Adsorption uptake of synthetic organic chemicals by carbon nanotubes and activated carbons

A number of different compounds were selected to compare uptake between traditional activated carbon and carbon nanotubes, and to determine the role that π–π interactions play in their uptake by CNTs. The physical and chemical properties of these compounds are tabulated in table 1, and compound structures are tabulated in supplemental information (available at stacks.iop.org/Nano/23/294008/mmedia). Atrazine (2-chloro-4-ethylamino-6-isopropylamino-s-triazine) is one of the most frequently applied and widely used herbicides in the world (Graymore et al 2001, Lazorko-Connon and Achari 2009, Chen et al 2009). Once applied in the environment, atrazine leaches into groundwater and surface water, and has thus been widely detected in drinking water supplies. It is recognized as an endocrine disruptor, toxic to humans and aquatic organisms (Graymore et al 2001, Lazorko-Connon and Achari 2009). The USEPA has suggested that exposure to the pesticide was associated with several birth-related health risks (Erickson 2010). Trichloroethylene (TCE) is also a common groundwater and surface water pollutant, and is a suspected carcinogen (Lavin et al 2000). These contaminants are representative of a wide variety of pollutants that are frequently identified in both surface and ground water drinking water sources, and for this reason, they have been chosen for this study.

To improve our understanding of π–π interactions between adsorbate aromatic rings and the CNT surface, which are expected to depend on the electron density in the carbon and in the aromatic structure of the organic compound, four compounds possessing significantly different π electron densities were selected: cyclohexene, benzene, naphthalene, and phenanthrene. Cyclohexene and benzene have one and three double bonds, respectively, which are expected to promote π–π interactions. Both naphthalene and phenanthrene are polycyclic aromatic hydrocarbons containing two and three rings, respectively. The adsorption trends of these molecules will provide insight into the significance of π–π interactions and the significance of hydrophobic versus π–π interactions during adsorption.

Four additional compounds were selected in order to examine the effects of methyl group and chlorine substitution in aromatic compounds, and to assess the effects of electron withdrawing and donating groups on π–π interactions: chlorobenzene, 1,4-dichlorobenzene, methylbenzene (toluene), and 1,4-dimethylbenzene (p-xylene). The presence of electron withdrawing groups on the benzene ring (e.g. –Cl) would be expected to increase the interactions of the adsorbate with carbon surfaces having high electron density, whereas the presence of an electron donating group (e.g.–CH₃) would exhibit the opposite trend. The inductive and resonance effects of a substituent can be quantified by their Hammett constants, which are tabulated in table 2.

Nine different carbon sorbents were used to define the range of performance available for these materials, and to compare uptake of activated carbons and carbon nanotubes for compounds of interest in drinking water treatment. These adsorbents were used, as received, to simulate how they would be employed in practice. Elemental analyses of the adsorbents were performed by Huffman Laboratories, Inc. (Golden, CO) and are given in table 3. Additional physical and chemical properties of the CNT materials are tabulated in table 4. The activated carbons include coal based granular materials (TOG, Calgon) from two different batches and having two different particle sizes, 80 × 325 (TOG80 × 325) and 20 × 50 (TOG 20 × 50). Zimmerman et al (2005) report an elemental carbon content of 87% and a specific surface area of 938 m² g⁻¹ for this material. A wood-based powdered carbon (Aqua Nuchar PAC, MeadWestvaco, Corington, VA USA) was also investigated. The manufacturer reports that this carbon is phosphoric acid activated followed by a hot water wash, which yields a mesoporous porosity (a predominance of pores in the 2–50 nm range) and a 0.4 ml g⁻¹ micropore volume. The acid activation results in a slightly acidic surface, having a pH of 4–6. The manufacturer reports a specific surface area of 1200 m² g⁻¹ and an ash content of less than 10%. Three different SWNTs were used with different degrees of purity.

Cheap Tubes nanotubes (Cheap Tubes, Brattleboro VT, USA) were made using the catalyzed chemical vapor deposition process (CVD). The as-produced material is purified by the manufacturer to a reported purity of 90%, a MWNT content of 5% and an amorphous carbon content of 3%. Jubete et al (2011) reported characterization data for the Cheap Tubes SWNTs. Raman spectra showed a strong tangential G band at ∼1580 cm⁻¹. Broad features at about 1345 cm⁻¹ are due to the disordered sp² carbon D-band of non-nanotube graphitic components, and indicates the presence of impurities. For this reason, the ratio of the amplitudes of the G-band and the D-band has been used as a measure of the purity (Itkis et al 2005). A low D : G peak intensity ratio of 0.06 was reported for the Cheap Tubes material. A bimodal G'-band was observed at ∼2680 cm⁻¹, suggesting a mixture of several types of SWCNTs in the starting material.

Carbon Solutions (Carbon Solutions, Inc., Riverside, CA, USA) nanotubes were electric arc synthesized using a Ni/Y catalyst. Carbonaceous purity of the as-produced materials is reportedly between 40 and 60%, with a metal content of 30%. Individual tube lengths range from 0.5 to 3 µm and have an average diameter of 1.4 nm. These SWNTs tend to occur as bundles with lengths of 1–5 µm and average bundle diameters of 2–10 nm. The density of the AP-SWNT is 1.2–1.5 g cm⁻³, and the ratio of semiconducting to metallic SWNTs produced using the electric arc discharge method is 2–1 (manufacturer's data).

Carbolex SWNTs obtained from Sigma-Aldrich were also produced by the electric arc method, with a diameter range of 1.2–1.5 nm (manufacturer's data). The carbonaceous purity of these nanotubes is reportedly between 50 and 70%; a metal content of <33% was reported by the manufacturer.

The NanoLab multi-wall nanotubes (NanoLab, Inc., Waltham, MA, USA) were produced by CVD to various length and diameter specifications, and were purified to >95%, as measured by TGA. Residuals may contain iron and sulfur, and the reported specific surface area is in the range of 200–400 m² g⁻¹. Three samples were investigated to probe effects of the MWNT length (L) and diameter (D), having D = 15 ± 5 nm, L = 1–5 µm (PD15L1-5); D = 30 ± 15 nm, L = 1–5 µm (PD30L1-5); or D = 30 ± 15 nm, L = 5–10 µm (PD30L5-20).

To investigate the role of π–π interactions, a SWNT (Carbon Nanotechnologies, Inc., Houston, TX, USA) was used. These were synthesized by high pressure CO conversion synthesis (HiPCO) and were purified before use. First, the SWNTs were heat treated in air at 300 °C for 30 min to reduce amorphous carbon, and facilitate the separation of the nanotubes from amorphous carbon coatings via surface cracking. This was followed by soaking in 18 wt% HCl for 20 min to solubilize the metals used as catalysts.

Isotherm experiments were conducted (1) to compare different CNTs and to compare CNTs with activated carbon, and (2) to investigate π–π interactions between CNTs and adsorbates having various electron donor/acceptor capabilities. Constant carbon dose aqueous adsorption isotherm experiments were conducted in completely mixed batch reactors (CMBR, 250 ml amber colored glass bottles with Teflon-lined screw caps). Each point of the adsorption isotherm represents an individual experiment in one CMBR. Stock solutions of each adsorbate were prepared in methanol. For single solute experiments, approximately 10 mg of carbon adsorbent (nanotube or activated carbon) was carefully added to each reactor, which was then filled approximately halfway with 1.0 mM phosphate buffer solution made with MQ water containing 100 ppm NaN₃. The pH of each CMBR was then adjusted to 5.7, 7.0, or 8.3 using HCL and NaOH, and the ionic strength was adjusted to 0.01 M using NaCl. Adsorbate was then added to each reactor via a microliter aliquot of stock solution. The bottle was then carefully filled without headspace with 1.0 mM phosphate buffer solution. The CMBRs were then placed on a jar mill roller/tumbler (Norton/St Gobain) and rotated end over and for a period of two weeks. Preliminary experiments indicated that this time was sufficient to reach an equilibrium conditions as measured by a lack of change in solution concentration with additional contact time (see supplemental information available at stacks.iop.org/Nano/23/294008/mmedia). At the end of the second week, the bottles were removed and allowed to sit for 24 h to allow the suspended carbon to settle out of the solution. Supernatant solutions were analyzed for TCE using gas chromatography (GC) with a μECD detector (Model 6890N, Agilent Technologies, Inc., Santa Clara, CA, USA). Atrazine was determined by high performance liquid chromatography (HPLC) (Model 1100 Agilent Technologies, Inc., Santa Clara, CA, USA) equipped with an UV-diode array detector set at a wavelength of 220 nm; the mobile phase was 60% methanol and 40% MQ water.

The organic compound sorption data was fitted using the Freundlich model:

$$\begin{eqnarray} &&{q}_{\mathrm{e}}={K}_{\mathrm{F}}{C}_{\mathrm{e}}^{n} \end{eqnarray} \tag{ 1 }$$

where q_e is the solid-phase concentration (µg —adsorbate/g—sorbent), C_e is the aqueous-phase concentration (mg—adsorbate/m³—water), K_F is a capacity factor ((µg g⁻¹) (m³ mg⁻¹)ⁿ), and n is an exponent related to the intensity of adsorption; lower values indicate a site energy distribution encompassing high energy sites (dimensionless). The Freundlich model has been widely used for describing the sorption of organic chemicals to carbonaceous materials such as activated carbon and CNT materials (Kilduff et al 1996, Yang et al 2006a) and other heterogeneous sorbents, including natural soils and sediments.

Single solute experiments were conducted to compare the uptake of atrazine by several different activated carbons and carbon nanotube materials; the results are plotted on an adsorbent mass basis in figure 1, and Freundlich isotherm parameters are tabulated in table 5. For the TOG 80 × 325 and Cheap Tubes SWNT isotherms, the data represent pooled data taken at three different pH values (pH 5.7, 7.0, and 8.3), and will be discussed in more detail below. For the other adsorbents, uptake was measured at pH 5.7. Over the concentration range examined, the three activated carbons exhibited higher uptake for atrazine when compared to the nanotube materials, although at higher concentrations the isotherms appear likely to converge. The smaller particle size TOG 80 × 325 carbon exhibited the highest uptake, followed by the TOG 20 × 50 and the Aqua Nuchar powdered carbon. The uptake was favorable, with Freundlich n-values between 0.30 (TOG 20 × 50) and 0.38 (Aqua Nuchar). The Cheap Tubes SWNT material performed the next best, with uptake lower than the activated carbons, but higher than the other nanotube materials at concentrations above about 500 mg m⁻³; however, the uptake by the Cheap Tubes SWNTs is similar to or lower than the other SWNTs at concentrations less than 100 µg l⁻¹. Additionally, the Cheap Tubes SWNTs had the highest Freundlich n-value of all adsorbents investigated, 0.54, although still quite favorable. The Carbolex and Carbon Solutions SWNTs exhibited similar uptake to each other, whereas the MWNT samples exhibited the lowest uptake, especially at concentrations less than 1000 mg m⁻³.

Zoom In Zoom Out Reset image size

The Cheap Tubes and Carbolex SWNTs have similar reported Raman D:G ratios (0.05 and 0.06, respectively); therefore, this parameter does not appear to correlate with adsorption from solution. The lower uptake (on a mass basis) of the Carbolex and Carbon Solutions SWNTs, as compared to the Cheap Tubes SWNTs, is most likely caused by their metal impurities leading to a lower carbon content (of the order of 60–70%, versus nearly 95% for the Cheap Tubes). The impurities contribute a significant mass but do not provide strong adsorption sites. Although the Nanolab MWNTs have high purity, it is likely that their concentric sheets contribute to adsorbent mass but are not accessible for adsorption. The MWNTs also have a significantly higher D:G ratio; this may indicate a greater degree of carbonaceous impurities, which are lower energy adsorption sites. Shown in greater detail in figure 2, uptake of atrazine by the MWNTs was comparable, and their Freundlich isotherms were statistically similar, providing evidence that there is little effect of the MWNT length or diameter on the uptake of atrazine over the range studied here.

Zoom In Zoom Out Reset image size

The potential effects of carbon content, surface area, and the combined effect of these were investigated by plotting the atrazine uptake normalized by these properties. The normalized isotherms are shown in figure 3, panels (A)–(C). The isotherms normalized by carbon content (panel (A)) appear quite similar to those in figure 1, with the exception that the normalized uptake increased significantly for the lower purity Carbolex and Carbon Solutions SWNTs. Still, the rank order of the adsorbents did not change appreciably, and the overall range of uptake among the different materials is about an order of magnitude. Normalizing by surface area (panel (B)) collapses the isotherms significantly, uptake is now within about a factor of two. Notably, most of the CNT data is comparable to or above the wood-based PAC. This could be explained by either a high surface area that is inaccessible to the adsorbate, a surface chemistry effect, or both. The acid activation of the wood-based material leads to oxygen containing functional groups on the surface which have been shown to reduce uptake (Kilduff and King 1997, Karanfil and Kilduff 1999, Karanfil et al 1999). Normalizing uptake by both carbon content and surface area collapses the data even further (panel (C)), indicating that these two characteristics drive many of the observed differences between activated carbons and CNT materials.

Zoom In Zoom Out Reset image size

The data plotted in figure 4 represents single solute experiments conducted at three different pH values (5.7, 7.0, and 8.3) for SWNT (Cheap Tubes) and for the TOG activated carbon; Freundlich model parameters are tabulated in table 6. As discussed above, on a mass basis, the activated carbon exhibited higher uptake for atrazine, as compared to the SWNT, over the range of concentrations examined. Also, the data plotted in figure 4 establishes that over the pH range from 5.7 to 8.3, there is no significant effect of pH on the uptake of atrazine by either the TOG 80 × 325 carbon or the Cheap Tubes SWNT material. This was verified by statistical analysis, quantified by overlapping 95% confidence intervals. Therefore, the Freundlich isotherm parameters reported in table 6 represent pooled data.

Zoom In Zoom Out Reset image size

Atrazine does not ionize over the pH range studied; therefore, changes in pH will not have significant effects on solute activity (fugacity) or solubility, and the lowest pH value is sufficiently high that sorption competition from H⁺ is not expected. Consequently, the only potential effect of pH is an effect on the adsorbent surface chemistry, such as ionization of surface functional groups. The results presented in figure 4 confirm that if such changes occur, they are negligibly small and subsequently have little effect on atrazine adsorption. This result was expected for TOG, because it is a 'basic' type activated carbon, meaning it is not exposed to air or other oxidizing gas during cooling after activation. Such carbons, in general, do not contain a high concentration of surface oxygen groups, in contrast to acid-activated wood-based carbons. The fact that pH had little effect on atrazine uptake by the SWNT suggests that, like the activated carbon, its surface does not contain a high concentration of ionizable groups. In this regard, it can be concluded that the surface chemistry of the SWNT and the (basic) activated carbon are similar.

Uptake by the TOG 80 × 325 activated carbon was well described by the Freundlich model; individual isotherms had r² values greater than 0.87, and pooled data yield an r² value of 0.91. The Freundlich model also fits the SWNT uptake data, but lower r² values were found due to greater variability inherent in the SWNT data; pooled data yield an r² value of 0.681. Because the same adsorbent masses were used, and the same protocols were followed, we conclude that the SWNT material is inherently more heterogeneous than the activated carbon material. One explanation for the lower uptake of atrazine by SWNTs on an adsorbent mass basis suggests that the material has a lower available surface area than the activated carbon. This may be due to a lower degree of microporosity, due to the formation of bundles or aggregates that reduce the available surface area, or both (Ji et al 2009a, 2009b, Wang et al 2010, Zhang et al 2009).

Single solute experiments were conducted at three different pH values (5.7, 7.0, and 8.3) to measure the uptake of TCE by SWNT (Cheap Tubes) and TOG 80 × 325 activated carbon. The data are plotted in figure 5; Freundlich isotherm model parameters are tabulated in table 7. As was observed for atrazine, there is no effect of pH on TCE uptake over the range studied, by either the TOG 80 × 325 or the Cheap Tubes SWNTs. The major potential effect of pH would be on the adsorbent surface chemistry, which could include ionization of surface functional groups. The results presented in figure 5 confirm that such pH effects, if present, are small and have a negligible effect on TCE uptake by either adsorbent. This is consistent with what was observed for atrazine, and further suggests that the surface chemistry of the Cheap Tubes SWNTs and the TOG active carbon are similar in their response to pH.

Zoom In Zoom Out Reset image size

The TCE uptake measured here was similar to that reported earlier for a similar water treatment carbon (Karanfil and Kilduff 1999, Kilduff and Karanfil 2002). The TOG activated carbon exhibited greater uptake than SWNT at low concentrations and comparable uptake at higher equilibrium concentrations. TCE uptake by the TOG activated carbon is well described by the Freundlich model, with an r² value of 0.889. The SWNT uptake is also well described by the Freundlich model with an r² value of 0.828; the lower correlation coefficient may be attributable to the greater variability that has been found in the SWNT data.

The potential effects of surface area were investigated by plotting TCE uptake normalized by this property, as was done for atrazine in figure 3. Because the carbon content of TOG GAC and Cheap Tubes SWNTs is similar, normalization by this parameter did not yield useful results. However, as shown in figure 6, after normalization by surface area, the data for TCE uptake by TOG GAC and Cheap Tubes SWNT correspond rather closely, consistent with our findings for atrazine.

Zoom In Zoom Out Reset image size

To assess specific interactions, uptake data, which was measured in the aqueous phase, was expressed in terms of concentration in an inert solvent (i.e. n-hexadecane). The inert solvent reference state minimizes differences in solute–solvent interactions for different solutes, thus enabling a comparison between sorbates in terms of their interactions with the sorbent (Borisover and Graber 2003):

$$\begin{eqnarray} &&{C}_{\mathrm{n}\text{-}\mathrm{hexa}}={C}_{\mathrm{aq}}\frac{{S}_{\mathrm{n}\text{-}\mathrm{hexa}}}{{S}_{\mathrm{aq}}} \end{eqnarray} \tag{ 2 }$$

where C_n-hexa is the solution concentration in an inert solvent (i.e. n-hexadecane) that would exist at the same fugacity as the measured aqueous-phase concentration; C_aq is the aqueous solution concentration; S_n-hexa and S_aq are the compound solubility in n-hexadecane and water, respectively. This approach is used to normalize differences in solubility and hydrophobic effects, and therefore provides a better comparison of interactions between the adsorbates and the sorbent surfaces. The ratio of solubilities could also be replaced by the ratio of air/water Henry's constant, H_w, and the air/hexadecane Henry's constant, H_n-hexa (Borisover and Graber 2003):

$$\begin{eqnarray} &&\frac{{S}_{\mathrm{n}\text{-}\mathrm{hexa}}}{{S}_{\mathrm{aq}}}=\frac{{H}_{\mathrm{w}}}{{H}_{i}}. \end{eqnarray} \tag{ 3 }$$

Equivalently, the normalization factor could be expressed as the n-hexadecane–water partition coefficient (K_HW,i (m³ m⁻³)) as done by Zhu and Pignatello (2005). This reveals more clearly the fact that the normalization factor is related to the change in chemical potential of a compound upon transfer from the infinitely dilute standard state in water to the infinitely dilute standard state in n-hexadecane (Borisover and Graber 2003). Freundlich model parameters for probe compounds are tabulated in table 8.

The uptake of model compounds, with the exception of chlorinated benzenes, is shown in figure 7. The data appear to cluster into two groups: over the range of concentration studied, benzene and toluene comprise one group, and exhibit higher uptake than the group consisting of cyclohexene, p-xylene, naphthalene, and phenanthrene. When normalized for hydrophobic effects, there does not appear to be any significant role of π-electron energy or of polarizability within a single group. Therefore, the hydrophobic effect appears to be the dominant mechanism controlling uptake. The higher uptake of benzene and toluene cannot be explained by specific interactions. It is possible that these compounds can access micropores that the other compounds cannot; or perhaps they have a higher affinity for amorphous carbon present (e.g. an ability to partition into such phases). The chlorinated benzenes exhibit flatter isotherms, which do not fit into either of the groups apparent in figure 7; their data tends to overlap both groups.

Zoom In Zoom Out Reset image size

For comparison, the data for uptake of similar compounds on activated carbon are plotted in figure 8 (Speth and Miltner 1990, Walters and Luthy 1984). With the exception of phenanthrene, the adsorbates are rather tightly grouped after normalizing for hydrophobic effects, although benzene exhibits somewhat lower uptake in the low concentration range. The lower uptake of benzene is in contrast to the findings for the SWNTs. The observation that compounds with either electron withdrawing or electron donating groups appear to exhibit somewhat higher uptake than benzene may indicate enhanced sorption caused by specific interactions, but the effect is not dramatic. The significantly lower uptake of phenanthrene by activated carbon is likely a result of size exclusion effects. Taken as a whole, however, it appears that hydrophobic interactions govern the uptake of the chemicals studied here by both activated carbons and CNTs.

Zoom In Zoom Out Reset image size

It is possible that uptake by SWNTs also involves size exclusion effects, which mask possible specific interactions with naphthalene and phenanthrene; i.e. it is possible that, in general, the observed coincidence of the isotherms in both figures 7 and 8 is a result of enhanced sorption due to the effect of functional groups on the electron density of the ring, partially canceled by size exclusion effects. This is difficult to assess in any further detail, and is not considered likely. However, we can conclude that at a minimum, size exclusion effects appear to be less significant for the SWNTs than for the GAC.