Introduction

Carbon nanotubes (CNTs) are pure carbon macromolecules consisting of sheets of carbon atoms covalently bonded in hexagonal arrays that are seamlessly rolled into a hollow, cylindrical shape with both ends rounded through pentagon ring inclusions. Variable CNT architectures with diameters in the nanometer range (ca. 1-100 nm) and lengths up to several tens of micrometers give rise to high length to diameter aspect ratios as compared to other carbon fullerenes such as C₆₀ (1). On the basis of their structure, CNTs are categorized into two main species; single-walled nanotubes (SWNTs) and multi-walled nanotubes (MWNTs). The latter results from a coaxial assembly of the multiple SWNTs (2).

CNTs are being considered for a range of applications due to their exceptional mechanical, electro-optical, and thermal properties (3, 4). Examples of such properties and corresponding applications include high tensile strength and elasticity suitable for aerospace and fiber industries (5); electronic conductance and unique semiconducting capaci ties ideal for nanoelectronics and semiconductors (5-7); hydrogen adsorption (storage) capacity for application in hydrogen-based fuel cells (8, 9); and electronic sensitivity in different chemical environments allowing for novel environmental sensors (10-13). With increasing commercial interests and industrial scale production facilities currently under construction (14), CNT supply and demand, by all accounts, are expected to grow very rapidly over the next decade (15, 16).

Unfortunately, a limited amount of information is cur rently available regarding the fate and transport of CNTs in the natural and engineering environment and the ultimate human health effects (14). To date, studies have shown that CNTs are biologically active as demonstrated by a pulmonary response via induction of pulmonary granulomas (17, 18) at a greater instance than quartz (1-3 m crystalline silica), which is a recognized chronic occupational health hazard (via inhalation routes). Both SWNT and MWNT were also attributed to cause loss of phagocytic ability and ultrastructure damage to alveola macrophages (19). Furthermore, CNT have induced observable toxic responses in other cell cultures (20, 21).

When considering industrial scale production and use, observed biological activities, and the fact that CNT and other fullerene structures have been identified in soot from common hydrocarbon combustion processes (22, 23), understanding their fate in the natural environment is necessary to assess possible routes for exposure to humans and the ecosystem. The CNTs have seldom been considered as potential contaminants in the aqueous phase. They are extremely hydrophobic and prone to aggregation, as they are subject to high Van der Waals interaction forces along the length axis, and thus are not readily dispersed (24, 25). However, facile dispersion of CNT in the aqueous phase can be achieved by augmenting the surface of the carbon structure through the addition of surfactants and polymers such as sodium dodecyl sulfate (SDS) (26-33), sodium dodecylbenzene sulfonate (NaDDBS) (34, 35), Triton X-100 (34, 36, 37), and polyvinyl pyrrolidone (PVP) (38) among others (34, 39). These surfactants and polymers not only create a thermodynamically suitable surface in water but also provide steric or electrostatic repulsion among dispersed CNTs, thus preventing aggregation (27, 30, 35, 36).

Given the previous observations that CNTs are stabilized in the aqueous phase by well-characterized surfactants and polymers, it is possible that similar interactions between CNTs and organic molecules present in natural systems will occur. This may result in aqueous dispersion and stabilization of CNTs following a demonstrated mechanism of hydrophobic surface shielding that has not been widely considered in environmental fate and transport studies to date. The objectives of this study were to verify MWNT stabilization, as an aqueous suspension, in both synthetic solutions containing model natural organic matter (NOM) and natural surface water with a high NOM background (Suwannee River) and to develop a method to quantify MWNTs suspended in NOM solutions based on thermal optical transmittance analysis.

Experimental Procedures

MWNTs, produced by a chemical vapor deposition (CVD) method with purity greater than 90%, were obtained from the MER Corporation (Tucson, AZ). The average diameter and average length were reported by the manufacturer to be 140 ± 30 nm (approximately 100 graphene layers per each molecules on average) and 7 ± 2 m, respectively. Standard Suwannee River NOM (SR-NOM) obtained from the International Humic Substances Society (IHSS) (St. Paul, MN) was used as a model NOM. Number average and weight average molecular weights of SR-NOM determined by gel permeation chromatography were reported as 1718 and 2703 Da, respectively (40). On the basis of the analytical informa tion provided by IHSS, SR-NOM is composed of 52.47 wt % carbon, 4.19 wt % hydrogen, 42.69 wt % oxygen, 1.10 wt % nitrogen, 0.65 wt % sulfur, and 0.02 wt % phosphate, and the ash content is 7.0 wt %. A significant amount of carbon in the SR-NOM is distributed at the carboxylic group (20%), which provides an acidic moiety to the SR-NOM, as well as aromatic (23%), aliphatic (27%), and heteroaliphatic (15%) groups. A 100 mg C/L stock solution was prepared by dissolving SR-NOM for 24 h and filtering the solution through a 0.2 m nylon membrane filter (Cole Parmer, Chicago, IL). For comparison with model SR-NOM, a grab sample of actual Suwannee River water was obtained from the official sampling site of the IHSS, located in the Okefenokee National Wildlife Refugee in Georgia (41). The water sample was transported in a cooler packed with ice from the sampling location and preserved in a 4 C temperature room after being filtered with a 0.2 m nylon membrane filter (Cole Parmer, Chicago, IL). ACS reagent grade (>99%) sodium dodecyl sulfate (SDS, CH₃(CH₂)₁₁OSO₃Na, FW = 288.38) (Aldrich Chemical Co., Milwaukee, WI), which has a critical micelle concentration of 0.24% at 25 C, was used as a representative surfactant to stabilize MWNTs in water. Ultrapure water (>18 M) produced by the Milli-Q water purification system (Millipore, Billerica, MA) was used for the preparation of all solutions.

MWNT suspensions were prepared by adding varying amounts of MWNT into 100 mL of a Milli-Q water, 1% SDS solution, a well-known surfactant to stabilize CNTs, as a positive control, solutions containing varying concentrations SR-NOM, and Suwannee River water in Erlenmeyer flasks and vigorously agitating the solutions for 1 h. After settling for 4 days, the unsettled supernatant (ca. 60% of total volume) was carefully removed by a syringe from the top of the flask. The solution was then filtered using a Whatman Model 541 filter paper (20-25 m, Florham Park, NJ) to remove any undispersed MWNT agglomerates, and the filtrate was collected for further analyses.

In situ images (i.e., suspended in the water phase) of MWNTs in suspension were obtained using a Leica DM IRM Differential Interference Contrast (DIC) Microscope (Wetzlar, Germany) operated in a reflective index mode and recorded with a Hamamatsu EM-CCD C9100 Camera (Hamamatsu City, Japan). The point-to-point resolution of the resulting image was 0.053 m. Suspended MWNTs were particularly easy to identify and record due to the contrasting reflective indexes of MWNT as compared to the aqueous background. Electron microscopic images were analyzed by a JEM 100C transmission electron microscope (TEM) (Jeol, Peabody, MA) using 100 kV electron beam at magnifications of 7200 and 100 000. TEM samples were prepared by placing a droplet of MWNT aqueous suspension on the 300 mesh copper carbon grid (Electron Microscopy Science, Hatfield, PA) and dried overnight at room temperature.

Concentrations of suspended MWNTs were determined by a Thermal Optical Transmittance (TOT) Analyzer (Sunset Laboratory, Tigard, OR), UV-vis absorbance, and turbidity measurements. TOT analysis was performed following the method described by NIOSH (National Institute for Oc cupational Safety and Health) (42). Samples for the TOT were prepared by filtering a known volume of MWNT suspension through a 25 mm diameter disc type Pallflex 2500 quartz filter (Pall Corporation, Ann Arbor, MI) with a nominal pore size of 0.3 m, which has good durability at high-temperature conditions of TOT, and drying the filter for 24 h at 90 C. An independent control test confirmed that virtually all MWNTs stabilized in the aqueous suspension were retained by this filter (i.e., the concentration of MWNT in the filtrate was less than the detection limit of the analytical methods used in this study). NOM and SDS that were not associated with MWNTs were also removed as filtrates during this step. After drying, each TOT specimen was prepared by cutting a 1.5 cm² rectangular area from the center of the glass filter and loading it on a glass sampling boat of the TOT. Before each set of measurements, the equipment was calibrated using a 10 L of a 5.0 mg/L sucrose solution. For each measurement, the flame ionization detector (FID) was calibrated using a known volume of CH₄. UV-vis absorbance and turbidity of MWNT suspensions were measured by an Agilent 8453 UV-vis spectroscopy system (Palo Alto, CA) and a Hach 2100N turbidimeter (Loveland, CO), respectively. Concentrations of NOM remaining in the solution phase, which did not adsorb onto MWNT, were quantified by UV-vis absorbances at 254 nm (UV₂₅₄) after removing the suspended MWNT and associated NOM with a GHP Acrodisc 0.2 m syringe filter (Pall Corporation, Ann Arbor, MI). UV₂₅₄ measurements were calibrated with DOC analysis (TOC-Vw analyzer, Shimadzu, Columbia, MD).

Results and Discussion

The stability of MWNTs in the aqueous phase was largely dependent on the presence of SDS or NOM (Figure 1). The MWNTs added to organic-free Milli-Q water at 500 mg/L (50 mg of MWNTs added to 100 mL of Milli-Q water) settled quickly, and the water became completely transparent in less than an hour (Figure 1a). Upon the addition of the MWNT at the same (equivalent) concentration, a 1% SDS solution immediately became dark and turbid. The solution gradually changed to a light gray suspension after 1 day of settling, and the color of the solution did not noticeably change for over a month (Figure 1b). The solution of 100 mg C/L SR-NOM originally appeared dark and turbid upon equivalent MWNT addition and gradually lightened with a corresponding loss of turbidity during the first 4 days of settling. However, after 4 days, the dark solution, which appeared to be due to the presence of MWNTs, with a yellowish background remained stable for over a month (Figure 1c).

Figure 1 Visual examination of (a) organic-free water, (b) 1% SDS solution, (c) 100 mg C/L SR-NOM solution, and (e) Suwannee River water after adding 500 mg/L MWNTs, agitating for 1 h, and quiescent settling for 4 days. The 100 mg C/L SR-NOM solution and Suwannee River water without MWNT addition are also shown in panels d and f.

In situ microscopic images of MWNTs suspended in the SDS and SR-NOM solutions are presented in Figure 2. Negligible morphological differences were observed between the MWNTs stabilized by SDS (Figure 2a) and those by SR-NOM (Figure 2b). Both samples contained relatively well-dispersed rod-shaped MWNTs as well as larger size flocs, albeit in much less frequency. A closer examination of MWNTs in a SR-NOM solution by TEM suggested that the majority of MWNTs was suspended as a single tube (a representative image shown in Figure 2c), as evidenced by the presence of a single hollow core (approximately 2 nm in diameter according to the manufacturer) in a single fibrous structure in these images. Bundles of MWNTs (i.e., pairing of several MWNTs along the length axis) were seldom observed. Bubble-like artifacts adjacent to MWNT surfaces might have originated from the sublimation of organic matter due to high-energy electron beam irradiation during TEM analysis. The agglomerates of MWNTs in which a single MWNT appeared to be connected to another MWNT, potentially by bridging through NOM, were also observed during TEM examination. While it is possible that some of these bridged structures might have been additionally generated during a drying process for TEM sample prepara tion, the similar structures were also observed in the in situ images.

Figure 2 (a) In situ microscope images of MWNTs suspended in 1% SDS solution and (b) 100 mg C/L SR-NOM solution (from Figure 1). The scale bars in the upper right corner of each image correspond to 5.3 m. (c) Representative TEM image of MWNTs stabilized in the SR-NOM solution.

In the presence of complex and undefined NOM, analytical approaches utilizing elemental and molecular characteristics may prove challenging to quantify the amount of CNT suspended in the aqueous phase. It should also be noted that the CNT sample used in this study consists of a mixture of molecules with different sizes. A method of using TOT relies on the fact that thermal stabilities of organic carbon (NOM) and elemental carbon (CNT) are different such that they can be differentially quantified (43-46). This instrument is widely used to examine carbon content and composition in various atmospheric samples (47-51).

The TOT analysis typically proceeds in two distinct stages (52). In the first stage, the temperature increases stepwise up to 820 C in a He atmosphere to volatilize organic carbon (OC), which is then oxidized to CO₂ via granular MnO₂ at 900 C. CO₂ is subsequently reduced to CH₄ by a Ni/firebrick methanator at 450 C and quantified by a flame ionization detector (FID). However, not all of OC is volatilized, as some becomes pyrolized in the O₂ deficient atmosphere. Elemental carbon (EC) measurements along with pyrolized organic carbon (PC) correction are performed in the next stage. The temperature is again raised stepwise in an O₂ (10%) and He (90%) mixed atmosphere. The total amount of thermally oxidized EC and PC is measured by the FID after reduction to CH₄. Utilizing He-Ne laser transmittance through the sample, EC from the original sample is differentiated from PC. In the first stage, as PC is generated and absorbs the light, the laser transmittance is decreased. However, as both EC and PC are volatilized, the laser transmittance increases in the second stage. The point at which the laser transmittance reaches the initial value (time = 0) is the separation point between PC and EC. CH₄ detected before this point is attributed to carbon originating from OC, and that detected after this point is from EC.

Profiles of temperature, laser transmittance, and FID signals for an entire cycle of a TOT measurement are presented in Figure 3 along with the split point between OC and EC. Control experiments were first performed with only SR-NOM (Figure 3a) and only MWNT (Figure 3b) samples. The sample containing only SR-NOM was prepared by placing 1 mL of 500 mg/L SR-NOM stock solution on top of the quartz filter without suction and drying at 90 C. The sample with only MWNT was prepared by adding MWNT to Milli-Q water, retrieving an arbitrary fraction onto the quartz filter, and drying it overnight at 90 C. Therefore, the exact concentra tions of MWNT were unknown, and quantitative comparison was not made for these samples. Nevertheless, these experi ments confirmed that signals from these different carbon classes do not overlap, allowing for quantitative differentia tion between MWNT and SR-NOM. During the analysis of the SR-NOM-only control (Figure 3a), peaks were generated in both the first and the second stages of analysis, and the second stage peak appeared to be due to the generation of PC, as the concentration estimated from the sum of two peaks agreed with a concentration measured as DOC. For the MWNT-only sample (Figure 3b), peaks were observed only in the second stage as EC (with O₂ present). When the sample specimen after one cycle was subject to another entire TOT cycle, no further change in laser transmittance was observed, and no more CH₄ was produced, confirming that MWNT conversion to CH₄ was complete, consistent with observations reported in the literature under similar combustion conditions (53, 54).

Figure 3 Representative TOT thermograms for (a) SR-NOM-only, (b) MWNT-only, and (c) MWNT associated with SR-NOM.

A representative TOT thermogram for the MWNT and SR-NOM adduct is shown in Figure 3c. Following this direct quantification method, concentrations of suspended MWNTs in synthetic solutions, prepared according to an orthogonal matrix of varying initial SR-NOM concentrations (10, 25, 50, and 100 mg C/L) and varying MWNT mass initially added (50, 100, 250, and 500 mg/L), were analyzed. Results summarized in Table 1 demonstrate that approximately 0.25-1.4% MWNTs initially added to the SR-NOM solution became suspended for the range of solution compositions investigated. The fraction of suspended MWNT to the initial mass increased as more NOM was available. However, NOM availability was certainly not a limiting factor since the concentration of suspended MWNT also increased as the initial MWNT dose increased for the same NOM concentra tion. This observation suggests that NOM association with MWNT is a result of a dynamic equilibrium process, similar to typical adsorbate-adsorbent interactions.

For the same set of samples, the concentration of SR-NOM that was not associated with MWNT was directly measured using absorbance of the filtrate (i.e., after removing MWNT and associated SR-NOM) at 254 nm. Because of the competitive adsorption of different fractions of SR-NOM onto MWNT, some bias can be involved in the quantification of residual SR-NOM with UV₂₅₄ absorbance. The mass of SR-NOM associated with MWNT (per unit mass) was further calculated by subtracting the SR-NOM concentration in the filtrate from the initial SR-NOM concentration and dividing the value by the initial MWNT concentration. For the calculation, homogeneous adsorption of SR-NOM onto the MWNT was assumed. The mass of SR-NOM associated with MWNT (Table 1) was observed to increase as the relative abundance of SR-NOM increased (i.e., SR-NOM association increased as the initial concentration of SR-NOM increased and/or the initial mass of MWNT added decreased). These results, taken with the TOT results, suggest that the MWNT-NOM suspension formation has two aspects of consideration: (i) the physical dispersion of the MWNT added to solution and (ii) the association/equilibrium processes of NOM to the surface of MWNTs rendering them stable in the aqueous phase.

A spectral analysis of MWNT in the SR-NOM solution showed a distinct, broad increase in the baseline of absorbance spectrum in the visible range (over 500 nm), which was not observed in the solution containing only SR-NOM. Since the increase in baselines appears to be a function of light scattering by MWNT suspension and no specific absorption peak was identified, the absorbance at 800 nm was arbitrarily selected and plotted against the suspended MWNT concentrations determined from TOT analyses, which resulted in a linear correlation (r² = 0.987) (Figure 4). Similarly, simple turbidity measurements resulted in a reasonable correlation (r² = 0.952) with the suspended MWNT concentration (results not shown).

Figure 4 Comparison of light absorbance of MWNTs dispersed in SR-NOM solution at 800 nm and concentration of suspended MWNTs measured by TOT.

The stability of MWNT was further investigated using an actual Suwannee River water sample (collected in situ as described previously) to exclude any potential artifacts that might have originated from the use of a model compound. The pH, conductivity, and DOC of the 0.2 m filtered sample were 3.42, 69.4 S, and 59.1 mg C/L, respectively. Figure 1e shows 500 mg/L MWNT added to Suwannee River water as compared to filtered Suwannee River water (Figure 1f). This picture was taken after agitation for 1 h and quiescent settling for 4 days. Similar to the model SR-NOM, the Suwannee River water quickly dispersed MWNT, and the resulting suspension was stable for over 1 month. When the initial amount of MWNT added to the Suwannee River water was varied at 500, 250, 100, and 50 mg/L, the concentration of suspended MWNT was determined at 6.9, 5.45, 2.27, and 1.76 mg/ L, respectively, which was consistent with observa tions made with model SR-NOM in that the suspended MWNT increased as the initial MWNT dose increased. The amount of NOM adsorbed per unit mass of MWNT was 0.033, 0.042, 0.060, and 0.104 mg C/mg for initial MWNT concentrations of 500, 250, 100, and 50 mg/L, which were also in reasonable agreement with the results obtained with the model NOM (Table 1).

The observed similarity in the dispersion nature of the MWNT in solutions containing NOM and SDS suggests that MWNT stabilization in the presence of NOM might follow a similar stabilization mechanism of MWNT surface shielding by these molecules, which not only leads to more thermodynamically favorable surfaces but also induces electrostatic and steric stabilization (27, 30, 35, 36). Consequently, an amphiphilic, surfactant-like fraction of NOM with nonpolar groups coexisting with polar, charged groups might play a critical role. It is noteworthy that NOM appeared to be a better stabilizing agent than SDS as is clearly demonstrated in Figure 1. TOT analysis also suggested that approximately 1.78 mg/L MWNT would be suspended when 200 mg/L MWNT was added to a 1% (10 000 mg/L) SDS solution. This value was approximately 3 times lower than that of a solution containing 100 mg C/L of SR-NOM, which was only 1/100 of the compared SDS concentration by mass.

The enhanced stabilizing propensity might be attributed to the presence of aromatic fractions of NOM, as compared to SDS which is aliphatic, as aromaticity and resulting - interactions have been identified as an important parameter in MWNT stabilization by various surfactant molecules (34, 39). Generally, ubiquitous aromatic fractions of NOM can range from ca. 10-40% (C/C) depending on the source and age, with the SR-NOM composition estimated to be 23% aromatic carbon via ¹³C- NMR analysis (55). Furthermore, a relatively high percent of charged functional groups, such as carboxyl moieties (20% carbon as carboxyl for SR-NOM (55)), might contribute to enhanced dispersion of resulting NOM-MWNT complexes. However, the exact mechanism for CNT interaction with NOM will depend on both NOM characteristics including aromaticity, charge density, and size as well as CNT characteristics such as aspect ratio (e.g., SWNT) and functional derivatization (e.g., through a commonly used acid treatment that induces tube shortening and end-group carboxylation as a way of stabilizing SWNT in the aqueous phase (36)). Understanding these interactions presents a challenge, especially as NOM is largely undefined and variable depending on the source, warranting further in-depth investigations.

Acknowledgment

This research was supported by the United States Environmental Protection Agency (U.S. EPA) STAR Grant D832526. The authors acknowledge Jaekyu Cho at the School of Chemical and Biomolecular Engineering at the Georgia Institute of Technology and Sangil Lee and Evan Cobb at the School of Civil and Environmental Engineering at the Georgia Institute of Technology for their assistance during some of the instrumental analyses.