Environ. Sci. Technol. 2008, 42, 5917–5923
Tannic Acid Adsorption and Its Role for Stabilizing Carbon Nanotube Suspensions D A O H U I L I N †,‡ A N D B A O S H A N X I N G * ,‡ Department of Environmental Science, Zhejiang University, Hangzhou, 310028, China, and Department of Plant, Soil and Insect Sciences, University of Massachusetts, Amherst, Massachusetts 01003
Received February 1, 2008. Revised manuscript received May 27, 2008. Accepted June 9, 2008.
Dissolved organic matter (DOM) has been reported to stabilize carbon nanotube (CNT) suspensions, which increases concern over the subsequent transport and behavior of CNTs. However, it is unknown exactly which compounds or functional groups cause the stabilization of CNTs in natural environments. Naturally occurring tannic acid (TA), which has a large number of aromatic functional groups, was used as a surrogate of DOM to investigate its interaction with CNTs. CNT suspendability in TA solution increased with increasing CNT diameter without the aid of sonication. Sorption affinity of CNTs for TA increased with decreasing CNT diameter, positively related to their surface area. A two-stage sorption model was proposed to illustrate the interaction between CNTs and TA. TA molecules may be adsorbed first onto CNTs with aromatic rings binding to the surface carbon rings via π-π interactions, until forming a monolayer; the TA monolayer then further sorbed the dissolved TA by hydrogen bonds and other polar interactions. The sorbed TA increased the steric repulsion between individual CNTs, which might disperse the relatively loose CNT aggregates and result in the stabilization of largediameter CNTs in TA solution. The sorption and suspending process were also examined by transmission electron microscopy, providing further evidence for the above proposed CNT-TA interactions. This study implies that widely distributed TA may promote the mobility and transport of CNTs in natural aqueous environments.
Introduction Carbon nanotubes (CNTs), with outer diameters in the nanometer range (ca. 1-100 nm) and lengths up to several tens of micrometers, are pure carbon macromolecules consisting of sheets of carbon atoms covalently bonded in hexagonal arrays that are seamlessly rolled into a hollow cylindrical shape. Since their discovery in 1991 (1), CNTs have stimulated intense interest in their unique and outstanding electronic, mechanical, and chemical properties and their potential applications such as energy conversion, quantum nanowires, catalyst supports, and biomedical use. With increasing production and potentially wide applications of CNTs, their environmental behavior and ecological risk are attracting great attention (2, 3). * Corresponding author tel: (413) 545-5212; fax: (413) 545-3958; e-mail:
[email protected]. † Zhejiang University. ‡ University of Massachusetts. 10.1021/es800329c CCC: $40.75
Published on Web 07/02/2008
2008 American Chemical Society
Engineered CNTs can potentially enter into aqueous environment through wastewater streams and runoff from areas around a manufacturing plant, waste dump, or other places where they are used or discarded. CNTs are extremely hydrophobic and prone to aggregation and deposition in water as they are subject to high van der Waals interaction forces along the length axis. However, stable CNT suspensions can be achieved by modifying the tube surface property with the addition of dispersants (4) such as sodium dodecylbenzene sulfonate (NaDDBS), sodium dodecyl sulfate (SDS), Triton X-100, and polyvinyl pyrrolidone (PVP). These surfactants or polymers not only create a thermodynamically suitable surface in water but also provide steric or electrostatic repulsion among dispersed CNTs, thus preventing aggregation. However, it has been pointed out that surfactant by itself is not capable of effectively dispersing CNT bundles without vigorous sonication (5). Sonication plays a key role in the stabilization of CNT suspensions by gradually disentangling and exfoliating CNTs from their aggregates and bundles. Though the exact dispersion process still remains unclear, an “unzippering” type of mechanism (6) has been developed. It is postulated that gaps or spaces can be created at bundle ends in the high shear environment of ultrasonicated solution. Surfactant adsorption and diffusion then propagate this space along the bundle length, thereby separating the individual nanotubes. To date, there is limited information available on the modes of dispersion/aggregation of the nanomaterials in natural aqueous environments. Recently, it was reported that one CNT could interact with dissolved organic matter (DOM) and be stabilized in Suwannee river water (7). DOM extracted from Sahan River in Ukraina was also found enhancing the stabilization of fullerene in water (8). These two studies imply that hydrophobic carbon nanomaterials may be transported rather than aggregate and deposit quickly after their release into aqueous environments, thus leading to potential exposure and ecotoxicity. However, it remains unclear exactly which structural components and/or functional groups of DOM cause the dispersion and stabilization of carbon nanomaterials. In addition, there are many types of CNTs, and DOM can differ markedly from location to location. Different types of CNTs and DOM have not been examined before. Natural surface-active materials, such as humic acid (8) and tannic acid (TA) (9), are widely distributed in the environment. They may adsorb onto individual CNTs, and thus alter their surface physicochemical properties and enhance their stabilization in water. Therefore, we selected TA, which has a simpler chemical structure than humic acid, as a model DOM to investigate its interaction with CNTs without using sonication.
Materials and Methods Materials. CNTs were purchased from Shenzhen Nanotech Port Co., China. They are one single-walled carbon nanotube (SWCNT) with claimed outer diameter of 1-2 nm and five multiwalled carbon nanotubes (MWCNTs) with claimed outer diameters of 50 >90h >95 >95 >95 >95 >95
5-15
1.4 (0.4
541
0.820
0.201
1.83
1-2 1-2 1-2 1-2 1-2
9.4 (1.8 20.9 (3.0 27.8 (6.0 42.7 (6.4 70.1 (9.5
357 126 86 73 58
0.951 0.364 0.285 0.162 0.114
0.142 0.051 0.034 0.029 0.023
0.29 0.15 0.14 0.03 0.09
g
elemental contentf (%) C
H
O
4.56
91.84
0.10
3.50
2.97 0.76 1.64 2.09 2.06
96.41 98.09 98.15 97.68 98.01
0.42 0.25 0.19 0.14 0.11
0.20 0.90 0.02 0.09 -
Provided by the supplier. b Measured by TEM, n ) 100. c Surface area (Asurf), mesopore volume (Vmeso), and micropore volume (Vmicro) were calculated from the adsorption-desorption isotherm of N2 at 77 K by multipoint BET method. d Water content was measured by drying the CNTs at 105 °C for 24 h. e Ash content was measured by heating the CNTs at 900 °C for 10 h. f Dry-weight-based elemental contents of the CNTs were determined using a Vario ELIII elemental analyzer (Elementar, Germany); O contents were calculated by mass difference. g SWCNT content. h CNT content. CNT is carbon nanotube; SWCNT is single-walled CNT; MWCNT is multiwalled CNT; the numbers after MWCNT are their outer diameters. a
volumes using the multipoint Brunauer-Emmett-Teller (BET) method (10). Dry-weight-based C, H, and N contents of the CNTs were determined using a Vario ELIII elemental analyzer (Elementar, Germany) with the oxygen content calculated by mass difference. Water contents were measured by drying the CNTs at 105 °C for 24 h. Ash contents were determined by heating the CNTs at 900 °C for 10 h. Functional groups of the CNTs were analyzed by diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) using a Perkin-Elmer Spectrum One spectrometer with a DRIFTS accessory (Spectros Instruments, Whitinsville, MA) following the procedure of Lin et al. (11). Ultrapure TA was from Alfa Aesar (Ward Hill, MA) with a structure characterized as five digallic acid units esterlinked to a glucose core (Figure S1, Supporting Information). It has a large number of phenolic hydroxyl groups with a molecular weight of 1701 (C76H52O46) and a solubility of 300 g/L in water. Due to its natural origin and wide distribution (12), TA has been widely used as a surrogate or model compound of DOM in environmental studies (13–15). Sorption and Stabilization Experiments. Sorption isotherms were obtained using a batch equilibration technique at 25 ( 1 °C. Eight mg of CNTs were added into 40-mL vials with 40 mL of TA solutions with initial concentrations of 0, 5, 10, 20, 50, 100, 200, 300, and 500 mg/L. Each concentration point, including blanks (i.e., without CNTs) was run in duplicate. The vials were sealed with aluminum-foil-lined Teflon screw caps and were shaken (120 rpm) for 7 days. Preliminary experiments indicated that apparent equilibrium of both sorption and suspension was reached within 7 days (Figure S2, Supporting Information). After equilibration, the vials were centrifuged at 3000 rpm for 20 min, and the resulting supernatants, the stable suspensions possibly with dispersed individual nanotubes, were taken out and measured with a UV-vis spectrometer (Agilent 8453, USA) at 800 nm. Measurement at 800 nm has been used to quantify CNTs in aqueous phase (5, 7). We have observed that tannic acid has no absorbance at 800 nm, and there was a good correlation between the absorbances at this wavelength and CNT concentrations. Therefore, we used the absorbance at 800 nm of CNT suspensions to compare the suspendability of the same CNT in various TA solutions and of different CNTs in the same TA solution. Preliminary experiments indicated that the deposition of CNT suspensions by centrifuging at 3000 rpm for 20 min was more than that by settling for 4 days. The 4-day period was used previously as the duration to examine the stabilization of CNT in an aqueous phase (7). Thus, the suspended CNTs in the supernatants after centrifuging at 3000 rpm for 20 min were arbitrarily considered to be stable in TA solutions. 5918
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 16, 2008
TA remaining in the supernatants was quantified by the absorbances at 275 nm after being filtered through 0.2-µm poly(tetrafluoroethylene) (PTFE) filters (Whatman). No absorbance at 800 nm of the filtrates was observed, indicating that the CNTs could be effectively removed from the supernatants by the filter. The solute loss, TA concentration differences in the blank solutions before and after the 7 day equilibration, were less than 5% of the initial concentrations, therefore, sorbed solute (TA) concentrations were determined by mass balance. CNT suspensions were further examined with TEM. The samples were prepared by air-drying a drop of the suspensions onto a copper TEM grid (Electron Microscopy Sciences, USA). Zeta potentials and colloid sizes of the CNT suspensions were measured with a ZetaSizer (Malvern Instrument Ltd., UK). The pH of pure TA solutions and TA-CNT mixtures was measured. The CNTs after TA sorption were washed and dried at 105 °C for 24 h, and then examined by DRIFTS. The dried TA-CNT complexes were again dispersed in water and observed by TEM. Statistical Analysis. Commonly used nonlinear isotherm models (Table S1, Supporting Information) were employed to fit the sorption experiment data. All estimated model parameter values and their probabilities of assuming the null hypothesis (p) were determined by a plotting software program (SigmaPlot 9.0). Statistical significance was accepted when p was less than 0.01. The goodness of fit was evaluated by the fitting parameter adjusted square of correlation coefficient (adj r2, also given by the SigmaPlot) which takes into account the number of independent variables reflecting the degrees of freedom.
Results and Discussion Characterization of Carbon Nanotubes. Characteristics of the CNTs were listed in Table 1. The morphologies (lengths and outer diameters) of the CNTs examined by TEM are consistent with that provided by the supplier. The surface area and pore volume increased with decreasing tube diameter. However, mesopore volume of SWCNT was less than that of MWCNT10. Meanwhile, SWCNT contained more impurities (ash and water) and oxygen, and less carbon than MWCNTs. Elemental compositions of MWCNTs were quite similar. Practically no nitrogen and very low hydrogen content could be detected in the CNTs. Low or no oxygen content was found for MWCNTs. The high temperature for ashing may result in overestimation of the ash contents due to the oxidation of residual catalysts in the CNTs, which may lead to underestimation of the oxygen contents. No large difference in DRIFTS spectra (Figure S3, Supporting Information) of the CNTs was observed. Strong graphitic structure of the sp2-hybridized carbon in the CNTs can be identified, while
FIGURE 1. Sorption isotherms of tannic acid (TA) by CNTs (A) and relationship of the monolayer adsorbed capacity with the surface area of CNTs (B). Solid lines in panel A are the isotherms fitted by DMM, among which the isotherms of MWCNT20 (red line) and MWCNT40 nearly overlap due to their similar sorption affinity. CNT is carbon nanotube; SWCNT is single-walled CNT; MWCNT is multiwalled CNT; the numbers after MWCNT are their outer diameters. the carboxylic functional group is not significant with a small shoulder (if any) of CdO bond (1710 cm-1) observed. Sorption. The sorption isotherms of TA on CNTs are shown in Figure 1A. At a given equilibrium concentration of TA, sorption increased in the order of MWCNT10 > SWCNT > MWCNT20 > MWCNT40 > MWCNT60 > MWCNT100. The TA sorption by MWCNTs generally decreased with increasing CNT diameters. Slightly lower carbon content and higher ash and oxygen contents may account for the lower sorption affinity of SWCNT than MWCNT10. However, the lower sorption of SWCNT than MWCNT10 might mainly come from its greater aspect ratio which could result in more entanglements and less sorption sites for TA. BET results showed less mesopore (accessible for TA) volume of SWCNT than MWCNT10. The TA sorption by these CNTs is higher than the organo-clays (16, 17) and comparable to or even higher than the activated carbons (18, 19) in terms of Langmuir sorption capacity or Freundlich coefficients. All sorption isotherms are apparently nonlinear (Figure 1A). Therefore, commonly used nonlinear isotherm models (Table S1), i.e., Freundlich (FM), Langmuir (LM), DualLangmuir (DLM), and Dual-mode (DMM) models, were tested to fit the experimental data. The fits of DMM to the sorption data are displayed in Figure 1A as examples. The fits of other models are shown in Figure S4. Respective fitting parameters are given in Table S2. Significant deviation of model estimation from the experimental data was observed for LM at high concentrations (Figure S4A), especially for MWCNTs with larger diameters. The adj r2 values of LM, ranging from 0.809 to 0.971, are the lowest among the four fitting models. A similar deviation was also observed when LM was used to fit the sorption isotherms of PAHs onto CNTs (10). The LM was originally developed to describe individual chemical adsorbents, and is applicable to physical adsorption (monolayer) within a low concentration range (16). The deviation indicates that the sorption of organic compounds, such as TA and PAHs, on these CNTs may not be monolayer formation on a homogeneous surface. The FM is an empirical approach for adsorbents with uneven adsorbing site energy, and is applicable to adsorption for a single-solute system within high- and middle-concentration environments (16). For FM, the goodness of fit varied with the CNTs (Figure S4B). The fit of FM for MWCNT20, MWNCT40, MWCNT60, and MWCNT100 was better than LM, while it was worse for SWCNT and MWCNT10. Good fit was obtained for DLM (Figure S4C) with adj r2 ranging from 0.978 to 0.998. However, two parameters among the four, i.e., adsorbed capacity (Q02) and DLM constant (b2) of site population 2 were significantly unreliable with p . 0.01, particularly for MWCNTs with larger diameters. Thus, the sorption process of TA by CNTs may not be limited by the two types of adsorption sites. The DMM had not only good fit for the experimental data by visual
examination (Figure 1A), but also high adj r2 (0.968-0.997) and reliable estimated parameters with p < 0.01 (Table S2). DMM was originally proposed for describing the sorption behavior of natural organic matter with both partition and Langmuir-type adsorption domains (20). Apparently, CNTs may not have a partitioning phase due to the limited amount of amorphous carbon. The inner pore of CNTs is usually closed from both of their ends, and inaccessible to small molecules of N2 or PAHs (21). TA with a molecular size of 1.6 nm (18) is unlikely sorbed to the CNTs by pore filling; similarly, TA was unable to get into the interlayers of organic clays (17). Two-stage sorption process was, thus, postulated as a model for the interaction between TA and CNTs through π-π interactions first and then polar interactions between TA molecules (Figure 2). Zhu and Pignatello (22) reported that π-π interaction was the main mechanism regulating adsorption of both π-acceptors and π-donors on the graphene surface. With a rolled hollow cylindrical graphite-like structure, CNTs were also found acting in an amphoteric role toward adsorbates of both π-acceptors and π-donors via π-π interaction (23–25). In addition, stacking of benzene rings onto the CNT surface was reported to greatly increase the surface coverage of surfactants and polymers (26, 27). Molecular structure of TA consists of three digallic acid units standing up onto one side of a glucose core while the other two digallic acid units directing to the other side (Figure S1). Thus, the sorption may begin with TA molecules adsorbing onto CNT wall with benzene rings anchoring to the surface carbon rings of CNTs via π-π interactions, until forming a monolayer. Because of stiff chemical bonds or restricted configuration flexibility, only 3 or 2 outer benzene rings in the 3 or 2 digallic acid units at one side of the glucose core, rather than all 10 benzene rings, may bind to the cylindrical-shape surface of CNTs (stage 1, Figure 2A). Good fit of LM in this stage (Figure S4A) indicates that it was a Langmuir-type adsorption in stage 1. The adsorption capacity in this stage would be mainly controlled by TA molecular size and the available CNT surface area. The maximum surface adsorption capacity Q0 estimated from DMM is positively related to the surface area of MWCNTs measured by the N2 method (multipoint BET) with r2 ) 0.986 (Figure 1b). SWCNT was excluded from Figure 1b for its large deviation from the linear relationship, possibly due to its high content of impurity and extremely high aspect ratio which could lead to tight agglomeration and reduced availability of surface area for TA adsorption. The estimated sorption capacity of CNTs from DMM is 1.1-1.7 times lower than the calculated monolayer adsorption capacity (Q, mg/ g) (Figure 1b). The calculation equation is Q ) S/(3.14 R2) × W/NA × 1021, where S (m2/g) is the surface area of CNT, R is the radius of projected area (assumed to be spherical) of VOL. 42, NO. 16, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5919
FIGURE 2. Schematic model for the interaction between carbon nanotube (CNT) and tannic acid (TA). TA molecule anchored onto the tube wall with benzene rings binding to the surface carbon rings via π-π interactions to form a monolayer (A); the adsorbed TA monolayer sorbed more dissolved TA molecules (B). Thick line in the figure is a chemical bond linking to a pyrogallol which interacts with the back side of CNT. a TA molecule onto the tube wall (0.8 nm), W is the molecular centrations. No micelle was formed at the experimental weight of TA (1701), and NA is the Avogadro constant. The concentrations of TA according to the surface tension overestimation of calculated capacity may stem from the measurements (Figure S5), which is consistent with the assumption that the CNT surfaces, available for N2 molecules reported result that no micelle could be observed for TA with during the surface area measurement, were fully covered by concentration lower than 3 × 10-4 M (9). The thickness of adsorbed TA layer on the CNT surfaces increased with TA molecules. However, the CNTs aggregated in the TA increasing TA concentration, which could be directly obsolutions, hence the macromolecules of TA may not be able served with TEM (Figure 3). At low concentration (5 mg/L), to reach the adsorption sites inside the CNT aggregates. the adsorbed TA onto individual CNTs was not obvious It has been widely recognized (28, 29) that the adsorbed (Figure 3B), while opaque layers covering the areas crowded surfactant can become a partitioning phase on sorbents for with CNTs indicate the strong interaction between CNTs both hydrophobic and hydrophilic organic compounds. The and TA. The TA coverage at 500 mg/L was so thick that hollow surfactant-derived organic phase is 10-30 times more cores of individual CNTs could not be observed (Figure 3C). effective on a unit weight basis than natural soil organic The drying process on TEM grid may lead to the coating of matter for organic compounds (28). Sorbed cationic surdissolved TA on CNTs, however, the water-washed TA-CNT factant at low levels behaves as a more powerful medium for complexes still showed similar increasing TA coverage (Figure sorbing organic compounds than the dissolved surfactant in S6), verifying the adsorption of TA on CNTs, consistent with micellar form (29). The adsorbed TA monolayer may continue our sorption isotherm data. DRIFTS results (Figure S7) also to sorb the dissolved TA molecules via hydrogen bonds and showed the adsorption of TA on the CNTs. CNTs after sorption other polar interactions (stage 2, Figure 2B). This stage would of TA showed stronger DRIFTS peaks of functional groups be controlled by the sorption affinity of adsorbed TA of -OH and tetrasubstituted benzene ring, both of which monolayer. This sorption affinity, indicated by sorption are typical for TA molecules. coefficient (Kd) in the DMM equation, increased with reducing CNT diameters (Table S2). The linear type sorption isotherms Stabilization. Capability of surfactants and polymers to at high TA concentrations indicate that the sorption capacity stabilize CNT suspensions and the influencing factors have of the adsorbed TA monolayer was not saturated at the TA been investigated (4). A few studies have been dedicated to concentration range used. Sorption behavior of surfactants development of models for explaining the role of dispersants usually changes greatly around the critical micellar conin suspending CNTs (5, 6, 26). However, limited information 5920
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 16, 2008
FIGURE 3. Representative TEM images of multiwalled carbon nanotubes with outer diameter of 60-100 nm (MWCNT100) in the background solution without tannic acid (TA) (A), and the supernatants of TA-MWCNT100 suspensions with initial TA concentrations of 5 mg/L (B) and 500 mg/L (C). Bundles of MWCNT100 can be observed in pure CNT suspension (A). Hollow cores of MWCNT100 in 5 mg/L TA solution are clear (the insert of panel B), while the hollow cores cannot be observed in 500 mg/L TA solution because of the thick coverage of TA (the insert of panel C).
FIGURE 4. Photo of 5 mg/L tannic acid (TA) solutions after adding 200 mg/L carbon nanotubes (CNTs), shaking for 7 days, centrifuging at 3000 rpm for 20 min, and settling for 30 days. CNT100, CNT60, CNT40, CNT20, and CNT10 are multiwalled CNTs with outer diameters of 60-100, 40-60, 20-40, 10-20, and
