑C value close to 1 was expected in the present pH 5 condition due to electrostatic attractions between clay and nanoparticles, diverging values were actually obtained, whatever the number, surface, or volume consideration of f (Table S7, Supporting Information). For example, values for αNP‑C ranged from 0.1 to 3 for nanoparticle concentrations from 0.8 to 4 mg/L in an edge surface ratio consideration. This indicates that, in the present conditions of initially dispersed clay platelets, the overall heteroaggregation rate does not follow a simple probabilistic correlation to the nanoparticle dosage as tested in eq 3. Nevertheless, eq 3 was shown well adapted to the determination of αNP‑C in the case of heteroaggregation of silica microspheres with nanoparticles. The difference is certainly that microspheres follow a random orientation upon collision, giving a probability for heteroaggregation indeed related to the surface area ratio of attached nanoparticles.14 Here, the lamellar shape and flexibility of the clay particles certainly implies much less orientation scenarios upon collision. In this case, the number of nanoparticle bridges with regard to the lamellar colloids could be driving. Indeed, the relation of αglobal to the number ratio shown in Figure 3 follows a power 3 law rising around a ratio of one nanoparticle per two clay platelets. A high heteroaggregation rate is obtained when clay colloids and bridging nanoparticles are mixed in the same order of number concentration. An interesting analogy could reasonably be done with the well-known theory of bridging flocculation where one bridging polymer is needed to bind two neighbor sites.36,37 At an ionic strength of 10−1 M NaCl and pH 5 (Figure 2b), both the clay and TiO 2 nanoparticles likely undergo homoaggregation when taken separately. Since the clay suspension was let to stabilize in this electrolyte before

The corresponding conformational characteristics for the clay in the different solutions used here were calculated. In 10−3 M NaCl, the dispersed clay is characterized by a number concentration of 6.6 × 1013 platelets/L and a total surface area of 35 m2/L. In 10−1 M NaCl and at pH 5, these characteristics for the aggregated clay are 1.4 × 108 units/L and 0.14 m2/L, respectively (Tables S3 and S4, Supporting Information). Nanoparticle/Clay Heteroaggregation. The heteroaggregation kinetics measured at pH 5 are presented in Figure 2. Results are presented for two background electrolyte concentrations: 10−3 or 10−1 M NaCl. Time 0 corresponds to the time at which the TiO2 nanoparticles were added to the mixture. In most cases, a significant aggregation of the dispersion was measured following nanoparticle addition. The kinetics and the extent of aggregation appear very dependent on the concentrations of both the nanoparticle and the salt. The solution having an electrolyte concentration of 10−3 M NaCl at pH 5 (Figure 2a) is far below the CCC of both the nanoparticles and the clay when taken separately (see previous sections). However, when they are mixed together in this electrolyte, a large aggregation is measured at nanoparticle concentrations greater than 0.5 mg/L, and the heteroaggregation kinetics are positively correlated with the nanoparticle concentration. Since the respective homoaggregation mechanisms for the nanoparticles and the clay are not expected in such a condition and depletion flocculation of the clay has no reason to occur at so low of a nanoparticle concentration, the heteroaggregation of clay with TiO2 nanoparticles stands as the only possible scenario to explain the increase in size. Indeed, at pH 5, the TiO2 nanoparticles are positively charged (Figure S1, Supporting Information), while the clay displays energetic edges partly negatively charged with deprotonated Si−O− groups and protonated Al−OH2+ groups and faces expressing the structural anionic charge.23,34,35 The nanoparticles are thus likely to adsorb to the clay via attractive electrostatic interactions, primarily on the more reactive sheet edges and also potentially on the faces that constitute 99% of the clay surface area. In the aggregation mechanism involved here (10−3 M NaCl), we assume that the primary heteroaggregates associate with each other through bridging interactions, forming larger secondary heteroaggregates. In this case, the nanoparticles serve as the bridge, being attached to more than one clay particle at a time. Moreover, these clay/nanoparticle/clay bonds should withstand the mechanical shearing imposed by the stirred system to assume the large heteroaggregate formation observed. It appears that these conditions are fulfilled for nanoparticle concentrations greater than or equal to 0.8 mg/L. A relatively narrow critical concentration can be defined at this dose, as the closest concentrations tested here below and above, 0.5 and 1 mg/L, lead to very different outcomes. A TiO2 concentration of 0.5 mg/L did not lead to any measurable heteroaggregation within the 30 min time given to the experiment, while the 0.8 mg/L concentration led to a slow and delayed heteroaggregation rate of 17 μm/min. Increasing the nanoparticle concentration up to 4 mg/L led to a progressive increase in the heteroaggregation kinetics up to 8600 μm/min (Table S6, Supporting Information). Such gradation in the kinetics of the aggregate growth with the nanoparticle concentration suggests that the number of nanoparticle bridges is a limiting parameter. 6612

DOI: 10.1021/acs.est.5b00357 Environ. Sci. Technol. 2015, 49, 6608−6616

Article

Environmental Science & Technology

concentration of the clay units when initially homoaggregated (Table S4, Supporting Information). The nanoparticle/clay number ratio is indeed increased by 5 orders of magnitude, in the present conditions, compared to 10−3 M NaCl. This implies a much lower consumption of nanoparticle bridges for heteroaggregation. Moreover, the ever-existing homoattractions between clay units also contribute to decrease the bridging threshold for aggregation. A range of αglobal values was returned by the SmoluCalc model from this set of heteroaggregation kinetics, increasing linearly from 0.018 to 0.6 with nanoparticle concentrations from 0.1 to 4 mg/L (Table S6, Supporting Information, and Figure 3). In this case, using eq 3 with a fraction f weighted by the volume returned αNP‑C leveling at 140 ± 15, while using the number ratio led to αNP‑C = 0.5 independent of f. These unrealistic αNP‑C values indicate inappropriate f considerations. Meanwhile, very interestingly, using in eq 3 the fraction f weighted by the surface area ratio of the attached nanoparticles returns αNP‑C values leveling between 0.9 and 1.1, while a theoretical value constant and close to 1 was still expected (pH = 5). These very satisfying values validate the use of eq 3 in the present condition, and the correlation of the heteroaggregation kinetics to the surface coverage of the clay units by the attached nanoparticles. This trend contrasts with the behavior observed in the heteroaggregation of nanoparticles with the dispersed clay platelets in the lower ionic strength solution (10−3 M NaCl) but follows that previously observed with the silica microspheres.14 This is likely to be attributed to the spherelike shape of the preaggregated clay colloids, which implies random orientation of the colliding units and thus a probability for bridging correlated to the surface ratio of the attached nanoparticles. At pH 8, the TiO2 nanoparticles were negatively charged (Figure S1, Supporting Information), as well was the entire clay particle surface.23 Thus, no or at most a weak affinity was expected between both components. Indeed, in the low ionic strength solution (10−3 M NaCl), no aggregation was measured upon addition of the nanoparticles at concentrations of 0.8 or 4 mg/L in the clay suspension, as the size of the latter remained at 0.5 μm Dv50 during the whole duration of the experiment (Figure 4). This is likely to be attributed to repulsive electrostatic interactions between not only the clay platelets and the nanoparticles but also between the clay platelets. The ζ potential values are −40.7 and −30 mV for montmorillonite and nanoparticles, respectively, under these conditions (see the Supporting Information). A similar observation and interpretation was proposed by Kim et al. on the heteroaggregation of nZVI with clay.19 Nevertheless, as expected from the Debye− Hückel approximation,42 increasing the ionic strength shortens these double layer-induced repulsions and reveals the attractive interactions due to van der Waals and Lewis acid−base forces.38 This was observed in the higher ionic strength solution (10−1 M NaCl), where the injection of nanoparticles into the prehomoaggregated clay led to further heteroaggregation. In analogy with the trends observed at pH 5 and 10−1 M NaCl, the aggregation rate and the final aggregate size increased with the nanoparticle dosage. However, lower sticking efficiencies were returned in the present case of weaker interaction. Here, αglobal values of 0.02 and 0.3 were returned for nanoparticle concentrations of 0.8 and 4 mg/L, in comparison to the values of 0.15 and 0.6, respectively, obtained at pH 5. Equation 3 based on the surface ratio was used again to get the αNP‑C values from αglobal. While αNP‑C is expected to be between 0 and 1 and

Figure 3. Sticking efficiencies returned from the heteroaggregation plots as a function of the number ratio of nanoparticle to colloid units, and of the NaCl salt concentration. αglobal is returned from the SmoluCalc model. αNP‑C is calculated from eq 3 based on a surface ratio of nanoparticle to clay. In both series, the nanoparticle concentration range is the same, 0.1−4 mg/L, while the nanoparticle/colloids number ratio differs due to the clay homoaggregation state. Linear fitting equation (red line): y = 1.14 × 10−6x + 0.014; cubic fitting equation (black line): y = 0.8366x3.

nanoparticle injection at time 0, the initial size distribution in the heteroaggregation measurements corresponds to that of the previously formed clay homoaggregates, that is, 15−19 μm Dv50 (Figure S2, Supporting Information). Then, for all the doses tested (0.1−4 mg/L), the injection of nanoparticles induced a rapid increase in aggregate size, followed by a partial reversal and stabilization to a steady-state size. We assume that the rapid increase in aggregate size results from the heteroaggregation of the preliminary formed clay homoaggregates with nanoparticles. In the present electrolyte, the interparticle forces mostly remain attractive, likely resulting from van der Waals forces, Lewis acid−base forces,38 and partially lowered electrostatic attractions (Figure S1, Supporting Information). The breakage toward a lower steady-state size is certainly due to the very rapid heteroaggregation regime occurring first, which led to fragile and unstable heteroaggregates under the given hydrodynamic shear stress (100 s−1).39,40 The analysis of the static light scattering plots enabled us to clarify this mechanism (Figure S4, Supporting Information). A 2.3 fractal dimension characterizes the preexisting 15 μm clay homoaggregates and remains unchanged along the heteroaggregation and breakage process. This suggests a relatively dense structure, which does not allow the nanoparticles to diffuse inside41 and which withstands the breakage step. The bridging nanoparticles thus certainly remain at the outermost surface of the clay units binding neighbors together. During the breakage step, the heteroaggregate reconformation occurred at a scale larger than the pre-existing 15 μm aggregates, tending to a denser association of these latter units. In the end, the heteroaggregate tensile strength results from the sum of the homo- and heteroattractions, which is logically higher than the initial condition with clay homoattractions only, and leads to a final steady-state size larger than the initial one. Interestingly, no critical dose of nanoparticle for heteroaggregation could be evidenced in the 10−1 M NaCl solution as even 0.1 mg/L nanoparticles led to secondary heteroaggregation. This might be due to the lower surface area and number 6613

DOI: 10.1021/acs.est.5b00357 Environ. Sci. Technol. 2015, 49, 6608−6616

Article

Environmental Science & Technology

only when equivalent number concentrations of nanoparticles and colloids are reached. Such a condition may be encountered in surface water when a nanoparticle to colloid mass ratio of 8 to 1000 is reached, considering the clay platelet and nanoparticle dimensions used in this work. In an aqueous system favoring aggregation of the clay colloids, the surface area available for interaction with nanoparticles is reduced. This favors secondary heteroaggregation with nanoparticles at a lower mass ratio. Such conditions may be encountered not only in marine or esturary environments where NaCl-induced aggregation of the colloids is favored but also in surface water where a calcium concentration higher than 10−4 M leads to clay aggregation.33,39,43 In all cases, the secondary heteroaggregates formed are larger than the clay homoaggregates naturally forming in the water column, certainly favoring faster sedimentation. These findings also point to the likely forms that engineered nanoparticles will take in influent streams to water treatment systems. Recent efforts44,45 on evaluating nanoparticle removal in flocculation/clarification processes have largely focused on homoaggregation and neglected heteroaggregation. This may lead to an underestimation of nanoparticle removal during water treatment. The dimensionless αNP‑C value is a key parameter needed to feed environmental fate models. It could be calculated in this work with natural clay colloids under limit conditions that were clarified. These considerations are of high importance in the field of risk assessment of engineered nanoparticles, as they contribute to better predict the exposure aspect in systems with ever improving relevance. In this way, future work should consider that, in environmental aquatic systems, clays might be assembled in heterogeneous SPM involving other types of mineral and organic components, which will certainly affect the heteroaggregation mechanism with nanoparticles.

Figure 4. Clay/nanoparticle (NP) heteroaggregation at pH 8 as a function of the nanoparticle concentration, the NaCl concentration, and the presence of natural organic matter humic acid (1 mg/L). Numerical fits obtained from the SmoluCalc model are shown in lines and labeled in the legend with the respective global sticking efficiency αglobal used. Dv50 is the median of the volume size distribution.

constant with nanoparticle dosage, the values 0.13 and 0.45 were obtained for 0.8 and 4 mg/L nanoparticles, respectively. While the order of magnitude looks reasonable, the slight divergence might be due to the rough approximation made on the surface area and compactness of the prehomoaggregated clay units (see the Supporting Information). An average αNP‑C = 0.29 ± 0.16 can be considered and compared to the higher value 0.8 obtained with the same approach between the same nanoparticles and microsilica.14 In the presence of 1 mg/L HA NOM, the heteroaggregation rate induced by nanoparticle injection at 0.8 mg/L in the clay suspension is lowered (Figure 4). The αglobal and αNP‑C values of 0.005 and 0.0076, respectively, were returned and calculated. This is the effect of the NOM that adsorbs on the surfaces of both clay units and nanoparticles, leading to increased interparticle repulsions driven by both electrostatic (Figure S1, Supporting Information) and steric effects.5 It is worthwhile to note that this stabilizing effect of NOM that is well-known in homoaggregation stands also for heteroaggregation. Environmental Implications. Considering the exposure assessment of manufactured nanoparticles in aqueous environments, the present data provide evidence for some key mechanisms that will affect the fate in the water column. At the predicted environmental concentrations in the microgram per liter range, the fate of released nanoparticles is mainly driven by a heteroaggregation phenomenon with inorganic colloids. Homoaggregation of nanoparticles has a negligible contribution, whatever the ionic composition of the dispersing medium, as the underlying kinetics are largely outplayed by the interaction with more concentrated inorganic colloids such as clays. The affinity of nanoparticles for the colloids is logically the limiting factor and depends both on the respective surface chemistries and on the solution chemistry (pH, ionic strength). Moreover, the shape and dispersion state of the clay colloids play important roles in determining the fate of the attached nanoparticles. In fresh water of low salinity favoring the clay platelets dispersion, colloidal transport of the nanoparticles likely takes place under the form of dispersed primary heteroaggregates. Secondary heteroaggregation further occurs

■

ASSOCIATED CONTENT

S Supporting Information *

Additional details on characterization of the materials, the SmoluCalc model approach, structural analysis of the aggregates, and compilation of the aggregation data. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.5b00357.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was funded by the French National Research Agency as the NANOHETER program under the frame of SIINN, the ERA-NET for a Safe Implementation of Innovative Nanoscience and Nanotechnology, and the Environmental Research and Educational Foundation (EREF). The France−U.S. bridge was funded by the Partner University Fund Program (PUF) “Mechanistic Assessment of Manufactured Nanomaterial Behavior in Engineered Environments”. The authors acknowledge Drs. A. Thill and A. Praetorius for the fruitful discussions. 6614

DOI: 10.1021/acs.est.5b00357 Environ. Sci. Technol. 2015, 49, 6608−6616

Article

Environmental Science & Technology

■

valent iron nano-particles in heterogeneous porous media. J. Colloid Interface Sci. 2012, 370, 1−10. (20) Quik, J. T. K.; Stuart, M. C.; Wouterse, M.; Peijnenburg, W.; Hendriks, A. J.; van de Meent, D. Natural colloids are the dominant factor in the sedimentation of nanoparticles. Environ. Toxicol. Chem. 2012, 31 (5), 1019−1022. (21) Velzeboer, I.; Quik, J. T. K.; van de Meent, D.; Koelmans, A. A. Rapid settling of nanoparticles due to heteroaggregation with suspended sediment. Environ. Toxicol. Chem. 2014, 33 (8), 1766− 1773. (22) Barton, L. E.; Therezien, M.; Auffan, M.; Bottero, J. Y.; Wiesner, M. R. Theory and methodology for determining nanoparticle affinity for heteroaggregation in environmental matrices using batch measurements. Environ. Eng. Sci. 2014, 31 (7), 421−427. (23) Tombacz, E.; Szekeres, M. Colloidal behavior of aqueous montmorillonite suspensions: The specific role of pH in the presence of indifferent electrolytes. Appl. Clay Sci. 2004, 27 (1−2), 75−94. (24) Von Smoluchowski, M. Versuch einer mathematischen theorie der koagulation. Z. Phys. Chem. 1917, 92, 129−155. (25) Gardner, K. H.; Theis, T. L. A unified kinetic model for particle aggregation. J. Colloid Interface Sci. 1996, 180 (1), 162−173. (26) Findlay, A. D.; Thompson, D. W.; Tipping, E. The aggregation of silica and haematite particles dispersed in natural water samples. Colloids Surf., A 1996, 118 (1−2), 97−105. (27) Therezien, M.; Thill, A.; Wiesner, M. R. Importance of heterogeneous aggregation for NP fate in natural and engineered systems. Sci. Total Environ. 2014, 485, 309−318. (28) Thill, A.; Moustier, S.; Aziz, J.; Wiesner, M. R.; Bottero, J. Y. Flocs restructuring during aggregation: Experimental evidence and numerical simulation. J. Colloid Interface Sci. 2001, 243 (1), 171−182. (29) Loosli, F.; Le Coustumer, P.; Stoll, S. Effect of natural organic matter on the disagglomeration of manufactured TiO2 nanoparticles. Environ. Sci.: Nano 2014, 1 (2), 154−160. (30) Keller, A. A.; Wang, H. T.; Zhou, D. X.; Lenihan, H. S.; Cherr, G.; Cardinale, B. J.; Miller, R.; Ji, Z. X. Stability and aggregation of metal oxide nanoparticles in natural aqueous matrices. Environ. Sci. Technol. 2010, 44 (6), 1962−1967. (31) Ates, M.; Daniels, J.; Arslan, Z.; Farah, I. O. Effects of aqueous suspensions of titanium dioxide nanoparticles on Artemia salina: Assessment of nanoparticle aggregation, accumulation, and toxicity. Environ. Monit. Assess. 2013, 185 (4), 3339−3348. (32) Chinnapongse, S. L.; MacCuspie, R. I.; Hackley, V. A. Persistence of singly dispersed silver nanoparticles in natural freshwaters, synthetic seawater, and simulated estuarine waters. Sci. Total Environ. 2011, 409 (12), 2443−2450. (33) Van den Bogaert, R.; Labille, J.; Cornu, S. Aggregation and dispersion behavior in the 0- to 2-μm fraction of luvisols. Soils Sci. Soc. Am. J. 2014, in press. (34) Bottero, J. Y.; Bruant, M.; Cases, J. M.; Canet, D.; Fiessinger, F. Adsorption of nonionic polyacrylamide on sodium montmorillonite: Relation between adsorption, ζ potential, turbidity, enthalpy of adsorption data and 13C-NMR in aqueous solution. J. Colloid Interface Sci. 1988, 124 (2), 515−527. (35) Perronnet, M.; Villieras, F.; Jullien, M.; Razafitianamaharavo, A.; Raynal, J.; Bonnin, D. Towards a link between the energetic heterogeneities of the edge faces of smectites and their stability in the context of metallic corrosion. Geochim. Cosmochim. Acta 2007, 71 (6), 1463−1479. (36) Healy, T. W.; La Mer, V. K. The energetics of flocculation and redispersion by biopolymers. J. Colloid Sci. 1964, 19 (4), 323−332. (37) Hogg, R. The role of polymer adsorption kinetics in flocculation. Colloids Surf., A 1999, 146 (1−3), 253−263. (38) Vanoss, C. J.; Giese, R. F.; Costanzo, P. M. DLVO and nonDLVO interactions in hectorite. Clay Clay Miner. 1990, 38 (2), 151− 159. (39) Labille, J.; Thomas, F.; Bihannic, I.; Santaella, C. Destabilization of montmorillonite suspensions by Ca2+ and succinoglycan. Clay Miner. 2003, 38 (2), 173−185.

REFERENCES

(1) The Project on Emerging Nanotechnology. Inventory of nanotechnology-based consumer products, updated on October 2013, http://www.nanotechproject.org/cpi/about/analysis/. (2) Wiesner, M. R. Responsible development of nanotechnologies for water and wastewater treatment. Water Sci. Technol. 2006, 53 (3), 45− 51. (3) Wiesner, M. R.; Lowry, G. V.; Alvarez, P.; Dionysiou, D.; Biswas, P. Assessing the risks of manufactured nanomaterials. Environ. Sci. Technol. 2006, 40 (14), 4336−4345. (4) Botta, C.; Labille, J.; Auffan, M.; Borschneck, D.; Miche, H.; Cabie, M.; Masion, A.; Rose, J.; Bottero, J. Y. TiO2-based nanoparticles released in water from commercialized sunscreens in a life-cycle perspective: Structures and quantities. Environ. Pollut. 2011, 159 (6), 1543−1548. (5) Labille, J.; Feng, J. H.; Botta, C.; Borschneck, D.; Sammut, M.; Cabie, M.; Auffan, M.; Rose, J.; Bottero, J. Y. Aging of TiO2 nanocomposites used in sunscreen. Dispersion and fate of the degradation products in aqueous environment. Environ. Pollut. 2010, 158 (12), 3482−3489. (6) Labille, J.; Brant, J. Stability of nanoparticles in water. Nanomedicine 2010, 5 (6), 985−998. (7) Gottschalk, F.; Sun, T. Y.; Nowack, B. Environmental concentrations of engineered nanomaterials: Review of modeling and analytical studies. Environ. Pollut. 2013, 181, 287−300. (8) Liu, H. H.; Cohen, Y. Multimedia environmental distribution of engineered nanomaterials. Environ. Sci. Technol. 2014, 48 (6), 3281− 3292. (9) Meesters, J. A. J.; Koelmans, A. A.; Quik, J. T. K.; Hendriks, A. J.; van de Meentt, D. Multimedia modeling of engineered nanoparticles with SimpleBox4nano: Model definition and evaluation. Environ. Sci. Technol. 2014, 48 (10), 5726−5736. (10) Quik, J. T. K.; van de Meent, D.; Koelmans, A. A. Simplifying modeling of nanoparticle aggregation-sedimentation behavior in environmental systems: A theoretical analysis. Water Res. 2014, 62, 193−201. (11) Praetorius, A.; Scheringer, M.; Hungerbühler, K. Development of environmental fate models for engineered nanoparticles−A case study of TiO2 nanoparticles in the Rhine River. Environ. Sci. Technol. 2012, 46 (12), 6705−6713. (12) Slomberg, D.; Ollivier, P.; Radakovitch, O.; Baran, N.; Sani-Kast, N.; Miche, H.; Borschneck, D.; Grauby, O.; Bruchet, A.; Scheringer, M.; Labille, J. Characterization of suspended particulate and natural organic matter in the Rhône River: Insights into selecting surface water analogues. Environ. Chem. 2015, submitted for publication. (13) Thill, A.; Moustier, S.; Garnier, J. M.; Estournel, C.; Naudin, J. J.; Bottero, J. Y. Evolution of particle size and concentration in the Rhone river mixing zone: Influence of salt flocculation. Cont. Shelf Res. 2001, 21 (18−19), 2127−2140. (14) Praetorius, A.; Labille, J.; Scheringer, M.; Thill, A.; Hungerbuhler, K.; Bottero, J. Y. Heteroaggregation of titanium dioxide nanoparticles with model natural colloids under environmentally relevant conditions. Environ. Sci. Technol. 2014, 48 (18), 10690− 10698. (15) Sani-Kast, N.; Scheringer, M.; Slomberg, D.; Labille, J.; Praetorius, A.; Ollivier, P.; Konrad, H. Addressing the complexity of water chemistry in environmental fate modeling for engineered nanoparticles. Sci. Total Environ. 2014, in press. (16) Zhou, D. X.; Abdel-Fattah, A. I.; Keller, A. A. Clay particles destabilize engineered nanoparticles in aqueous environments. Environ. Sci. Technol. 2012, 46 (14), 7520−7526. (17) Afrooz, A. R. M. N.; Khan, I. A.; Hussain, S. M.; Saleh, N. B. Mechanistic heteroaggregation of gold nanoparticles in a wide range of solution chemistry. Environ. Sci. Technol. 2013, 47 (4), 1853−1860. (18) Huynh, K. A.; McCaffery, J. M.; Chen, K. L. Heteroaggregation of multiwalled carbon nanotubes and hematite nanoparticles: Rates and mechanisms. Environ. Sci. Technol. 2012, 46 (11), 5912−5920. (19) Kim, H. J.; Phenrat, T.; Tilton, R. D.; Lowry, G. V. Effect of kaolinite, silica fines and pH on transport of polymer-modified zero 6615

DOI: 10.1021/acs.est.5b00357 Environ. Sci. Technol. 2015, 49, 6608−6616

Article

Environmental Science & Technology (40) Bouyer, D.; Coufort, C.; Line, A.; Do-Quang, Z. Experimental analysis of floc size distributions in a 1-L jar under different hydrodynamics and physicochemical conditions. J. Colloid Interface Sci. 2005, 292 (2), 413−428. (41) Veerapaneni, S.; Wiesner, M. R. Hydrodynamics of fractal aggregates with radially varying permeability. J. Colloid Interface Sci. 1996, 177 (1), 45−57. (42) Hiemenz, P. C.; Rajagopalan, R. Priniples of Colloids and Surface Chemistry, 3rd ed.; Marcel Dekker Inc.: New York, 1997; p 605. (43) Labille, J.; Thomas, F.; Milas, M.; Vanhaverbeke, C. Flocculation of colloidal clay by bacterial polysaccharides: Effect of macromolecule charge and structure. J. Colloid Interface Sci. 2005, 284 (1), 149−156. (44) Honda, R. J.; Keene, V.; Daniels, L.; Walker, S. L. Removal of TiO2 nanoparticles during primary water treatment: Role of coagulant type, dose, and nanoparticle concentration. Environ. Eng. Sci. 2014, 31 (3), 127−134. (45) Kinsinger, N.; Honda, R. J.; Keene, V.; Walker, S. L. Titanium dioxide nanoparticle removal in primary prefiltration stages of water treatment: Role of coating, natural organic matter, source water, and solution chemistry. Environ. Eng. Sci. 2015, 32 (4), 292−300.

6616

DOI: 10.1021/acs.est.5b00357 Environ. Sci. Technol. 2015, 49, 6608−6616