Science of the Total Environment 497–498 (2014) 133–138
Contents lists available at ScienceDirect
Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv
Sorption behavior of carbon nanotubes: Changes induced by functionalization, sonication and natural organic matter Melanie Kah a,⁎,1, Xiaoran Zhang a,b, Thilo Hofmann a,⁎,1 a
Department of Environmental Geosciences, University of Vienna, Althanstrasse 14, Vienna 1090, Austria School of Environment and Energy Engineering, Key Laboratory of Urban Stormwater System and Water Environment, Ministry of Education, Beijing University of Civil Engineering and Architecture, Zhanlanguan Road 1, Xicheng District, Beijing, 100044, China
b
H I G H L I G H T S
G R A P H I C A L
A B S T R A C T
• Low levels of functionalization significantly affected sorption affinity for pyrene. • CNTs suspensions remain difficult to characterize. • Sonication significantly increased the surfaces accessible for pyrene sorption. • Sorption affinity was very high across all conditions (7.5 b log Kd b 9). • Passive sampling allowed sorption to be studied over 5 orders of magnitude.
a r t i c l e
i n f o
Article history: Received 9 July 2014 Received in revised form 28 July 2014 Accepted 28 July 2014 Available online xxxx Editor: D. Barcelo Keywords: Dispersion Adsorption Passive sampling Humic acid PAH Pyrene
a b s t r a c t The effect of functionalization on the sorption behavior of carbon nanotubes (CNTs) remains poorly understood, especially when combined with other factors affecting dispersion. The sorption behavior of a series of functionalized CNTs towards pyrene has therefore been systematically evaluated over a wide range of concentrations and dispersion scenarios. When studied as purchased (in the absence of humic acids and sonication treatment), sorption isotherms showed that OH-, COOH- and NH2-CNTs exhibited significantly different sorption affinity for pyrene. Sonication greatly increased both the sorption affinity and the maximum capacity of all types of functionalized CNTs, to an extent that overwhelmed the differences initially observed (increase of up to 1.5 orders of magnitude lead to log Kd values close to 9 L/kg). Results demonstrate that a significant proportion of the CNT surface was unavailable to pyrene prior to sonication. The presence of humic acids enhanced dispersion but decreased sorption, especially when combined with sonication. Sorption affinity, however, remained very high in all cases (log Kd N 7.5 L/kg), suggesting that CNTs can act as strong sorbents under a wide range of conditions. © 2014 Elsevier B.V. All rights reserved.
⁎ Corresponding authors. E-mail addresses: [email protected] (M. Kah), [email protected] (T. Hofmann). 1 Tel.: +43 1 427753320.
http://dx.doi.org/10.1016/j.scitotenv.2014.07.112 0048-9697/© 2014 Elsevier B.V. All rights reserved.
1. Introduction Carbon nanotubes (CNTs) have potential applications in many areas including material science, analytical chemistry and environmental
134
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
science (Dervishi et al., 2009). CNTs exhibit a very strong affinity for organic contaminants and have been proposed as a superior sorbent for waste water treatment (Mauter and Elimelech, 2008) and for use in solid phase extraction cartridges (Ma et al., 2010a). Understanding the interactions between organic contaminants and CNTs is therefore essential for evaluating the potential environmental impact of CNTs (through, for example, facilitated nanoparticle-bound cotransport of organic contaminants), as well as their potential efficiency as a superior sorbent. CNTs tend to form large aggregates in water due to van der Waals forces between the tubes (Dervishi et al., 2009). Many industrial applications, however, require CNTs to be dispersed, which is typically achieved through mechanical treatment and functionalization (Ma et al., 2010b). In addition to intentional modifications, CNT surfaces can also be altered when exposed to oxidative conditions (Petersen et al., 2011), which is likely to occur during purification treatments and/or after their release into an aquatic environment (Hüffer et al., 2013; Smith et al., 2009). Taking into account the effect that surface functionalization of CNTs has on their sorption behavior is thus essential when evaluating their fate. Investigations into the effects that surface functionalization has on the sorption behavior of CNTs are, however, relatively scarce. While the effects of oxygen-containing functional groups obtained by acid treatment have been studied in prior research, the sorption behavior of commercially available functionalized CNTs (e.g., aminoated by plasma treatment) has yet to be evaluated. In addition, investigations into the influence of surface oxidation on sorption have led to mixed interpretation (Pan and Xing, 2010). Consistently with most literature on the interactions between hydrophobic compounds and carbonaceous sorbents, authors who observed a decrease in sorption affinity with increasing level of CNTs oxidation (Cho et al., 2008; Zhang et al., 2009) proposed that water clusters around functional vgroups can prevent electron donor– acceptor interactions from taking place. In contrast, the increase in sorption affinity with increasing levels of CNT oxidation observed for toluene, ethylbenzene and xylene (Lu et al., 2008; Yu et al., 2011) has been explained as being due to an enhancement of π–π and n–π electron donor–acceptor interactions. It is important to distinguish the effect that functionalization has on the maximum sorption capacity (the space available for sorption) from its effect on the sorption affinity (the strength of attractive forces between sorbent and sorbates). This has proved difficult when sorption isotherms are only determined over a relatively narrow range of concentrations and fitted with the Freundlich model (an experimental model that assumes multiple types of site acting in parallel and an unlimited total number of sorption sites). Based on fits with a Polanyi theory-based model, Wu et al. (2012) recently proposed that surface oxidation induced competition with water molecules, which consequently decreased the maximum sorption capacity of CNTs, but had only a minimal effect on sorption affinity. The effect that surface functionalization has on sorption affinity deserved further research as it is key to assess sorption behavior for low concentrations of sorbates, which are particularly relevant when considering the release or application of CNTs in the environment. The effect that functionalization has on the CNT dispersion status also needs to be taken into account, as functionalization can increase the maximum sorption capacity by increasing the exposed surface area (Yu et al., 2011). Most studies to date have only referred to aggregated CNTs studied through batch sorption set-ups that may not allow full separation of the nano-scaled sorbent from the aqueous phase. We have previously examined the sorption behavior of both partially and fully dispersed CNTs using a passive sampling set-up (Kah et al., 2011; Zhang et al., 2012). Mechanical treatment (i.e., sonication) and the presence of natural dispersants (i.e., humic acids, HA) increased the stability of the CNT suspension, with a large impact on the sorption behavior (Zhang et al., 2012). The full dispersion of a CNT suspension requires a combination of functionalization and sonication, either with or without the addition
of dispersants (Hou et al., 2013; Schwyzer et al., 2011; Zhou et al., 2012). Investigating the effects of functionalization, both on its own and in combination with other factors affecting dispersion, will thus allow a better understanding of CNT sorption behavior under conditions likely to occur during their production and following their release into the environment. The aim of our research was therefore to fill in some of these gaps in our knowledge and to investigate the influence that surface functionalization has on the dispersion and sorption behavior of CNTs. The specific objectives were (i) to distinguish the effects that different types of functionalization (i.e., OH-, COOH- and NH2-functionalized CNTs) have on CNT sorption behavior and to investigate how these effects vary with further dispersion by (ii) sonication and (iii) the use of natural dispersants. Sorption isotherms were measured over a wide range of sorbates concentrations in order to distinguish the effects on sorption affinity and maximum capacity. CNT suspensions were extensively characterized by microscopic, spectroscopic, size and size distribution measurements in order to assist in mechanistic interpretations of the results. 2. Materials and methods 2.1. Sorbents and chemicals A series of functionalized multiwalled CNTs was selected to investigate the effect of different functional groups and different methods of functionalization. Hydroxylated (OH-), carboxylated (COOH-) and aminoated (NH2-) CNTs were functionalized by plasma treatment and contained 7 ± 1.5 wt% functional groups according to X-ray photoelectron spectroscopy (XPS) and titration results provided by the supplier (Cheaptubes, Brattleboro, USA). All CNTs were synthesized by chemical vapor deposition; they had purities N99% and mean outer diameters ranging from 13 to 18 nm (Table S1 in the Supporting Information, SI). Humic acid standard II “2S101H” (HA) was selected as a model natural dispersant (Suwannee River, from the International Humic Substance Society). HA solutions of between 5 and 40 mg/L were prepared in 25 mg/L NaN3 background (as biocide). First, 300 mg of HA were dissolved in 5 mL of 0.1 M NaOH. The solution was then diluted with background solution to obtain a 1 g/L HA stock solution. The pH was adjusted from acidic to neutral (pH 7.06) using 0.1 M NaOH. The HA stock solution was then gradually diluted. A polyoxymethylene sheet (POM: thickness 0.5 mm, density 1.41 g/cm 3 ) purchased from Vink Kunststoffen BV (Didam, The Netherlands) was cut into strips (1 cm × 1 cm) and cold-extracted with hexane (30 min) and methanol (3 times for 30 min), as previously described by Jonker and Koelmans (2001). Pyrene (99.0%) and pyrene-d10 (99.5%) were purchased from Dr. Ehrenstorfer (Germany). All stock solutions were prepared in methanol. The hexane and methanol were of residue analysis grade (from Lab Scan, Dublin, Ireland, and Acros Organics, Geel, Belgium). 2.2. Sorption experiments All experiments were carried out at 20 ± 1 °C using a MilliQ water background solution prepared with 25 mg/L NaN3 as biocide (pH 7.06). One milligram of CNTs and 50 mL of HA solution (0, 5 and 40 mg/L) were added to glass vials. Samples were then pretreated by (i) shaking by hand for 1 minute or (ii) sonication for 2 h. For the sonication, 20 sample vials were immersed into the sonication bath (Bandelin sonorex super rk106 bath, 35 kHz, diameter 24.5 cm). A POM strip (approximately 100 mg) was subsequently added to each sample and pyrene was spiked. The concentration of methanol (spiking solvent) was kept at 0.16% for all samples in order to minimize solvent effects. For each of the two pretreatments, sorption coefficients were measured for an initial pyrene concentration of 50 μg/L. In addition, full sorption isotherms for pyrene were measured at 0 mg/L HA (pyrene equilibrium
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
concentration ranged over five orders of magnitude from 0.0001 to 20 μg/L). All vials were shaken horizontally (180 rpm) for 28 days to ensure equilibration. The POM strips were then removed from the vials, rinsed with deionized water and wiped with a wet tissue. They were then extracted with methanol by accelerated solvent extraction (ASE 200, Dionex, USA; 1500 psi, 100 °C) using pyrene-d10 as an internal standard. Extracts were concentrated under N2 prior to gas chromatography–mass spectrometry analysis. Details about the method validation (e.g., equilibration time, extraction efficiency, recovery and reproducibility) can be found in our previous study (Kah et al., 2011). Blanks containing no pyrene were prepared for each pretreatment and used for characterization (see the following section). Control samples containing pyrene indicated that losses were b 6% after 28 days of shaking. The sorption of pyrene to CNTs was calculated based on mass balance, as described in our previous study (Zhang et al., 2012). Our previous experience with the method applied to similar materials indicated a typical maximum 10% error for triplicates (Hüffer et al., 2013; Kah et al., 2011). Instead of repeating samples, priority was here given to generate data for more concentration levels along the isotherm in order to support more robust isotherm fits. 2.3. Characterization of CNTs The CNTs were imaged by scanning electron microscopy (SEM) after the sorption experiments, for each type of CNT and each pretreatment (more details available in the SI). The absorbance of CNT suspensions was measured by UV–vis spectrometry at 800 nm (Varian Cary 50 UV–vis spectrophotometer) as an indication for suspension stability. With the aim to further compare the stability of the suspensions, the average hydrodynamic diameter of CNT aggregates remaining in suspension after 2 days of settling was measured by dynamic light scattering (DLS, Malvern ZetaSizer Nano). Size distributions for the whole suspensions were determined using a particle size analyzer (Eyetech, based on the time of transition principle) and a Malvern Mastersizer (MasterSizer 2000, based on a laser light scattering principle). The CNT surface functional groups were detected with Fourier transform infrared spectroscopy (FTIR, Bruker, Tensor 27) and X-ray photoelectron spectroscopy (XPS, Kratos Axis Ultra electron spectrometer). The bulk chemistry of the CNTs was further characterized by energy dispersive X-ray spectroscopy (EDX) and isotope-ratio mass spectrometer (IRMS: further details available in the SI). The specific surface area and pore volume were measured by N2 sorption (NOVA2000, Quantachrome). 2.4. Sorption models and statistics Six sorption models that have previously been applied to describe sorption by carbonaceous materials were fitted to the isotherms: these being the Freundlich, Langmuir, dual Langmuir, Toth, dual-mode and Dubinin–Ashtakhov models. Descriptions of the models and their parameters are available in Table S2. The goodness of fit was evaluated and compared on the basis of the r2 values, mean weighted square errors and Akaike's information criterion, as described previously by Kah et al.
135
(2011). All statistical tests and parameter optimizations were performed with SigmaPlot 11.0 for Windows. 3. Results and discussion Investigations into the sorption behavior of CNTs are often based on the Freundlich model. However, Fig. S1 and S2 clearly show that this model is not suitable for fitting pyrene sorption isotherms over a wide range of concentrations. Of the six sorption models tested, and based on a multicriteria comparison of the goodness of fit (Kah et al., 2011), the Dubinin–Ashtakhov and Toth models gave the best isotherm fits for all types of CNTs (Table S3 and Fig. S1–S2). The parameters derived by fitting (Table 1) are used below to discuss the discrepancies observed in terms of sorption affinity (E and Kt), maximum capacity (Q0) and heterogeneity (t). 3.1. Effects of functional groups (without sonication) The effect that CNT oxidation has on sorption affinity for nonpolar compounds remains unclear as previous work has reported an increase (Lu et al., 2008; Yu et al., 2011), a decrease (Cho et al., 2008) and no effect (Wu et al., 2012). Fig. 1 suggests that after shaking only (i.e., with no sonication or addition of HA, open symbols), the sorption behavior varies with the type of CNT functionalization. In order to better understand these variations, each type of CNTs has been extensively characterized. Analyzing the level of oxidation of CNTs is known to be a difficult task (Wepasnick et al., 2010), with each technique having its own strengths and limitations. We therefore applied several different approaches (i.e., EDX and IRMS as bulk analysis, and XPS as surface analysis) in combination with the data provided by the supplier (XPS and titration; Tables S1 and S4). Larger levels of oxidation were generally measured for COOH-CNTs. However, neither absolute values nor trends in oxygen content were consistent across methods for the other types of CNTs. The relatively large O% measured for NH2-CNTs by all methods was unexpected. XPS analysis also indicates that NH2-CNTs contain much less NH2 groups than suggested by the supplier (about 0.25% against 7%). Variability from batch to batch is likely to occur when producing CNTs. Our results highlight the importance of characterizing purchased materials using several complementary techniques, which can quickly become a very time- and budget-consuming task. In literature, the concentration of specific functional groups at the surface of CNTs is often solely derived from XPS data. However, because of ambiguities in spectral interpretation (Wepasnick et al., 2010) and the fact that the standard deviation is often not reported, there is a risk to over interpret differences between materials. In addition to spectroscopic approaches, a number of indirect indications for surface functionalization can also be explored. For instance, it is worth noting that prior to sonication, the zeta potential of all types of CNT aggregates was close to neutral or only slightly negative (Table S5). The pH in this study being close to neutral, the zeta potentials of OHand COOH-CNTs were expected to be negative and that of NH2-CNTs was expected to be positive (Schaffer and Licha, 2014). The lack of a
Table 1 Fitting parameters ± standard error and r2 for Dubinin–Ashtakhov and Toth models fitting the isotherms of pyrene on three types of CNTs pretreated by shaking only or by sonication. Dubinin–Ashtakhov model
Toth model
Shaking only
log Q0
E (kJ/mol)
b
r2
log Q0
Kt (L/μg)
t
r2
OH-CNTs COOH-CNTs NH2-CNTs
7.54 ± 0.08 7.58 ± 0.06 7.57 ± 0.05
18.38 ± 0.93 19.48 ± 0.63 22.02 ± 0.47
2.24 ± 0.25 2.27 ± 0.16 3.01 ± 0.18
0.94 0.96 0.99
7.55 ± 0.12 7.59 ± 0.06 7.56 ± 0.10
1.67 ± 0.35 2.21 ± 0.29 4.40 ± 2.19
0.58 ± 0.13 0.54 ± 0.07 0.78 ± 0.21
0.95 0.97 0.99
Sonication
log Q0
E (kJ/mol)
b
r2
log Q0
Kt (L/μg)
t
r2
OH-CNTs COOH-CNTs NH2-CNTs
7.73 ± 0.10 7.72 ± 0.03 7.70 ± 0.04
23.60 ± 1.31 23.42 ± 0.45 23.48 ± 0.56
2.28 ± 0.29 2.18 ± 0.09 2.32 ± 0.13
0.99 0.99 0.99
7.88 ± 0.21 7.95 ± 0.08 7.84 ± 0.09
3.49 ± 1.04 3.16 ± 0.30 3.48 ± 0.47
0.30 ± 0.07 0.26 ± 0.02 0.31 ± 0.03
0.99 0.99 0.99
136
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
Adsorbed Concentrations,C CNT (µg/kg)
1e+8
1e+7
increases (Smith et al., 2009; Wu et al., 2012), no effect (Cho et al., 2008) and decreases (Gotovac et al., 2007; Lu et al., 2008; Zhang et al., 2009) in the surface area after CNT functionalization have all been reported by different authors. In view of the uncertainty relating to BETnitrogen surface area measurements and the current lack of a wellaccepted alternative (especially for measurements in suspension), great care is required when attempting to interpret normalization by the specific surface area. Finally, no significant difference in surface heterogeneity between the different types of functionalization (t, p N 0.05, Table 1) was observed, probably due to insufficient levels of functionalization. Overall, the type of CNT functionalization had a significant effect on the sorption affinity for pyrene (after shaking only), but no effect on the maximum sorption capacity or the surface heterogeneity. Elucidating the driving mechanism(s) will remain difficult until more reliable synthesis and characterization techniques for suspensions are developed, permitting more accurate descriptions of the nature, density and distribution of functional groups on, and within, the CNT aggregates.
Sonication OH-CNTs COOH-CNTs NH2-CNTs
1e+6
Shaking OH-CNTs COOH-CNTs NH2-CNTs
1e+5
Dubinin-Ashtakhov model Toth model 1e+4 1e-5
1e-4
1e-3
1e-2
1e-1
1e+0
1e+1
1e+2
Equilibrium Concentrations, Cw (µg/L) Fig. 1. Sorption isotherms of pyrene to three types of CNTs pretreated by shaking and by sonication. The lines represent the fits by the Dubinin–Ashtakhov (DAM) and Toth (TM) models.
clear surface charge indicates that the number of functional groups located at the surface of the aggregates was low and/or that a mixture of cationic and anionic functional groups co-existed. Negative zeta potentials for NH2 functionalized CNTs have previously been reported by other authors across a wide pH range (Liao et al., 2008; Wang et al., 2012, 2013). The application of three spectroscopic methods here confirmed that negative charges are much likely due to the presence of dissociated oxygen-containing functional groups on NH2-CNTs. Sorption affinity increased in the following order: OH-CNTs b COOHCNTs b NH2-CNTs (E and Kt, p b 0.001; Fig. 1 and Table 1). Overall, there appeared to be no clear relationship between sorption affinity and functionalization, suggesting that additional factors need to be taken into account in order to explain the variations in sorption affinity. The strongest affinity of pyrene, which was for NH2-CNTs, could be explained by the overall higher density of electron-donating groups (N- and Ocontaining groups) but are unlikely to be mainly driven by NH2 groups (which would have been concluded if relying solely on the data from the supplier). The maximum sorption capacity of functionalized CNTs for hydrophobic compounds was expected to be mainly influenced by (i) water clusters surrounding functional groups (negative effect, Cho et al., 2008; Wu et al., 2012; Zhang et al., 2009) and (ii) an increase in the total accessible surface area through enhanced dispersion (positive effect, Yu et al., 2011). However, we observed similar maximum sorption capacity for all types of CNTs (p N 0.05, Table 1). None of the CNT suspensions achieved a good dispersion status after 6 days shaking only (see pictures on Fig. S3). SEM observations (Fig. S4a) suggest very similar aggregate sizes and structures. After 28 days shaking, the proportion of CNTs remaining in suspension increased slightly in the following order: OH-CNTs b COOH-CNTs b NH2-CNTs (based on UV absorbance measurements, Fig. S5a) but remained very low overall. The specific surface area after shaking decreased in the following order: OH-CNTs N COOH-CNTs N NH2-CNTs (Table S6). The normalization of the maximum sorption capacity by the specific surface area of CNTs can be useful for ruling out the influence of variations in surface area with the type of CNT functionalization. We found that the surface-normalized sorption capacity followed the order OH-CNTs b COOH-CNTs b NH2CNTs, consistent with the sorption affinity. It is important to note that
3.2. Effect of sonication Sonication efficiently dispersed the three types of functionalized CNTs, as indicated by higher UV absorbance values and smaller size distributions (Figs. S5c, S6 and S7). A slightly better dispersion of NH2-CNTs was observed immediately after sonication, as indicated by smaller aggregate size, higher absorbance and more negative zeta potential (Figs. S5, S7 and Table S5). However, the stability of all CNT suspensions was poor and reaggregation occurred rapidly (Figs. S5d and S7). Sonication significantly increased both the maximum sorption capacity and the sorption affinity of all types of CNTs for pyrene (p b 0.05, Table 1 and Fig. 1). These increases could be well explained by an enlargement of the pore spaces that occur between the aggregates or between individual CNTs, after the dispersion–reaggregation event. It has been previously reported that sorbates have similar access to surfaces located within CNT aggregates and those located at their surfaces (Cho et al., 2008). It is, however, important to stress that this conclusion was made after the application of a sonication treatment, and for a relatively small sorbates (e.g., naphthalene as in Cho et al., 2008). The increase of both the maximum sorption capacity and the sorption affinity observed here for pyrene suggests that a significant proportion of the CNT aggregate surface was unavailable prior to sonication. The effect of sonication on the sorption behavior of CNTs with different functionalization has not previously been reported. We saw in the section above that the sorption affinity for pyrene significantly increased in the order OH-CNTs b COOH-CNTs b NH2-CNTs, but no more differences in the sorption isotherms could be distinguished following the sonication treatment (p N 0.05, Fig. 1). It is also worth mentioning that isotherms are identical to those we previously obtained after sonication of non-functionalized CNTs, with similar physical characteristics, but obtained from an another provider (see Table S3 and Zhang et al., 2012). This similarity in behavior after sonication can be explained by the strong and irreversible effects that sonication has on both the surface chemistry and aggregate morphology of CNTs. A number of indicators suggest that sonication had a significant effect on the surface chemistry of all types of CNTs: the zeta potential became more negative (from about − 3 to − 25 mV, Table S5), the oxygen content increased (according to EDX and IRMS techniques, Table S4) and the FTIR spectra suggested the generation of OH and C = O groups (Fig. S8). The increase in surface heterogeneity for all types of CNTs (decrease in t value, Table 1) could also be explained by the generation of oxygen-containing functional groups, as well as defects such as bending, buckling and breaking (Vichchulada et al., 2010). The degree of initial functionalization (possibly complemented by oxidation during sonication) remained, however, relatively low. We believe that it is very likely that the pore enlargement—possibly maintained after reaggregation owing to the defects generated during
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
sonication—was mainly responsible for the observed large increase in sorption. Hou et al. (2013) recently proposed that looser aggregates allow sorbate molecules to simultaneously interact with several CNTs. Such a phenomenon could also explain the enhanced sorption affinity that we observed following sonication. The effect of pore enlargement on sorption, however, is difficult to characterize as it involves determining the effective surface area available for the sorption of a given sorbate. On one hand, the nitrogen adsorption method determines the surface area accessible to N2 (i.e., a relatively small sorbate) and is subject to uncertainties due to unavoidable artifacts introduced during sample preparation (as shown in Table S6), while on the other hand, values derived from the size distribution of the aggregates only reflect the external surface areas and certainly underestimate the total surface area available for sorption. This is particularly true for small sorbates that can access a higher proportion of pores than larger sorbates. It has previously been suggested that surface chemistry has a greater effect than dispersion status on the sorption behavior of CNTs (Zhang et al., 2009). However, no measurements on dispersed CNT systems were yet available at that time. Our results highlight the importance of CNT aggregate morphology, which can be greatly affected by a dispersion event. After sonication, the physical characteristics of the CNTs (e.g., their diameters and the number of similar nanotube walls in the CNTs investigated) became the main drivers of their sorption behavior. Sonication is frequently applied when studying CNTs, and the impact that this has on aggregate structures and consequent sorption behavior should therefore not be neglected (changes of up to 1.5 orders of magnitude have been observed, based on the single Kd values presented in Fig. 2). Different observations could be expected for highly functionalized CNTs that disperse spontaneously in water and expose all of their surface areas to potential sorbates. Functionalization by plasma or acid treatments has, to date, resulted in a maximum of 15% wt. functionalization, which is insufficient for spontaneous dispersion in water (Sahoo et al., 2010). 3.3. Effects of HA The presence of HA efficiently dispersed all types of CNTs, but only when combined with a sonication treatment (Fig. S3). Absorbance and size measurements indicate that the stability of OH-, COOH- and NH2CNT suspensions were relatively similar (Figs. S5c, S6 and S7a). The reaggregation of all types of CNTs occurred during shaking after sonication but to a lesser extent than in the absence of HA (Figs. S3, S5d, S6 and S7b), probably due to the electrosteric stabilization provided by HA molecules adsorbed onto CNTs. The addition of HA certainly increased the negative charge of CNT aggregates significantly (more negative zeta potential in Table S5). After shaking only, there was a clear increase
log KCNT (L/Kg)
10
9
8
7
6 OH-
COOH-
NH2-
Fig. 2. Single distribution coefficients of pyrene to three types of CNTs (log KCNT, L/kg), as affected by HA concentrations (□, , are 0, 5, 40 mg/L HA, respectively) and two pretreatments (□ shaking and sonication).
137
in the negative surface charge following the order OH-CNTs b COOHCNTs b NH2-CNTs. Following sonication in the presence of HA, the surface charges of all types of CNTs were similar (i.e., around − 50 mV and − 60 mV at 5 and 40 mg/L of HA, respectively). This contrasts with the results obtained by Smith et al. (2012), who reported that the effective surface charge of oxidized CNTs decreased as the concentration of natural organic matter increased. Changes in surface charge can be expected to depend on the nature of the organic matter. That investigated by Smith et al. (2012) was from Great Dismal Swamp (Virginia) and presumably possessed a lower density of negatively charged groups than the Suwanee River HA used in our investigations. Increasing the concentration of HA greatly suppressed the sorption affinity for pyrene of OH-, COOH- and NH2-CNTs (Fig. 2, p b 0.001), especially after sonication. This contrasts with results we previously obtained for non-functionalized CNTs (Zhang et al., 2012), which showed that sorption of pyrene was relatively insensitive to the presence of HA (a decrease in sorption was only observed after sonication). Discrepancies in behavior between functionalized and non-functionalized CNTs could be, at least partially, explained by differences in their interactions with HA. For instance, sonication was previously reported to increase the amount of peat HA sorbed to CNTs by a factor of approximately two (Zhou et al., 2012). Data on the sorption of natural organic matter to functionalized CNTs remain limited. While a similar affinity of peat HA for pristine and COOH-functionalized CNTs has previously been reported (Zhou et al., 2012), a negative impact of CNT oxidation on the sorption of NOM has also been observed (Smith et al., 2012; Wang et al., 2013). Integrating our results for suspension stability (Fig. S7), size of aggregates (Figs. S6 and S7) and suppression of pyrene sorption (Fig. 2) suggests an affinity of HA for CNTs that increases in the following order: NH2-CNTs b OH-CNTs ≈ COOH-CNTs. Further research will be required to elucidate the mechanisms involved in the interactions between CNTs and natural organic matter, and to understand how these vary with the type of CNT, with the type of natural organic matter and over a range of water chemistries. Overall, increasing the HA concentration generally decreased the sorption affinity for pyrene, much likely due to a combination of competition and pore blockage. It is important to note, however, that even at a relatively high HA concentration (40 mg/L), the sorption affinity for pyrene remained very strong, about three orders of magnitude larger than that of pyrene to HA (i.e., log KCNT N 7.4 compared to log KHA 4.7; Zhang et al., 2012). 4. Conclusions One of our initial objectives was to generate robust sorption data for NH2-CNTs, for which no sorption data for hydrophobic compounds had previously been published. However, comprehensive surface characterization indicated that purchased NH2-CNTs contained only a small amount of amino group and a relatively large amount of oxygen containing groups, highlighting the necessity to extensively characterize materials using complementary techniques. When studied as purchased (in the absence of HA and sonication treatment), sorption isotherms over a wide range of pyrene concentrations showed that OH-, COOH- and NH2-CNTs exhibited significantly different sorption affinity. Despite extensive characterization, discrepancies due to surface chemistry and dispersion were difficult to distinguish. Elucidating the mechanisms involved will require further work with materials that are functionalized more homogeneously, and to a higher level. Commercially available CNTs are nevertheless essential to study, in view of the large quantities produced and consequent high environmental relevance (relative to purer materials produced in the laboratory in much smaller quantities). Sonication greatly increased both the sorption affinity and maximum capacity of all types of functionalized CNTs, to an extent that overwhelmed the differences initially observed. Results demonstrate that a significant proportion of the CNT surface was unavailable to
138
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
pyrene prior to sonication. Sonication treatment is frequently applied when studying CNTs, and its irreversible effect on sorption behavior should not be neglected (increase of up to 1.5 orders of magnitude was observed, leading to a log Kd value close to 9). Even though the presence of HA reduced the sorption affinity, the exceptional sorption properties of CNTs were still maintained, at least in part, which suggests that CNTs can act as strong sorbents under a wide range of conditions. Acknowledgments The authors would like thank Petra Körner for her invaluable support in the laboratory. We are grateful to Gerlinde Habler and Frank von der Kammer for electron microscope imaging, to Eugen Libowitzky for FTIR measurements (University of Vienna, AT) and to Andrey Shchukarev for XPS analysis (Umeå University, SE). Xiaoran Zhang was financially supported by the China Scholarship Council. Appendix A. Supplementary data Details of CNT characterization, isotherm fits and parameters for the six models evaluated. Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.scitotenv.2014.07.112 References Cho HH, Smith BA, Wnuk JD, Fairbrother DH, Ball WP. Influence of surface oxides on the adsorption of naphthalene onto multiwalled carbon nanotubes. Environ Sci Technol 2008;42:2899–905. Dervishi E, Li ZR, Xu Y, Saini V, Biris AR, Lupu D, et al. Carbon nanotubes: synthesis, properties, and applications. Part Sci Technol 2009;27:107–25. Gotovac S, Yang CM, Hattori Y, Takahashi K, Kanoh H, Kaneko K. Adsorption of polyaromatic hydrocarbons on single wall carbon nanotubes of different functionalities and diameters. J Colloid Interface Sci 2007;314:18–24. Hou L, Zhu D, Wang X, Wang L, Zhang C, Chen W. Adsorption of phenanthrene, 2-naphthol, and 1-naphthylamine to colloidal oxidized multiwalled carbon nanotubes: Effects of humic acid and surfactant modification. Environ Toxicol Chem 2013;32:493–500. Hüffer T, Kah M, Hofmann T, Schmidt TC. How redox conditions and irradiation affect sorption of PAHs by nC60. Environ Sci Technol 2013;47:6935–42. Jonker MTO, Koelmans AA. Polyoxymethylene solid phase extraction as a partitioning method for hydrophobic organic chemicals in sediment and soot. Environ Sci Technol 2001;35:3742–8. Kah M, Zhang XR, Jonker MTO, Hofmann T. Measuring and modelling adsorption of PAHs to carbon nanotubes over a six order of magnitude wide concentration range. Environ Sci Technol 2011;45:6011–7. Liao Q, Sun J, Gao L. Adsorption of chlorophenols by multi-walled carbon nanotubes treated with HNO3 and NH3. Carbon 2008;46:553–5. Lu C, Su F, Hu S. Surface modification of carbon nanotubes for enhancing BTEX adsorption from aqueous solutions. Appl Surf Sci 2008;254:7035–41.
Ma JP, Xiao RH, Li JH, Yu JB, Zhang YQ, Chen LX. Determination of 16 polycyclic aromatic hydrocarbons in environmental water samples by solid-phase extraction using multiwalled carbon nanotubes as adsorbent coupled with gas chromatography–mass spectrometry. J Chromatogr A 2010a;1217:5462–9. Ma PC, Siddiqui NA, Marom G, Kim JK. Dispersion and functionalization of carbon nanotubes for polymer-based nanocomposites: a review. Compos A: Appl Sci Manuf 2010b;41:1345–67. Mauter MS, Elimelech M. Environmental applications of carbon-based nanomaterials. Environ Sci Technol 2008;42:5843–59. Pan B, Xing BS. Manufactured nanoparticles and their sorption of organic chemicals. Adv Agron 2010;108(108):137–81. Petersen EJ, Zhang LW, Mattison NT, O'Carroll DM, Whelton AJ, Uddin N, et al. Potential release pathways, environmental fate, and ecological risks of carbon nanotubes. Environ Sci Technol 2011;45:9837–56. Sahoo NG, Rana S, Cho JW, Li L, Chan SH. Polymer nanocomposites based on functionalized carbon nanotubes. Prog Polym Sci 2010;35:837–67. Schaffer M, Licha T. A guideline for the identification of environmentally relevant, ionizable organic molecule species. Chemosphere 2014;103:12–25. Schwyzer I, Kaegi R, Sigg L, Magrez A, Nowack B. Influence of the initial state of carbon nanotubes on their colloidal stability under natural conditions. Environ Pollut 2011; 159:1641–8. Smith B, Wepasnick K, Schrote KE, Cho HH, Ball WP, Fairbrother DH. Influence of surface oxides on the colloidal stability of multi-walled carbon nanotubes: a structure–property relationship. Langmuir 2009;25:9767–76. Smith B, Yang J, Bitter JL, Ball WP, Fairbrother DH. Influence of surface oxygen on the interactions of carbon nanotubes with natural organic matter. Environ Sci Tech 2012; 46:12839–47. Vichchulada P, Cauble MA, Abdi EA, Obi EI, Zhang QH, Lay MD. Sonication power for length control of single-walled carbon nanotubes in aqueous suspensions used for 2-dimensional network Formation. J Phys Chem C 2010;114:12490–5. Wang F, Zhu DQ, Chen W. Effect of copper ion on adsorption of chlorinated phenols and 1-naphthylamine to surface-modified carbon nanotubes. Environ Toxicol Chem 2012;31:100–7. Wang F, Yao J, Chen H, Yi Z, Xing B. Sorption of humic acid to functionalized multi-walled carbon nanotubes. Environ Pollut 2013;180:1–6. Wepasnick KA, Smith BA, Bitter JL, Fairbrother DH. Chemical and structural characterization of carbon nanotube surfaces. Anal Bioanal Chem 2010;396:1003–14. Wu WH, Chen W, Lin DH, Yang K. Influence of surface oxidation of multiwalled carbon nanotubes on the adsorption affinity and capacity of polar and nonpolar organic compounds in aqueous phase. Environ Sci Technol 2012;46:5446–54. Yu F, Ma J, Wu Y. Adsorption of toluene, ethylbenzene and m-xylene on multi-walled carbon nanotubes with different oxygen contents from aqueous solutions. J Hazard Mater 2011;192:1370–9. Zhang SJ, Shao T, Bekaroglu SSK, Karanfil T. The impacts of aggregation and surface chemistry of carbon nanotubes on the adsorption of synthetic organic compounds. Environ Sci Technol 2009;43:5719–25. Zhang XR, Kah M, Jonker MTO, Hofmann T. Dispersion state and humic acids concentrationdependent sorption of pyrene to carbon nanotubes. Environ Sci Technol 2012;46: 7166–73. Zhou XZ, Shu L, Zhao HB, Guo XY, Wang XL, Tao S, et al. Suspending multi-walled carbon nanotubes by humic acids from a peat soil. Environ Sci Technol 2012;46: 3891–7.
Contents lists available at ScienceDirect
Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv
Sorption behavior of carbon nanotubes: Changes induced by functionalization, sonication and natural organic matter Melanie Kah a,⁎,1, Xiaoran Zhang a,b, Thilo Hofmann a,⁎,1 a
Department of Environmental Geosciences, University of Vienna, Althanstrasse 14, Vienna 1090, Austria School of Environment and Energy Engineering, Key Laboratory of Urban Stormwater System and Water Environment, Ministry of Education, Beijing University of Civil Engineering and Architecture, Zhanlanguan Road 1, Xicheng District, Beijing, 100044, China
b
H I G H L I G H T S
G R A P H I C A L
A B S T R A C T
• Low levels of functionalization significantly affected sorption affinity for pyrene. • CNTs suspensions remain difficult to characterize. • Sonication significantly increased the surfaces accessible for pyrene sorption. • Sorption affinity was very high across all conditions (7.5 b log Kd b 9). • Passive sampling allowed sorption to be studied over 5 orders of magnitude.
a r t i c l e
i n f o
Article history: Received 9 July 2014 Received in revised form 28 July 2014 Accepted 28 July 2014 Available online xxxx Editor: D. Barcelo Keywords: Dispersion Adsorption Passive sampling Humic acid PAH Pyrene
a b s t r a c t The effect of functionalization on the sorption behavior of carbon nanotubes (CNTs) remains poorly understood, especially when combined with other factors affecting dispersion. The sorption behavior of a series of functionalized CNTs towards pyrene has therefore been systematically evaluated over a wide range of concentrations and dispersion scenarios. When studied as purchased (in the absence of humic acids and sonication treatment), sorption isotherms showed that OH-, COOH- and NH2-CNTs exhibited significantly different sorption affinity for pyrene. Sonication greatly increased both the sorption affinity and the maximum capacity of all types of functionalized CNTs, to an extent that overwhelmed the differences initially observed (increase of up to 1.5 orders of magnitude lead to log Kd values close to 9 L/kg). Results demonstrate that a significant proportion of the CNT surface was unavailable to pyrene prior to sonication. The presence of humic acids enhanced dispersion but decreased sorption, especially when combined with sonication. Sorption affinity, however, remained very high in all cases (log Kd N 7.5 L/kg), suggesting that CNTs can act as strong sorbents under a wide range of conditions. © 2014 Elsevier B.V. All rights reserved.
⁎ Corresponding authors. E-mail addresses: [email protected] (M. Kah), [email protected] (T. Hofmann). 1 Tel.: +43 1 427753320.
http://dx.doi.org/10.1016/j.scitotenv.2014.07.112 0048-9697/© 2014 Elsevier B.V. All rights reserved.
1. Introduction Carbon nanotubes (CNTs) have potential applications in many areas including material science, analytical chemistry and environmental
134
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
science (Dervishi et al., 2009). CNTs exhibit a very strong affinity for organic contaminants and have been proposed as a superior sorbent for waste water treatment (Mauter and Elimelech, 2008) and for use in solid phase extraction cartridges (Ma et al., 2010a). Understanding the interactions between organic contaminants and CNTs is therefore essential for evaluating the potential environmental impact of CNTs (through, for example, facilitated nanoparticle-bound cotransport of organic contaminants), as well as their potential efficiency as a superior sorbent. CNTs tend to form large aggregates in water due to van der Waals forces between the tubes (Dervishi et al., 2009). Many industrial applications, however, require CNTs to be dispersed, which is typically achieved through mechanical treatment and functionalization (Ma et al., 2010b). In addition to intentional modifications, CNT surfaces can also be altered when exposed to oxidative conditions (Petersen et al., 2011), which is likely to occur during purification treatments and/or after their release into an aquatic environment (Hüffer et al., 2013; Smith et al., 2009). Taking into account the effect that surface functionalization of CNTs has on their sorption behavior is thus essential when evaluating their fate. Investigations into the effects that surface functionalization has on the sorption behavior of CNTs are, however, relatively scarce. While the effects of oxygen-containing functional groups obtained by acid treatment have been studied in prior research, the sorption behavior of commercially available functionalized CNTs (e.g., aminoated by plasma treatment) has yet to be evaluated. In addition, investigations into the influence of surface oxidation on sorption have led to mixed interpretation (Pan and Xing, 2010). Consistently with most literature on the interactions between hydrophobic compounds and carbonaceous sorbents, authors who observed a decrease in sorption affinity with increasing level of CNTs oxidation (Cho et al., 2008; Zhang et al., 2009) proposed that water clusters around functional vgroups can prevent electron donor– acceptor interactions from taking place. In contrast, the increase in sorption affinity with increasing levels of CNT oxidation observed for toluene, ethylbenzene and xylene (Lu et al., 2008; Yu et al., 2011) has been explained as being due to an enhancement of π–π and n–π electron donor–acceptor interactions. It is important to distinguish the effect that functionalization has on the maximum sorption capacity (the space available for sorption) from its effect on the sorption affinity (the strength of attractive forces between sorbent and sorbates). This has proved difficult when sorption isotherms are only determined over a relatively narrow range of concentrations and fitted with the Freundlich model (an experimental model that assumes multiple types of site acting in parallel and an unlimited total number of sorption sites). Based on fits with a Polanyi theory-based model, Wu et al. (2012) recently proposed that surface oxidation induced competition with water molecules, which consequently decreased the maximum sorption capacity of CNTs, but had only a minimal effect on sorption affinity. The effect that surface functionalization has on sorption affinity deserved further research as it is key to assess sorption behavior for low concentrations of sorbates, which are particularly relevant when considering the release or application of CNTs in the environment. The effect that functionalization has on the CNT dispersion status also needs to be taken into account, as functionalization can increase the maximum sorption capacity by increasing the exposed surface area (Yu et al., 2011). Most studies to date have only referred to aggregated CNTs studied through batch sorption set-ups that may not allow full separation of the nano-scaled sorbent from the aqueous phase. We have previously examined the sorption behavior of both partially and fully dispersed CNTs using a passive sampling set-up (Kah et al., 2011; Zhang et al., 2012). Mechanical treatment (i.e., sonication) and the presence of natural dispersants (i.e., humic acids, HA) increased the stability of the CNT suspension, with a large impact on the sorption behavior (Zhang et al., 2012). The full dispersion of a CNT suspension requires a combination of functionalization and sonication, either with or without the addition
of dispersants (Hou et al., 2013; Schwyzer et al., 2011; Zhou et al., 2012). Investigating the effects of functionalization, both on its own and in combination with other factors affecting dispersion, will thus allow a better understanding of CNT sorption behavior under conditions likely to occur during their production and following their release into the environment. The aim of our research was therefore to fill in some of these gaps in our knowledge and to investigate the influence that surface functionalization has on the dispersion and sorption behavior of CNTs. The specific objectives were (i) to distinguish the effects that different types of functionalization (i.e., OH-, COOH- and NH2-functionalized CNTs) have on CNT sorption behavior and to investigate how these effects vary with further dispersion by (ii) sonication and (iii) the use of natural dispersants. Sorption isotherms were measured over a wide range of sorbates concentrations in order to distinguish the effects on sorption affinity and maximum capacity. CNT suspensions were extensively characterized by microscopic, spectroscopic, size and size distribution measurements in order to assist in mechanistic interpretations of the results. 2. Materials and methods 2.1. Sorbents and chemicals A series of functionalized multiwalled CNTs was selected to investigate the effect of different functional groups and different methods of functionalization. Hydroxylated (OH-), carboxylated (COOH-) and aminoated (NH2-) CNTs were functionalized by plasma treatment and contained 7 ± 1.5 wt% functional groups according to X-ray photoelectron spectroscopy (XPS) and titration results provided by the supplier (Cheaptubes, Brattleboro, USA). All CNTs were synthesized by chemical vapor deposition; they had purities N99% and mean outer diameters ranging from 13 to 18 nm (Table S1 in the Supporting Information, SI). Humic acid standard II “2S101H” (HA) was selected as a model natural dispersant (Suwannee River, from the International Humic Substance Society). HA solutions of between 5 and 40 mg/L were prepared in 25 mg/L NaN3 background (as biocide). First, 300 mg of HA were dissolved in 5 mL of 0.1 M NaOH. The solution was then diluted with background solution to obtain a 1 g/L HA stock solution. The pH was adjusted from acidic to neutral (pH 7.06) using 0.1 M NaOH. The HA stock solution was then gradually diluted. A polyoxymethylene sheet (POM: thickness 0.5 mm, density 1.41 g/cm 3 ) purchased from Vink Kunststoffen BV (Didam, The Netherlands) was cut into strips (1 cm × 1 cm) and cold-extracted with hexane (30 min) and methanol (3 times for 30 min), as previously described by Jonker and Koelmans (2001). Pyrene (99.0%) and pyrene-d10 (99.5%) were purchased from Dr. Ehrenstorfer (Germany). All stock solutions were prepared in methanol. The hexane and methanol were of residue analysis grade (from Lab Scan, Dublin, Ireland, and Acros Organics, Geel, Belgium). 2.2. Sorption experiments All experiments were carried out at 20 ± 1 °C using a MilliQ water background solution prepared with 25 mg/L NaN3 as biocide (pH 7.06). One milligram of CNTs and 50 mL of HA solution (0, 5 and 40 mg/L) were added to glass vials. Samples were then pretreated by (i) shaking by hand for 1 minute or (ii) sonication for 2 h. For the sonication, 20 sample vials were immersed into the sonication bath (Bandelin sonorex super rk106 bath, 35 kHz, diameter 24.5 cm). A POM strip (approximately 100 mg) was subsequently added to each sample and pyrene was spiked. The concentration of methanol (spiking solvent) was kept at 0.16% for all samples in order to minimize solvent effects. For each of the two pretreatments, sorption coefficients were measured for an initial pyrene concentration of 50 μg/L. In addition, full sorption isotherms for pyrene were measured at 0 mg/L HA (pyrene equilibrium
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
concentration ranged over five orders of magnitude from 0.0001 to 20 μg/L). All vials were shaken horizontally (180 rpm) for 28 days to ensure equilibration. The POM strips were then removed from the vials, rinsed with deionized water and wiped with a wet tissue. They were then extracted with methanol by accelerated solvent extraction (ASE 200, Dionex, USA; 1500 psi, 100 °C) using pyrene-d10 as an internal standard. Extracts were concentrated under N2 prior to gas chromatography–mass spectrometry analysis. Details about the method validation (e.g., equilibration time, extraction efficiency, recovery and reproducibility) can be found in our previous study (Kah et al., 2011). Blanks containing no pyrene were prepared for each pretreatment and used for characterization (see the following section). Control samples containing pyrene indicated that losses were b 6% after 28 days of shaking. The sorption of pyrene to CNTs was calculated based on mass balance, as described in our previous study (Zhang et al., 2012). Our previous experience with the method applied to similar materials indicated a typical maximum 10% error for triplicates (Hüffer et al., 2013; Kah et al., 2011). Instead of repeating samples, priority was here given to generate data for more concentration levels along the isotherm in order to support more robust isotherm fits. 2.3. Characterization of CNTs The CNTs were imaged by scanning electron microscopy (SEM) after the sorption experiments, for each type of CNT and each pretreatment (more details available in the SI). The absorbance of CNT suspensions was measured by UV–vis spectrometry at 800 nm (Varian Cary 50 UV–vis spectrophotometer) as an indication for suspension stability. With the aim to further compare the stability of the suspensions, the average hydrodynamic diameter of CNT aggregates remaining in suspension after 2 days of settling was measured by dynamic light scattering (DLS, Malvern ZetaSizer Nano). Size distributions for the whole suspensions were determined using a particle size analyzer (Eyetech, based on the time of transition principle) and a Malvern Mastersizer (MasterSizer 2000, based on a laser light scattering principle). The CNT surface functional groups were detected with Fourier transform infrared spectroscopy (FTIR, Bruker, Tensor 27) and X-ray photoelectron spectroscopy (XPS, Kratos Axis Ultra electron spectrometer). The bulk chemistry of the CNTs was further characterized by energy dispersive X-ray spectroscopy (EDX) and isotope-ratio mass spectrometer (IRMS: further details available in the SI). The specific surface area and pore volume were measured by N2 sorption (NOVA2000, Quantachrome). 2.4. Sorption models and statistics Six sorption models that have previously been applied to describe sorption by carbonaceous materials were fitted to the isotherms: these being the Freundlich, Langmuir, dual Langmuir, Toth, dual-mode and Dubinin–Ashtakhov models. Descriptions of the models and their parameters are available in Table S2. The goodness of fit was evaluated and compared on the basis of the r2 values, mean weighted square errors and Akaike's information criterion, as described previously by Kah et al.
135
(2011). All statistical tests and parameter optimizations were performed with SigmaPlot 11.0 for Windows. 3. Results and discussion Investigations into the sorption behavior of CNTs are often based on the Freundlich model. However, Fig. S1 and S2 clearly show that this model is not suitable for fitting pyrene sorption isotherms over a wide range of concentrations. Of the six sorption models tested, and based on a multicriteria comparison of the goodness of fit (Kah et al., 2011), the Dubinin–Ashtakhov and Toth models gave the best isotherm fits for all types of CNTs (Table S3 and Fig. S1–S2). The parameters derived by fitting (Table 1) are used below to discuss the discrepancies observed in terms of sorption affinity (E and Kt), maximum capacity (Q0) and heterogeneity (t). 3.1. Effects of functional groups (without sonication) The effect that CNT oxidation has on sorption affinity for nonpolar compounds remains unclear as previous work has reported an increase (Lu et al., 2008; Yu et al., 2011), a decrease (Cho et al., 2008) and no effect (Wu et al., 2012). Fig. 1 suggests that after shaking only (i.e., with no sonication or addition of HA, open symbols), the sorption behavior varies with the type of CNT functionalization. In order to better understand these variations, each type of CNTs has been extensively characterized. Analyzing the level of oxidation of CNTs is known to be a difficult task (Wepasnick et al., 2010), with each technique having its own strengths and limitations. We therefore applied several different approaches (i.e., EDX and IRMS as bulk analysis, and XPS as surface analysis) in combination with the data provided by the supplier (XPS and titration; Tables S1 and S4). Larger levels of oxidation were generally measured for COOH-CNTs. However, neither absolute values nor trends in oxygen content were consistent across methods for the other types of CNTs. The relatively large O% measured for NH2-CNTs by all methods was unexpected. XPS analysis also indicates that NH2-CNTs contain much less NH2 groups than suggested by the supplier (about 0.25% against 7%). Variability from batch to batch is likely to occur when producing CNTs. Our results highlight the importance of characterizing purchased materials using several complementary techniques, which can quickly become a very time- and budget-consuming task. In literature, the concentration of specific functional groups at the surface of CNTs is often solely derived from XPS data. However, because of ambiguities in spectral interpretation (Wepasnick et al., 2010) and the fact that the standard deviation is often not reported, there is a risk to over interpret differences between materials. In addition to spectroscopic approaches, a number of indirect indications for surface functionalization can also be explored. For instance, it is worth noting that prior to sonication, the zeta potential of all types of CNT aggregates was close to neutral or only slightly negative (Table S5). The pH in this study being close to neutral, the zeta potentials of OHand COOH-CNTs were expected to be negative and that of NH2-CNTs was expected to be positive (Schaffer and Licha, 2014). The lack of a
Table 1 Fitting parameters ± standard error and r2 for Dubinin–Ashtakhov and Toth models fitting the isotherms of pyrene on three types of CNTs pretreated by shaking only or by sonication. Dubinin–Ashtakhov model
Toth model
Shaking only
log Q0
E (kJ/mol)
b
r2
log Q0
Kt (L/μg)
t
r2
OH-CNTs COOH-CNTs NH2-CNTs
7.54 ± 0.08 7.58 ± 0.06 7.57 ± 0.05
18.38 ± 0.93 19.48 ± 0.63 22.02 ± 0.47
2.24 ± 0.25 2.27 ± 0.16 3.01 ± 0.18
0.94 0.96 0.99
7.55 ± 0.12 7.59 ± 0.06 7.56 ± 0.10
1.67 ± 0.35 2.21 ± 0.29 4.40 ± 2.19
0.58 ± 0.13 0.54 ± 0.07 0.78 ± 0.21
0.95 0.97 0.99
Sonication
log Q0
E (kJ/mol)
b
r2
log Q0
Kt (L/μg)
t
r2
OH-CNTs COOH-CNTs NH2-CNTs
7.73 ± 0.10 7.72 ± 0.03 7.70 ± 0.04
23.60 ± 1.31 23.42 ± 0.45 23.48 ± 0.56
2.28 ± 0.29 2.18 ± 0.09 2.32 ± 0.13
0.99 0.99 0.99
7.88 ± 0.21 7.95 ± 0.08 7.84 ± 0.09
3.49 ± 1.04 3.16 ± 0.30 3.48 ± 0.47
0.30 ± 0.07 0.26 ± 0.02 0.31 ± 0.03
0.99 0.99 0.99
136
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
Adsorbed Concentrations,C CNT (µg/kg)
1e+8
1e+7
increases (Smith et al., 2009; Wu et al., 2012), no effect (Cho et al., 2008) and decreases (Gotovac et al., 2007; Lu et al., 2008; Zhang et al., 2009) in the surface area after CNT functionalization have all been reported by different authors. In view of the uncertainty relating to BETnitrogen surface area measurements and the current lack of a wellaccepted alternative (especially for measurements in suspension), great care is required when attempting to interpret normalization by the specific surface area. Finally, no significant difference in surface heterogeneity between the different types of functionalization (t, p N 0.05, Table 1) was observed, probably due to insufficient levels of functionalization. Overall, the type of CNT functionalization had a significant effect on the sorption affinity for pyrene (after shaking only), but no effect on the maximum sorption capacity or the surface heterogeneity. Elucidating the driving mechanism(s) will remain difficult until more reliable synthesis and characterization techniques for suspensions are developed, permitting more accurate descriptions of the nature, density and distribution of functional groups on, and within, the CNT aggregates.
Sonication OH-CNTs COOH-CNTs NH2-CNTs
1e+6
Shaking OH-CNTs COOH-CNTs NH2-CNTs
1e+5
Dubinin-Ashtakhov model Toth model 1e+4 1e-5
1e-4
1e-3
1e-2
1e-1
1e+0
1e+1
1e+2
Equilibrium Concentrations, Cw (µg/L) Fig. 1. Sorption isotherms of pyrene to three types of CNTs pretreated by shaking and by sonication. The lines represent the fits by the Dubinin–Ashtakhov (DAM) and Toth (TM) models.
clear surface charge indicates that the number of functional groups located at the surface of the aggregates was low and/or that a mixture of cationic and anionic functional groups co-existed. Negative zeta potentials for NH2 functionalized CNTs have previously been reported by other authors across a wide pH range (Liao et al., 2008; Wang et al., 2012, 2013). The application of three spectroscopic methods here confirmed that negative charges are much likely due to the presence of dissociated oxygen-containing functional groups on NH2-CNTs. Sorption affinity increased in the following order: OH-CNTs b COOHCNTs b NH2-CNTs (E and Kt, p b 0.001; Fig. 1 and Table 1). Overall, there appeared to be no clear relationship between sorption affinity and functionalization, suggesting that additional factors need to be taken into account in order to explain the variations in sorption affinity. The strongest affinity of pyrene, which was for NH2-CNTs, could be explained by the overall higher density of electron-donating groups (N- and Ocontaining groups) but are unlikely to be mainly driven by NH2 groups (which would have been concluded if relying solely on the data from the supplier). The maximum sorption capacity of functionalized CNTs for hydrophobic compounds was expected to be mainly influenced by (i) water clusters surrounding functional groups (negative effect, Cho et al., 2008; Wu et al., 2012; Zhang et al., 2009) and (ii) an increase in the total accessible surface area through enhanced dispersion (positive effect, Yu et al., 2011). However, we observed similar maximum sorption capacity for all types of CNTs (p N 0.05, Table 1). None of the CNT suspensions achieved a good dispersion status after 6 days shaking only (see pictures on Fig. S3). SEM observations (Fig. S4a) suggest very similar aggregate sizes and structures. After 28 days shaking, the proportion of CNTs remaining in suspension increased slightly in the following order: OH-CNTs b COOH-CNTs b NH2-CNTs (based on UV absorbance measurements, Fig. S5a) but remained very low overall. The specific surface area after shaking decreased in the following order: OH-CNTs N COOH-CNTs N NH2-CNTs (Table S6). The normalization of the maximum sorption capacity by the specific surface area of CNTs can be useful for ruling out the influence of variations in surface area with the type of CNT functionalization. We found that the surface-normalized sorption capacity followed the order OH-CNTs b COOH-CNTs b NH2CNTs, consistent with the sorption affinity. It is important to note that
3.2. Effect of sonication Sonication efficiently dispersed the three types of functionalized CNTs, as indicated by higher UV absorbance values and smaller size distributions (Figs. S5c, S6 and S7). A slightly better dispersion of NH2-CNTs was observed immediately after sonication, as indicated by smaller aggregate size, higher absorbance and more negative zeta potential (Figs. S5, S7 and Table S5). However, the stability of all CNT suspensions was poor and reaggregation occurred rapidly (Figs. S5d and S7). Sonication significantly increased both the maximum sorption capacity and the sorption affinity of all types of CNTs for pyrene (p b 0.05, Table 1 and Fig. 1). These increases could be well explained by an enlargement of the pore spaces that occur between the aggregates or between individual CNTs, after the dispersion–reaggregation event. It has been previously reported that sorbates have similar access to surfaces located within CNT aggregates and those located at their surfaces (Cho et al., 2008). It is, however, important to stress that this conclusion was made after the application of a sonication treatment, and for a relatively small sorbates (e.g., naphthalene as in Cho et al., 2008). The increase of both the maximum sorption capacity and the sorption affinity observed here for pyrene suggests that a significant proportion of the CNT aggregate surface was unavailable prior to sonication. The effect of sonication on the sorption behavior of CNTs with different functionalization has not previously been reported. We saw in the section above that the sorption affinity for pyrene significantly increased in the order OH-CNTs b COOH-CNTs b NH2-CNTs, but no more differences in the sorption isotherms could be distinguished following the sonication treatment (p N 0.05, Fig. 1). It is also worth mentioning that isotherms are identical to those we previously obtained after sonication of non-functionalized CNTs, with similar physical characteristics, but obtained from an another provider (see Table S3 and Zhang et al., 2012). This similarity in behavior after sonication can be explained by the strong and irreversible effects that sonication has on both the surface chemistry and aggregate morphology of CNTs. A number of indicators suggest that sonication had a significant effect on the surface chemistry of all types of CNTs: the zeta potential became more negative (from about − 3 to − 25 mV, Table S5), the oxygen content increased (according to EDX and IRMS techniques, Table S4) and the FTIR spectra suggested the generation of OH and C = O groups (Fig. S8). The increase in surface heterogeneity for all types of CNTs (decrease in t value, Table 1) could also be explained by the generation of oxygen-containing functional groups, as well as defects such as bending, buckling and breaking (Vichchulada et al., 2010). The degree of initial functionalization (possibly complemented by oxidation during sonication) remained, however, relatively low. We believe that it is very likely that the pore enlargement—possibly maintained after reaggregation owing to the defects generated during
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
sonication—was mainly responsible for the observed large increase in sorption. Hou et al. (2013) recently proposed that looser aggregates allow sorbate molecules to simultaneously interact with several CNTs. Such a phenomenon could also explain the enhanced sorption affinity that we observed following sonication. The effect of pore enlargement on sorption, however, is difficult to characterize as it involves determining the effective surface area available for the sorption of a given sorbate. On one hand, the nitrogen adsorption method determines the surface area accessible to N2 (i.e., a relatively small sorbate) and is subject to uncertainties due to unavoidable artifacts introduced during sample preparation (as shown in Table S6), while on the other hand, values derived from the size distribution of the aggregates only reflect the external surface areas and certainly underestimate the total surface area available for sorption. This is particularly true for small sorbates that can access a higher proportion of pores than larger sorbates. It has previously been suggested that surface chemistry has a greater effect than dispersion status on the sorption behavior of CNTs (Zhang et al., 2009). However, no measurements on dispersed CNT systems were yet available at that time. Our results highlight the importance of CNT aggregate morphology, which can be greatly affected by a dispersion event. After sonication, the physical characteristics of the CNTs (e.g., their diameters and the number of similar nanotube walls in the CNTs investigated) became the main drivers of their sorption behavior. Sonication is frequently applied when studying CNTs, and the impact that this has on aggregate structures and consequent sorption behavior should therefore not be neglected (changes of up to 1.5 orders of magnitude have been observed, based on the single Kd values presented in Fig. 2). Different observations could be expected for highly functionalized CNTs that disperse spontaneously in water and expose all of their surface areas to potential sorbates. Functionalization by plasma or acid treatments has, to date, resulted in a maximum of 15% wt. functionalization, which is insufficient for spontaneous dispersion in water (Sahoo et al., 2010). 3.3. Effects of HA The presence of HA efficiently dispersed all types of CNTs, but only when combined with a sonication treatment (Fig. S3). Absorbance and size measurements indicate that the stability of OH-, COOH- and NH2CNT suspensions were relatively similar (Figs. S5c, S6 and S7a). The reaggregation of all types of CNTs occurred during shaking after sonication but to a lesser extent than in the absence of HA (Figs. S3, S5d, S6 and S7b), probably due to the electrosteric stabilization provided by HA molecules adsorbed onto CNTs. The addition of HA certainly increased the negative charge of CNT aggregates significantly (more negative zeta potential in Table S5). After shaking only, there was a clear increase
log KCNT (L/Kg)
10
9
8
7
6 OH-
COOH-
NH2-
Fig. 2. Single distribution coefficients of pyrene to three types of CNTs (log KCNT, L/kg), as affected by HA concentrations (□, , are 0, 5, 40 mg/L HA, respectively) and two pretreatments (□ shaking and sonication).
137
in the negative surface charge following the order OH-CNTs b COOHCNTs b NH2-CNTs. Following sonication in the presence of HA, the surface charges of all types of CNTs were similar (i.e., around − 50 mV and − 60 mV at 5 and 40 mg/L of HA, respectively). This contrasts with the results obtained by Smith et al. (2012), who reported that the effective surface charge of oxidized CNTs decreased as the concentration of natural organic matter increased. Changes in surface charge can be expected to depend on the nature of the organic matter. That investigated by Smith et al. (2012) was from Great Dismal Swamp (Virginia) and presumably possessed a lower density of negatively charged groups than the Suwanee River HA used in our investigations. Increasing the concentration of HA greatly suppressed the sorption affinity for pyrene of OH-, COOH- and NH2-CNTs (Fig. 2, p b 0.001), especially after sonication. This contrasts with results we previously obtained for non-functionalized CNTs (Zhang et al., 2012), which showed that sorption of pyrene was relatively insensitive to the presence of HA (a decrease in sorption was only observed after sonication). Discrepancies in behavior between functionalized and non-functionalized CNTs could be, at least partially, explained by differences in their interactions with HA. For instance, sonication was previously reported to increase the amount of peat HA sorbed to CNTs by a factor of approximately two (Zhou et al., 2012). Data on the sorption of natural organic matter to functionalized CNTs remain limited. While a similar affinity of peat HA for pristine and COOH-functionalized CNTs has previously been reported (Zhou et al., 2012), a negative impact of CNT oxidation on the sorption of NOM has also been observed (Smith et al., 2012; Wang et al., 2013). Integrating our results for suspension stability (Fig. S7), size of aggregates (Figs. S6 and S7) and suppression of pyrene sorption (Fig. 2) suggests an affinity of HA for CNTs that increases in the following order: NH2-CNTs b OH-CNTs ≈ COOH-CNTs. Further research will be required to elucidate the mechanisms involved in the interactions between CNTs and natural organic matter, and to understand how these vary with the type of CNT, with the type of natural organic matter and over a range of water chemistries. Overall, increasing the HA concentration generally decreased the sorption affinity for pyrene, much likely due to a combination of competition and pore blockage. It is important to note, however, that even at a relatively high HA concentration (40 mg/L), the sorption affinity for pyrene remained very strong, about three orders of magnitude larger than that of pyrene to HA (i.e., log KCNT N 7.4 compared to log KHA 4.7; Zhang et al., 2012). 4. Conclusions One of our initial objectives was to generate robust sorption data for NH2-CNTs, for which no sorption data for hydrophobic compounds had previously been published. However, comprehensive surface characterization indicated that purchased NH2-CNTs contained only a small amount of amino group and a relatively large amount of oxygen containing groups, highlighting the necessity to extensively characterize materials using complementary techniques. When studied as purchased (in the absence of HA and sonication treatment), sorption isotherms over a wide range of pyrene concentrations showed that OH-, COOH- and NH2-CNTs exhibited significantly different sorption affinity. Despite extensive characterization, discrepancies due to surface chemistry and dispersion were difficult to distinguish. Elucidating the mechanisms involved will require further work with materials that are functionalized more homogeneously, and to a higher level. Commercially available CNTs are nevertheless essential to study, in view of the large quantities produced and consequent high environmental relevance (relative to purer materials produced in the laboratory in much smaller quantities). Sonication greatly increased both the sorption affinity and maximum capacity of all types of functionalized CNTs, to an extent that overwhelmed the differences initially observed. Results demonstrate that a significant proportion of the CNT surface was unavailable to
138
M. Kah et al. / Science of the Total Environment 497–498 (2014) 133–138
pyrene prior to sonication. Sonication treatment is frequently applied when studying CNTs, and its irreversible effect on sorption behavior should not be neglected (increase of up to 1.5 orders of magnitude was observed, leading to a log Kd value close to 9). Even though the presence of HA reduced the sorption affinity, the exceptional sorption properties of CNTs were still maintained, at least in part, which suggests that CNTs can act as strong sorbents under a wide range of conditions. Acknowledgments The authors would like thank Petra Körner for her invaluable support in the laboratory. We are grateful to Gerlinde Habler and Frank von der Kammer for electron microscope imaging, to Eugen Libowitzky for FTIR measurements (University of Vienna, AT) and to Andrey Shchukarev for XPS analysis (Umeå University, SE). Xiaoran Zhang was financially supported by the China Scholarship Council. Appendix A. Supplementary data Details of CNT characterization, isotherm fits and parameters for the six models evaluated. Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.scitotenv.2014.07.112 References Cho HH, Smith BA, Wnuk JD, Fairbrother DH, Ball WP. Influence of surface oxides on the adsorption of naphthalene onto multiwalled carbon nanotubes. Environ Sci Technol 2008;42:2899–905. Dervishi E, Li ZR, Xu Y, Saini V, Biris AR, Lupu D, et al. Carbon nanotubes: synthesis, properties, and applications. Part Sci Technol 2009;27:107–25. Gotovac S, Yang CM, Hattori Y, Takahashi K, Kanoh H, Kaneko K. Adsorption of polyaromatic hydrocarbons on single wall carbon nanotubes of different functionalities and diameters. J Colloid Interface Sci 2007;314:18–24. Hou L, Zhu D, Wang X, Wang L, Zhang C, Chen W. Adsorption of phenanthrene, 2-naphthol, and 1-naphthylamine to colloidal oxidized multiwalled carbon nanotubes: Effects of humic acid and surfactant modification. Environ Toxicol Chem 2013;32:493–500. Hüffer T, Kah M, Hofmann T, Schmidt TC. How redox conditions and irradiation affect sorption of PAHs by nC60. Environ Sci Technol 2013;47:6935–42. Jonker MTO, Koelmans AA. Polyoxymethylene solid phase extraction as a partitioning method for hydrophobic organic chemicals in sediment and soot. Environ Sci Technol 2001;35:3742–8. Kah M, Zhang XR, Jonker MTO, Hofmann T. Measuring and modelling adsorption of PAHs to carbon nanotubes over a six order of magnitude wide concentration range. Environ Sci Technol 2011;45:6011–7. Liao Q, Sun J, Gao L. Adsorption of chlorophenols by multi-walled carbon nanotubes treated with HNO3 and NH3. Carbon 2008;46:553–5. Lu C, Su F, Hu S. Surface modification of carbon nanotubes for enhancing BTEX adsorption from aqueous solutions. Appl Surf Sci 2008;254:7035–41.
Ma JP, Xiao RH, Li JH, Yu JB, Zhang YQ, Chen LX. Determination of 16 polycyclic aromatic hydrocarbons in environmental water samples by solid-phase extraction using multiwalled carbon nanotubes as adsorbent coupled with gas chromatography–mass spectrometry. J Chromatogr A 2010a;1217:5462–9. Ma PC, Siddiqui NA, Marom G, Kim JK. Dispersion and functionalization of carbon nanotubes for polymer-based nanocomposites: a review. Compos A: Appl Sci Manuf 2010b;41:1345–67. Mauter MS, Elimelech M. Environmental applications of carbon-based nanomaterials. Environ Sci Technol 2008;42:5843–59. Pan B, Xing BS. Manufactured nanoparticles and their sorption of organic chemicals. Adv Agron 2010;108(108):137–81. Petersen EJ, Zhang LW, Mattison NT, O'Carroll DM, Whelton AJ, Uddin N, et al. Potential release pathways, environmental fate, and ecological risks of carbon nanotubes. Environ Sci Technol 2011;45:9837–56. Sahoo NG, Rana S, Cho JW, Li L, Chan SH. Polymer nanocomposites based on functionalized carbon nanotubes. Prog Polym Sci 2010;35:837–67. Schaffer M, Licha T. A guideline for the identification of environmentally relevant, ionizable organic molecule species. Chemosphere 2014;103:12–25. Schwyzer I, Kaegi R, Sigg L, Magrez A, Nowack B. Influence of the initial state of carbon nanotubes on their colloidal stability under natural conditions. Environ Pollut 2011; 159:1641–8. Smith B, Wepasnick K, Schrote KE, Cho HH, Ball WP, Fairbrother DH. Influence of surface oxides on the colloidal stability of multi-walled carbon nanotubes: a structure–property relationship. Langmuir 2009;25:9767–76. Smith B, Yang J, Bitter JL, Ball WP, Fairbrother DH. Influence of surface oxygen on the interactions of carbon nanotubes with natural organic matter. Environ Sci Tech 2012; 46:12839–47. Vichchulada P, Cauble MA, Abdi EA, Obi EI, Zhang QH, Lay MD. Sonication power for length control of single-walled carbon nanotubes in aqueous suspensions used for 2-dimensional network Formation. J Phys Chem C 2010;114:12490–5. Wang F, Zhu DQ, Chen W. Effect of copper ion on adsorption of chlorinated phenols and 1-naphthylamine to surface-modified carbon nanotubes. Environ Toxicol Chem 2012;31:100–7. Wang F, Yao J, Chen H, Yi Z, Xing B. Sorption of humic acid to functionalized multi-walled carbon nanotubes. Environ Pollut 2013;180:1–6. Wepasnick KA, Smith BA, Bitter JL, Fairbrother DH. Chemical and structural characterization of carbon nanotube surfaces. Anal Bioanal Chem 2010;396:1003–14. Wu WH, Chen W, Lin DH, Yang K. Influence of surface oxidation of multiwalled carbon nanotubes on the adsorption affinity and capacity of polar and nonpolar organic compounds in aqueous phase. Environ Sci Technol 2012;46:5446–54. Yu F, Ma J, Wu Y. Adsorption of toluene, ethylbenzene and m-xylene on multi-walled carbon nanotubes with different oxygen contents from aqueous solutions. J Hazard Mater 2011;192:1370–9. Zhang SJ, Shao T, Bekaroglu SSK, Karanfil T. The impacts of aggregation and surface chemistry of carbon nanotubes on the adsorption of synthetic organic compounds. Environ Sci Technol 2009;43:5719–25. Zhang XR, Kah M, Jonker MTO, Hofmann T. Dispersion state and humic acids concentrationdependent sorption of pyrene to carbon nanotubes. Environ Sci Technol 2012;46: 7166–73. Zhou XZ, Shu L, Zhao HB, Guo XY, Wang XL, Tao S, et al. Suspending multi-walled carbon nanotubes by humic acids from a peat soil. Environ Sci Technol 2012;46: 3891–7.
