Sorption of phenanthrene on single-walled carbon nanotubes modified by DOM: effects of DOM molecular weight and contact time

Abstract

The impacts on the sorption of phenanthrene (PHE) onto single-walled carbon nanotubes (SCNT) by loading of dissolved organic matter (DOM) fractions with different molecular weights (MWs) were studied. Moreover, the influence of contact time on the sorption capacity of DOM-modified SCNT was evaluated for the first time. Preloading of DOM on SCNT for 2 days led to a reduction in PHE sorption and the nonlinearity of the sorption isotherm; however further increasing the contact time between the DOM and SCNT from 2 days to 20 days enhanced the sorption capacity, and the sorption nonlinearity increased also. This is due to the transferring of DOM molecules from the external surface to interstitial channels trapped in SCNT particles, restoring the sorption sites. The loading of DOM fraction with MWs larger than 14 000 Da (DOM_>14 000) brought a stronger suppression of PHE sorption by SCNT compared to smaller ones (DOM_<1000 and DOM_{1000–14 000}), with the Freundlich nonlinear index n increasing from 0.236 to 1.093 and the Freundlich affinity coefficient log K_F decreasing from to 7.34 to 4.57. These are similar to the sorption characteristics of DOM_>14 000, suggesting a complete coverage of the SCNT by DOM_>14 000 molecules. Though the loading of DOM fractions with smaller MWs made SCNT more polar, the reduction in PHE sorption on DOM-preloaded SCNT was limited due to their lesser sorbed amount and smaller molecular size.

---

Environmental impact

Once CNT are released into the environment, they will inevitably interact with DOM, which would in turn affect the environmental behavior of the CNT and their interaction with organic pollutants. Both the chemical composition and steric conformation of DOM were found to influence its interaction with SCNT. DOM fractions with larger MWs (DOM_>14 000) showed stronger interactions with SCNT and a greater binding capacity for PHE than smaller ones (DOM_<1000 and DOM_{1000–14 000}), leading to the strongest suppression of sorption of PHE. Increasing the contact time between the DOM and SCNT from 2 days to 20 days enhanced the sorption capacity and increased the sorption nonlinearity of PHE on DOM-preloaded SCNT.

Introduction

Carbon nanotubes (CNT) have attracted great attention due to their unique properties, such as electrical conductivity, optical activity, and mechanical strength, since their discovery by Iijima in 1991.¹ This fascinating new class of materials has shown promising applications in many areas, including battery and fuel electrodes, catalysts, super capacitors, conductive coating, sensors, and adsorbents in water treatment.^2,3

Serious environmental concerns have been aroused regarding the fate and impacts of CNT since they might be released into the environment with increasing manufacture and applications. CNT exhibit great sorption capacity for organic contaminants, especially hydrophobic organic contaminants (HOCs) because of their great surface areas and highly hydrophobic nature, and the sorption coefficients on CNT for most HOCs are usually hundredfold that of those on natural geosorbents.^4,5 Hence, one major concern is that the existence of CNT would alter the fate, transport and bioavailability of HOCs in the environment. Both inhibition and enhancement of the bioavailability of organic contaminants due to the sorption onto CNT have been reported. For example, addition of 3.0 mg g⁻¹ CNT in soil substantially decreased the bioaccumulation of pyrene by earthworms.⁶ This is in agreement with the common recognition that sorbed organic chemicals on solids are less bioavailable. However, opposite results were also reported. The existence of CNT increased accumulation of sorbed pyrene in Lumbriculus variegatus and the authors ascribed this to the easier passage of nanosized CNT into and through organisms.⁷ Similarly, the mobility of HOCs may change greatly which is controlled by both the sorption capacity and the stability of CNT in the water phase. Hence the sorption of organic chemicals on CNT has drawn much attention, and considerable literature has come forth.^6–9 Multiple mechanisms have been proposed for the sorption by CNT, such as hydrophobic interactions, π–π electron donor–acceptor interactions, hydrogen bonds and electrostatic interactions.⁸

Dissolved organic matter (DOM) is ubiquitous in the aquatic environment,^10,11 and it is composed of a complex mixture of organic compounds with varying molecular weights (MWs), from small hydrophilic acids, amino acids, proteins to large humic and fulvic acids. DOM affects the distribution of organic compounds between the interface of solution and solids through several pathways. Kilduff and Wigton found that the presence of DOM could reduce the sorption of trichloroethylene (TCE) on activated carbon due to pore blockage of the activated carbon, which can reduce the effective surface area available to TCE.¹² Pignatello et al. reported that humic substances in adsorbed and coflocculated states inhibited adsorption of organic chemicals onto black carbon because of the similar mechanisms.¹³ An opposite result has also been reported by Murphy et al. that DOM coating would enhance the adsorption of HOCs on natural mineral particles.¹⁴ The effects of DOM depend on the nature of the solids as well as the physicochemical properties of DOM, such as polarity, molecular size and configuration, aromaticity and alkyl moieties. It has been reported that different DOM fractions show different properties. Kang et al. found that the macromolecule fraction of DOM contained higher amounts of aromatic moieties and carbohydrate-like substances and more hydrophobic groups.¹⁵ This will lead to different effects on the fate of organic chemicals. Chin et al. observed a strong correlation between the pyrene binding coefficient, K_doc, MWs and aromaticity of the aquatic humic substances.¹⁶ Chen et al. reported that DOM fractions with high MW had more effect on the mobility of prometryne in soils than those with low MW.¹⁷

Once CNT are released into the environment, they will inevitably interact with DOM, which would in turn affect the environmental behavior of CNT and their interaction with organic pollutants. Hyung et al. reported that CNT can interact with DOM in aquatic systems which could enhance their stability and transport in aquatic systems.¹⁸ It was also observed that CNT had higher sorption capacity for humic acids of higher MW and greater hydrophobicity than for fulvic acids.¹⁹ Zhang et al. reported that the presence of natural organic matter (NOM) greatly suppressed the adsorption of phenanthrene (PHE), biphenyl, and 2-phenylphenol by CNT and the effects of NOM were determined by the planarity and hydrophobicity of these three organic chemicals.²⁰ Wang et al. reported that peptone coating made the strongest suppression of the sorption of PHE, naphthalene, and 1-naphthol by CNT among three DOMs (humic acid, peptone, and α-phenylalanine).²¹ However, so far no study has examined the influence of DOM fractions with different MWs on the adsorptive behavior of HOCs by CNT.

Besides, it is well known that many interactions in the environment are long-term processes, which are quite different from laboratory experiments with a short time usually of one or two days.^22,23 It has been well documented that both the sorption amount and combination state of organic chemicals on geosorbents change with increasing contact time (aging effect); and the change in the combination state of sorbed chemicals leads to decreased desorption and bioavailability.^24,25 In our earlier study, we found that extending contact time between soil and black carbon led to a reduced sorption capacity of the mixture in the presence of water, but no significant change was found under dry conditions.²⁶ The reduced sorption capacity of the mixture was ascribed to the interaction of soil organic matter with black carbon, which was facilitated by water molecules. The interaction of DOM and CNT may also change with extended time, which may influence the environmental behavior of their complex in the environment. However, sorption on DOM–CNT complexes under different contact times has never been studied.

Therefore, the objectives of this study were to investigate the influence of contact time and DOM chemical properties on the sorption capacity of DOM-loaded CNT for organic chemicals and the mechanisms therein. To this end, we studied the sorption of PHE on original commercial single-walled carbon nanotubes (SCNT), SCNT preloaded with bulk DOM (extracted from a peat) and its fractions with different MWs, and SCNT preloaded with DOM experiencing different contact times.

Materials and methods

Materials

The SCNT (2–10 nm diameter × 1–5 μm length) were purchased from Sigma-Aldrich Chemical Company (MO, USA) and were used as received. PHE (purity >98%) was purchased from Acros Corporation (NJ, USA).

The DOM was extracted from a peat material collected from Mount Wuyi, China with double-distilled water using a solid to water ratio of 1 : 2.5 (w/v, dry weight basis) on a magnetic stirrer for 24 h at room temperature. After the suspension was centrifuged at 3000 rpm for 30 min, the supernatant was filtered through a 0.45 μm membrane, and the filtrate was stored at 4 °C. Cellulose dialysis membrane bags (Spectrum Inc., Texas, USA) were used to separate DOM fractions at MWs of 1000 and 14 000 Da as described by Cheng and Wong.²⁷ The bags were cleaned by overnight soaking in deionized water and thorough rinsing before use. Fifty milliliters of the DOM extract was added to a dialysis bag with a MW cutoff of 1000 Da, and dialysis was conducted in 100 mL of deionized water in the dark at 10 °C with magnetic stirring for 40 h. Then, the solution that remained in the bag was transferred to another dialysis bag with a MW cutoff of 14 000 Da, and the dialysis was conducted under the same conditions as described above. The three fractions are designated as DOM_<1000, DOM_{1000–14 000}, and DOM_>14 000, respectively. The NaN₃ used in the aqueous stock solutions as the biocide to inhibit microorganisms was of analysis grade.

Characterization of SCNT

The SCNT were measured for Brunauer–Emmett–Teller (BET) surface areas (S_BET) and pore size distribution based on nitrogen sorption–desorption isotherms at 77 K using a high-resolution gas adsorption analyzer (AutoSorb-1-MP, Quantachrome, USA). DFT (density functional theory) was used to analyze the pore size distribution using a Quantachrome NOVA automated gas sorption system, NOVA for Windows Version 1.12.²⁸ All samples were outgassed at 115 °C for 10 h before N₂ sorption. Bulk elemental contents (C, H, and N) of the samples were measured using a CHN elemental analyzer (Elementar Vario EL CUBE, Germany) and are reported as w/w percentages. The amount of ash was determined by mass loss after heating the samples at 900 °C for 10 h. Oxygen content was calculated by mass difference. To get information on surface functionalities and elemental composition of the samples, X-ray photoelectron spectra (XPS) were obtained using a Kratos Axis Ultra DLD multi-technique X-ray photoelectron spectrometer (Shimadzu Corporation, Japan). Binding energies were calibrated using the C1s hydrocarbon peak at 284.8 eV with deconvolution processing of the C1s spectra by the Xpspeak software package, and the C1s binding energy levels were assigned as follows: 284.6 eV to C–C, 286.2 eV to C–H, 287.6 eV to C=O, and 289.1 eV to COO–.²⁹

Sorption experiments

Sorption experiments were carried out in 40 mL US EPA standard sample vials with Teflon-lined septa and screw caps. Five milligrams of SCNT powder was weighed into the vials containing 40 mL of 200 mg L⁻¹ NaN₃ solution, and specific amounts of PHE stock solution were added to the vials as experimental design to initiate the sorption. The vials were shaken for 8 days in a shaker operated at 150 rpm and 20 ± 0.5 °C in the dark. The data of the sorption kinetic experiments showed that it took 5 days for PHE to obtain apparent sorption equilibrium on SCNT and DOM-loaded SCNT. When the impacts of the contact time of DOM with SCNT were studied, the extracted bulk DOM was allowed to react with the SCNT for 2 and 20 days to represent fresh and aged DOM_bulk–SCNT. For the influence of DOM fractions with different MWs, the solutions of DOM MW fractions were allowed to react with SCNT for 2 days, which are denoted as DOM_>14 000–SCNT, DOM_{1000–14 000}–SCNT, and DOM_<1000–SCNT, respectively. The levels of DOM before and after interaction with SCNT were recorded. Phenanthrene was spiked directly in vials without separation of DOM residue in the solution phase to initiate the sorption. This design is thought more realistic since DOM exists both in the solution phase and on the solid phase in the natural environment. After sorption experiments, the solution and the SCNT were separated by centrifugation at 2000 rpm for 20 min. An appropriate aliquot of the supernatant was removed, and PHE concentration was analyzed. All sorption experiments were conducted in duplicates and all the data in figures are the average of two replicates.

PHE binding to bulk DOM and its MW fractions

To elucidate the binding capacity and possible impact of DOM on PHE, PHE binding to DOM_bulk and its MW fractions in solution were measured. Concentrations of DOM_bulk and its MW fractions were adjusted by dilution in deionized water, and a series of DOM solutions with concentrations ranging from 0 to 40 mg C L⁻¹ were used in the experiments. The initial concentration of PHE was set at 1 mg L⁻¹. The sorption experiments were conducted in 22 mL US EPA standard sample vials, and after hand-shaking, the vials were allowed to interact for 10 min at 25 °C before analysis. Experiments were conducted in duplicates. PHE blanks without DOM and DOM blanks without PHE were prepared.

Analysis

Aqueous PHE was quantified by HPLC (Agilent 1200, USA) equipped with a fluorescence detector (G1321A) and a reverse-phase column (Agela, Venusil XBP C-18, 150 mm × 4.6 mm × 5 μm). The flow rate of the mobile phase composed of 90% methanol and 10% HPLC pure water was 1.0 mL min⁻¹, and the injection volume was 10 μL. The excitation and emission wavelengths for the fluorescence detector were 250 nm and 364 nm, respectively.

Concentration of DOM was measured by a TOC analyzer (TOC-5000A, Shimadzu, Japan). The equilibrium DOM concentration after interaction with SCNT for 2 or 20 days was analyzed after removing the SCNT with a 0.45 μm Acrodisc nylon membrane syringe filter. The sorbed amount of DOM onto the SCNT (Q) was calculated based on the difference in DOM concentrations before and after the interaction with SCNT. To characterize the aromaticity of DOM, the absorbances of DOM solutions were measured at 254 nm using a Varian UV-005 spectrophotometer (Varian, USA).

To quantify the interaction of PHE with DOM in solution, the fluorescence intensities of PHE–DOM mixtures and blank samples without DOM were measured at 25 °C using a fluorescence spectrometer (Perkin-Elmer LS 55, UK) with 294 and 365 nm as excitation and emission wavelengths, respectively. The amount of bound PHE could be calculated from the reduction in fluorescence in the PHE–DOM mixtures as compared to the blank, and the K_DOM [L (kg C)⁻¹] of PHE were calculated using the Stern–Volmer plot (eqn (1)) described by Lakowicz:³⁰

where F₀ is the initial fluorescence intensity of PHE blanks without DOM, f is the fluorescence intensity of PHE–DOM mixtures, and [DOM] is the concentration of DOM (mg C L⁻¹). The fluorescence of PHE was assumed to be entirely quenched once it was combined with DOM.

Isotherm modeling

The Freundlich model was employed to analyze sorption isotherm data:

Q = K_FC_eⁿ (2)

where Q (μg kg⁻¹) and C_e (μg L⁻¹) are equilibrium concentrations of PHE on the solid phase and in aqueous solution, respectively; K_F ((μg kg⁻¹)/(μg L⁻¹)ⁿ) is the Freundlich affinity coefficient, and n (unitless) is the Freundlich linearity index.

Results and discussion

Properties characterization of SCNT and DOM-preloaded SCNT

The structural parameters for the original SCNT, and SCNT preloaded with bulk DOM of different contact times, and SCNT preloaded with DOM fractions of different MWs are shown in Table 1. The bulk of the original SCNT was rich in carbon and ash but low in oxygen and hydrogen contents; while the surface of the original SCNT was rich in both carbon and oxygen (Table 1), which was different from the bulk element composition, indicating that the residue O-containing moieties were mainly located on the outer surface of the SCNT. The preloading of extracted bulk DOM on SCNT for 2 days increased the oxygen content from 0.62% to 3.53% and nitrogen content from 0 to 1.21%, respectively, indicating that substantial O-containing and N-containing moieties were introduced to the SCNT by DOM, leading to an increase in polarity, as indicated by the increased polar index [(O + N)/C]. The original SCNT had a BET surface area and micropore volume of 243 m² g⁻¹ and 0.026 cm³ g⁻¹, respectively; while the fresh DOM_bulk–SCNT had a substantially reduced surface area (110 m² g⁻¹) and micropore volume (0 cm³ g⁻¹). The average pore diameter of aged DOM_bulk–SCNT with a contact time of 20 days increased and the BET surface area and total pore volume decreased compared to those of fresh DOM_bulk–SCNT.

Table 1 Selected physical and chemical characteristics of SCNT

Sorbents	S BET (m^2 g^−1)	V total (cm^3 g^−1)	V micro (cm^3 g^−1)	D average (Å)	A cal (m^2 g^−1)	V cal (cm^3 g^−1)	Bulk element composition (%)
							C	H	N	O	Ash	H/C	(O + N)/C
a S BET, BET surface area. b V total, total pore volume. c V micro, micropore volume. d D average, average pore diameter. e A cal and Vcal, the calculated surface area and micropore volume from the surface area of SCNT (243 m^2 g^−1) and micropore volume of SCNT (0.026 cm^3 g^−1) and the adsorbed DOM content by the following equation: Acal = 243 × (1 − fss) and Vcal = 0.026 × (1 − fss), fss = QDOM/100.^9
Original SCNT	243	0.413	0.026	67.9	243	0.026	67.8	2.29	0	0.62	29.3	0.405	0.007
Fresh DOMbulk–SCNT	110	0.467	0	170	160	0.017	65.4	1.81	1.21	3.53	28.0	0.332	0.056
Aged DOMbulk–SCNT	93.1	0.410	0	176	129	0.014	64.7	1.84	1.55	3.67	28.2	0.341	0.063
DOM<1000–SCNT	125	0.494	0	158	214	0.023	67.2	1.47	1.53	5.04	24.7	0.262	0.076
DOM1000–14 000–SCNT	133	0.477	0	144	189	0.020	64.4	1.49	1.11	2.29	30.7	0.278	0.041
DOM>14 000–SCNT	136	0.456	0	134	159	0.017	67.2	1.55	1.65	0.96	28.6	0.277	0.032

---

Surface functionalities and element composition (%)
	C–C	C–O	C=O	O–C=O	C	N	O	O/C	(O + N)/C
Original SCNT	59.4	29.4	5.41	5.80	83.2	0	16.8	0.15	0.15
Fresh DOMbulk–SCNT	35.0	24.6	34.6	5.70	80.2	0.62	19.2	0.18	0.18
Aged DOMbulk–SCNT	54.3	5.99	28.5	11.3	81.3	1.46	17.3	0.16	0.17
DOM<1000–SCNT	47.3	41.0	0.65	11.0	84.7	0.54	14.7	0.13	0.14
DOM1000–14 000–SCNT	53.8	25.4	17.2	3.56	83.9	0.87	15.2	0.14	0.14
DOM>14 000–SCNT	58.7	4.27	20.7	16.3	83.0	1.37	15.6	0.14	0.15

---

The micropore volume was zero for all the DOM-preloaded SCNT. This is because of the diffusion limitation of N₂ molecules to pass through loaded DOM layer into micropores at 77 K, and hence the micropore volume and surface area of the sorbents may be underestimated by this method.³¹ The calculated surface area A_cal and micropore volume V_cal of fresh DOM_bulk–SCNT based on the unit loading mass of DOM were 160.0 m² g⁻¹ and 0.017 cm³ g⁻¹ (Table 1), suggesting that DOM sorption occurred through both surface-adsorption and micropore-blocking processes, which is consistent with previous reports on DOM adsorption by CNT.^20,21 This can also be proved by DFT differential pore volume distribution of the original SCNT, fresh DOM_bulk–SCNT and aged DOM_bulk–SCNT (Fig. 1), showing that after reacting with DOM, the pore volume of SCNT with pore width within 14 Å sharply decreased. The total pore volume of fresh DOM_bulk–SCNT slightly increased from 0.413 to 0.467 cm³ g⁻¹, and average pore diameter substantially increased from 67.9 to 170 Å (Table 1), indicating that DOM-preloading had changed the structural features of the SCNT greatly. After aging, these parameters changed further, indicating the location and conformation of the loaded DOM changed during the increased contact time.

Fig. 1 DFT differential pore volume distribution of original SCNT (◆), fresh DOM_bulk–SCNT (□) and aged DOM_bulk–SCNT (△).

Binding of PHE with DOM fractions with different MWs in solution

The DOM extracted from peat consisted of significantly more DOM_>14 000 (46.9%) and relatively less of DOM_<1000 (35.2%) and DOM_{1000–14 000} (18.0%), as calculated based on the TOC values of each respective fraction. Notable differences in the PHE binding capacity existed for different DOM MW fractions; however, K_DOM (eqn (1)) values of PHE were still in the same magnitude order, being 3.16 × 10⁴, 2.79 × 10⁴, and 2.19 × 10⁴ L (kg C)⁻¹ for DOM_>14 000, DOM_{1000–14 000}, and DOM_<1000 (Table 2), which descended with decreasing MW of the DOM. The sum of the K_DOMs of the different DOM MW fractions based on their percentage weights was 2.76 × 10⁴ L (kg C)⁻¹, which agrees closely with the measured K_DOM for DOM_bulk [2.45 × 10⁴ L (kg C)⁻¹]. Several interactions may control the binding of PHE to DOM, such as van der Waals forces, solubilization (partition) effect, specific forces like π–π and n–π electron donor–acceptor (EDA) interactions, as well as physical engagement in the steric structure of DOM aggregates.^32,33 Hence, both chemical composition and steric conformation of DOM influence the binding of organic chemicals. The UV light absorbance of DOM depends on the electronic structure of its macromolecules. The UV absorbance in the range of 254–280 nm is in the region of π–π* electronic transitions in conjugated systems of bulky DOM and can be used as an indicator of aromaticity.^34,35 The UV absorbance at 254 nm followed the order: DOM_>14 000 > DOM_1000–14000 > DOM_bulk > DOM_<1000 (Table 2), and significant positive correlation between the molar absorption of different DOM MW fractions at 254 nm and their K_DOMs for PHE occurred (r² = 0.94), which agrees with the view that the aromaticity of DOM is a primary factor in controlling its binding capacity to hydrophobic aromatic hydrocarbons. This finding is consistent with former reports,^36–38 but contrary to a study on pyrethroid, which indicated that aromaticity was not the main property influencing pyrethroid binding.³⁹ The different results in the latter study may be because of the specific reactions between DOM and polar groups in pyrethroid.

Table 2 The parameters of the Stern–Volmer plot for PHE combined to different DOM MW fractions and their UV light absorbance at 254 nm^a

DOM fraction	UV254nm (L (m mg)^−1)	Stern–Volmer plot	R ^2	K DOM [L (kg C)^−1]	Log KDOM
a x is the concentration of DOM (mg C L^−1) and Y is F0/f (eqn (1)).
DOMbulk	345	Y = 0.0245x + 0.99	0.995	2.45 × 10^4	4.39
DOM<1000	281	Y = 0.0219x + 0.99	0.989	2.19 × 10^4	4.34
DOM1000–14 000	358	Y = 0.0279x + 0.98	0.987	2.79 × 10^4	4.44
DOM>14 000	392	Y = 0.0316x + 0.99	0.994	3.16 × 10^4	4.50

---

Sorption of PHE on original SCNT and fresh DOM_bulk–SCNT

Sorption isotherms of PHE on original SCNT and fresh DOM_bulk–SCNT are shown in Fig. 2. Parameters of the sorption of DOM on original SCNT and PHE on unmodified and modified SCNT are listed in Table 3.

Fig. 2 Sorption of PHE on original SCNT (▲), fresh DOM_bulk–SCNT (□) and aged DOM_bulk–SCNT (◇).

Table 3 Sorption parameters of DOM on SCNT and PHE on different kinds of SCNT

Sorbents	DOM			PHE (the Freundlich equation (eqn (2)) was employed)
	C 0 (mg C L^−1)	C e (mg C L^−1)	Q (mg C g^−1)	Log KF	n	R ^2	K d (L kg^−1)
							C e = 100 μg L^−1	C e = 1000 μg L^−1
a C 0 (mg C L^−1) is the initial concentration of DOM before reacting with SCNT. b C e (mg C L^−1) is the equilibrium concentration of DOM in the aqueous solution. c Q (mg C g^−1) is the equilibrium concentration of DOM on SCNT. d K d (L kg^−1) values were derived from the Freundlich model: Kd = Q/Ce = KF(Ce)^n−1.
Original SCNT	—	—	—	7.34	0.236	0.966	6.49 × 10^5	1.12 × 10^5
Fresh DOMbulk–SCNT	18.4	14.2	34.2	6.75	0.402	0.959	3.54 × 10^5	8.92 × 10^4
Aged DOMbulk–SCNT	18.4	12.5	46.9	7.14	0.283	0.946	5.06 × 10^5	9.72 × 10^4
DOM<1000–SCNT	12.0	10.5	11.8	7.22	0.273	0.952	5.87 × 10^5	1.10 × 10^5
DOM1000–14 000–SCNT	12.1	9.29	22.2	7.05	0.343	0.924	5.41 × 10^5	1.17 × 10^5
DOM>14 000–SCNT	12.1	7.82	34.6	4.57	1.093	0.963	5.69 × 10^4	7.04 × 10^4

---

The original SCNT showed great sorption capacity for PHE, with log K_F being 7.34 (Table 3). The sorption was highly nonlinear, with the Freundlich nonlinear index n being 0.23, suggesting heterogeneous sorption sites existing on the SCNT. The heterogeneous sites of SCNT for PHE may come from the external surfaces and interstitial channels trapped in the SCNT. Besides, aggregation of SCNT may create heterogeneous sites with different sorption energies for PHE sorption.⁴⁰ From the thermodynamic view, at low PHE concentration, high-energy sites played a dominant role in adsorption, hence, K_d at C_e = 100 μg L⁻¹ was much higher than that at C_e = 1000 μg L⁻¹.

After loading of DOM, the Freundlich nonlinear index n of PHE adsorption on fresh DOM_bulk–SCNT increased, suggesting that DOM molecules were preferably sorbed to high-energy adsorption sites of SCNT making sorption sites of SCNT less heterogeneous for PHE. Yang et al. also showed that cetylpyridinium chloride precoating on SCNT increased the linearity of the naphthalene adsorption isotherm due to the occupancy of high-energy adsorption sites.⁹

When preloaded with DOM, PHE sorption on SCNT declined significantly, and log K_F decreased from 7.34 to 6.75 for the fresh DOM-preloaded SCNT (Table 3). Several mechanisms may be involved for the reduction. First, DOM preloading introduced polar moieties to SCNT, as indicated by the increased O content (Table 1). A slight increasing in O-containing moieties would greatly facilitate the formation of water clusters on the surface of SCNT. The water clusters and polar functionalities on the SCNT surface would affect the surface hydrophobicity of SCNT, and thus inhibit the hydrophobic interactions between SCNT and PHE. Zhang et al. found that surface functionalized SCNT showed reduction in adsorption of PHE but enhancement in adsorption of biphenyl and 2-phenylphenol due to the increased O-containing functional group content on the surface of SCNT.³⁹ The latter is due to the strengthened polar interactions. Secondly, as DOM sorption occurred through both surface-adsorption and micropore-blocking processes, DOM molecules would occupy sorption sites of SCNT and block the micropores, which suppressed the sorption of PHE by “competition”. This could be confirmed by the reduced micropore volume, especially the portion with micro-size (<14 Å) and surface area (Table 1 and Fig. 1). Besides the change in individual SCNT particles, DOM-preloading might reduce the aggregation of CNT particles and enhance their repulsion. Enhanced dispersion of SCNT by DOM was visually observed in our DOM-preloading experiments. This may have two effects. First, as mentioned above, the interstitial spacing between neighboring nanotubes may sequester PHE and contribute partially to PHE sorption. Reduction in SCNT aggregation led to a loss of this sorption mechanism. On the other hand, reduction in aggregation resulted in more surface sites being exposed,^18,41 which should have favored the sorption of PHE. Hence, the apparent impact on PHE sorption on SCNT by DOM preloading is an integrated result of the above multiple processes. In the present study, the increase in sorption of PHE due to the newly exposed sites was much weaker than the decrease in sorption. Our observation is consistent with the previous findings that coating DOM on black char strongly reduced its affinity for hydrophobic organic molecules.⁴² Kilduff and Wigton also reported that sorption of trichloroethylene by humic-preloaded activated carbon was inhibited.¹²

Sorption of PHE on aged DOM_bulk–SCNT

Although sorption kinetics have shown that apparent sorption equilibrium occurred in less than 2 days, increasing the contact time made SCNT sorb more DOM molecules, as indicated by the increased Q values of aged DOM_bulk–SCNT compared to that for fresh DOM_bulk–SCNT (Table 3). According to this, we have expected that the sorption capacity and nonlinearity of PHE on aged DOM–SCNT should have decreased. However, PHE sorption on the aged DOM–SCNT increased significantly as compared to that on the fresh DOM–SCNT (Fig. 2), with the log K_F value increasing from 6.75 to 7.14. The nonlinearity of the sorption isotherm increased simultaneously, with n decreasing from 0.402 to 0.283 (Table 3). To explain this, the properties were compared between the aged and fresh DOM_bulk–SCNT. Micropore volume (Fig. 1, pore width within 12 Å) in aged DOM_bulk–SCNT, which is thought to contribute primarily to the sorption of organic chemicals like PHE (169.5 ų), increased slightly compared to that of the fresh DOM_bulk–SCNT (Fig. 1). Bulk O content of fresh and aged DOM_bulk–SCNT was 3.53 and 3.67, respectively; whereas the corresponding surface O content was 19.2 and 17.3 (Table 1). These different variation tendencies of O content in bulk and surface of DOM-preloaded SCNT suggested that DOM molecules that contained O atoms had transferred from the external surface to interstitial channels in the SCNT aggregate, making the sorption sites restore their heterogeneity and become more available for PHE molecules.

Sorption of PHE on SCNT preloaded with DOM fractions with different MWs

The interaction of SCNT and DOM fractions of different MWs differed a lot (Table 3), with the sorption amount following the order: DOM_>14 000 (Q = 34.6 mg C g⁻¹) > DOM_{1000–14 000} (Q = 22.2 mg C g⁻¹) > DOM_<1000 (Q = 11.8 mg C g⁻¹). Though the most was loaded onto SCNT for DOM_>14 000, the O content of DOM_>14 000–SCNT was the least (0.96%) compared to those of DOM_{1000–14 000}–SCNT (2.29%) and DOM_<1000–SCNT (5.04%) (Table 1). This is due to the differences in the chemical composition of the DOM fractions with different MWs. Our parallel study found that C content increased and O content decreased with increasing DOM MW, indicating that DOM_>14 000 was less hydrophilic followed by DOM_{1000–14 000} and DOM_<1000.⁴³ The difference in the polarity of DOM fractions could explain the difference in their affinity to SCNT. Thus, the preloading of DOM_>14 000 made SCNT more hydrophobic than the other DOM fractions did, as indicated by the increase in polar index [(O + N)/C] from 0.032 for DOM_>14 000–SCNT to 0.076 for DOM_<1000–SCNT. As has been discussed above, the absorbance of the different DOM fractions at 254 nm followed the order: DOM_>14 000 > DOM_{1000–14 000} > DOM_<1000 (Table 2), suggesting that DOM_>14 000 was more aromatic than DOM_{1000–14 000} and DOM_<1000. As π–π interactions exist between SCNT and DOM, and this could also explain why SCNT had a stronger interaction with DOM_>14 000 than with DOM_{1000–14 000} and DOM_<1000. Hence, we attribute the strongest sorption of DOM_>14 000 on SCNT to hydrophobic and π–π interactions. This result is consistent with the previous reports showing that hydrophobic and π–π interactions were responsible for DOM sorption by activated carbon and multi-walled carbon nanotubes.^44,45 We also observed that the loading of DOM_>14 000 more significantly reduced the aggregation degree of SCNT particles than DOM_<1000 and DOM_{1000–14 000} did, as indicated by the color of the SCNT suspensions after standing. This may lead to an enhancement of repulsion and increasing the number of sorption sites on the SCNT. Therefore, the sorption of PHE on SCNT preloaded with different DOM MW fractions was the result of the balance of these two opposite effects.

The loading of all the three DOM fractions with different MWs inhibited PHE sorption on SCNT; however, there were notable differences in the sorption of PHE on SCNT loaded with DOM fractions with different MWs (Fig. 3, Table 3). DOM_>14 000–SCNT showed the weakest sorption capacity, with log K_F decreasing from 7.34 for the original SCNT to 4.57; whereas the log K_F for DOM_{1000–14 000}–SCNT and DOM_<1000–SCNT was 7.05 and 7.22, respectively, which did not differ a lot from the value for the original SCNT. Combining the changes in SCNT properties, the most negative effect of DOM_>14 000 on PHE sorption was ascribed to the occupying of sorption sites and blockage of micropores of SCNT by DOM_>14 000 due to its greater sorbed amount and larger MW. The log K_F of DOM_>14 000–SCNT (4.57 in Table 3) is similar to the log K_DOM of DOM_>14 000 (4.50 in Table 2), suggesting that the SCNT was completely occupied by DOM_>14 000, and the sorption sites of SCNT were inaccessible to PHE. The Freundlich linearity sorption index n of PHE on DOM_>14 000–SCNT is 1.093, indicating a linear sorption. This confirms that the highly heterogeneous sorption sites of SCNT with high energy were almost completely covered by the DOM_>14 000 molecules. Though the loading of DOM_{1000–14 000} and DOM_<1000 makes SCNT more polar, which can partially explain the reduction in PHE sorption: most of the sorption sites of SCNT are still accessible to PHE due to the lesser sorbed amount and smaller MWs of these two fractions. Hence, the reduction in PHE sorption is very limited.

Fig. 3 Effect of DOM fractions of different molecular weights on the sorption of PHE as compared to bulk DOM. Fresh DOM_bulk–SCNT (◆); DOM_<1000–SCNT (△); DOM_{1000–14 000}–SCNT (□); DOM_>14 000–SCNT (○).

There was no significant correlation between the sorption capacity of PHE on different SCNT and their structural parameters, such as surface area, micropore volume, surface and bulk polarity among DOM-preloaded SCNT (data not shown). This confirms that PHE sorption by SCNT was determined by multiple factors and no single SCNT characteristic controls PHE sorption. In addition, the surface area and micropore volume measured from nitrogen sorption–desorption isotherms at 77 K may be underestimated. Yang et al. reported that the adsorbed volume capacity of PHE on SCNT was greater than both the monolayer adsorption volume capacity of N₂ and micropore volume of SCNT.⁴ Pignatello et al. showed that N₂ molecules were unable to access micropores after condensation of humic substances on black carbon surface.¹³

Conclusions

DOM sorption on SCNT occurred through both surface adsorption and micropore blocking, which reduced the sorption of PHE on DOM-preloaded SCNT by occupying sorption sites and blockage of micropores. Both the chemical composition and steric conformation of DOM were found to influence its interaction with SCNT. The DOM fraction with larger MW (DOM_>14 000) showed stronger interaction with SCNT and a greater binding capacity of PHE than smaller ones (DOM_<1000 and DOM_{1000–14 000}), leading to the strongest suppression of sorption of PHE on DOM-preloaded SCNT. Increasing the contact time between DOM and SCNT from 2 days to 20 days enhanced the sorption capacity and increased the sorption nonlinearity of PHE on DOM-preloaded SCNT due to the transferring of DOM molecules from the external surface to interstitial channels trapped in SCNT, making the sorption sites restore their heterogeneity and be more available for PHE molecules. In addition, sorption of PHE by SCNT was determined by multiple factors and no single SCNT characteristic controls PHE sorption.

Acknowledgements

This study was supported by Natural Science Foundation of China (no. 41073087 and 41225014) and by the Fundamental Research Funds for the Central Universities in China.

Notes and references

1. S. Iijima, Nature, 1991, 354, 56–58.
2. R. H. Baughman, A. A. Zakhidov and W. A. de Heer, Science, 2002, 297, 787–792.
3. T. Masciangioli and W. X. Zhang, Environ. Sci. Technol., 2003, 37, 102–108.
4. K. Yang, L. Z. Zhu and B. S. Xing, Environ. Sci. Technol., 2006, 40, 1855–1861.
5. S. Gotovac, J. Honda, Y. Hattori, K. Takahashi, H. Kanoh and K. Kaneko, Nano Lett., 2007, 7, 583–587.
6. E. J. Petersen, R. A. Pinto and P. F. Landrum, Environ. Sci. Technol., 2009, 43, 4181–4187.
7. E. J. Petersen, Q. G. Huang and W. J. Weber, Environ. Health Perspect., 2008, 116, 496–500.
8. B. Pan and B. S. Xing, Environ. Sci. Technol., 2008, 42, 9005–9013.
9. K. Yang, Q. Jing and W. Wu, Environ. Sci. Technol., 2010, 44, 681–687.
10. E. A. Ghabbour and G. Davies, Humic Substances: Versatile Components of Plants, Soil and Water, Royal Society of Chemistry, Cambridge, UK, 2000.
11. K. Kalbitz, S. Solinger, J. H. Park, B. Michalzik and E. Matzner, Soil Sci., 2000, 165, 277–304.
12. J. E. Kilduff and A. Wigton, Environ. Sci. Technol., 1999, 33, 250–256.
13. J. J. Pignatello, S. Kwon and Y. F. Lu, Environ. Sci. Technol., 2006, 40, 7757–7763.
14. E. M. Murphy, J. M. Zachara, S. C. Smith and J. L. Philips, Sci. Total Environ., 1992, 117–118, 413–423.
15. K. H. Kang, H. S. Shin and H. Park, Water Res., 2002, 36, 4023–4032.
16. Y. P. Chin, G. R. Aiken and K. M. Danielsen, Environ. Sci. Technol., 1997, 31, 1630–1635.
17. G. Chen, C. Lin and L. Chen, Chemosphere, 2010, 79, 1046–1055.
18. H. Hyung, J. D. Fortner, J. B. Hughes and J. H. Kim, Environ. Sci. Technol., 2007, 41, 179–184.
19. H. Hyung and J. H. Kim, Environ. Sci. Technol., 2008, 42, 4416–4421.
20. S. J. Zhang, T. Shao, S. Sule and B. Kaplan, Water Res., 2010, 44, 2067–2074.
21. X. L. Wang, J. L. Lu and B. S. Xing, Environ. Sci. Technol., 2008, 42, 3207–3212.
22. M. Alexander, Environ. Sci. Technol., 2000, 34, 4259–4264.
23. H. W. Sun, T. Masafumi, I. Michihiko and F. Masanori, Water Res., 2003, 37, 2960–2968.
24. J. G. Li, H. W. Sun and Y. Zhang, Soil Sediment Contam., 2007, 16, 1–9.
25. H. W. Sun and J. G. Li, Water, Air, Soil Pollut., 2005, 166, 353–365.
26. W. Zhang and H. W. Sun, Soil Sediment Contam., 2012, 21, 407–418.
27. K. Y. Cheng and J. W. C. Wong, Chemosphere, 2006, 62, 1907–1916.
28. P. I. Ravikovitch, G. L. Haller and A. V. Neimark, Adv. Colloid Interface Sci., 1998, 76–77, 203–226.
29. C. H. Cheng, J. Lehmann, J. E. Thies, S. D. Burton and M. H. Engelhard, Org. Geochem., 2006, 37, 1477–1488.
30. J. R. Lakowicz, Principles of Fluorescence Spectroscopy, Stern–Volmer, Plenum Press, New York, 2006.
31. W. J. Braida, J. J. Pignatello, Y. Lu, P. I. Ravikovitch, A. V. Neimark and B. S. Xing, Environ. Sci. Technol., 2003, 37, 409–417.
32. T. D. Gauthier, W. R. Seitz and C. L. Grant, Environ. Sci. Technol., 1987, 21, 243–248.
33. D. Zhu, S. Hyun and J. J. Pignatello, Environ. Sci. Technol., 2004, 38, 4361–4368.
34. D. J. Dryer, G. V. Korshin and M. Fabbricino, Environ. Sci. Technol., 2008, 42, 6644–6649.
35. J. L. Weishaar, G. R. Aiken and B. A. Bergamaschi, Environ. Sci. Technol., 2003, 37, 4702–4708.
36. T. Polubesova, N. M. Sherman and B. Chefetz, Environ. Sci. Technol., 2007, 41, 5389–5394.
37. S. Kang and B. S. Xing, Environ. Sci. Technol., 2005, 39, 134–140.
38. G. Chen, C. Lin and L. Chen, Chemosphere, 2010, 79, 1046–1055.
39. M. L. Delgado, L. Wu and J. Gan, Environ. Sci. Technol., 2010, 44, 8473–8478.
40. S. J. Zhang, T. Shao, S. S. K. Bekarogle and T. Karanfil, Environ. Sci. Technol., 2009, 43, 5719–5725.
41. B. Pan and B. S. Xing, Environ. Sci. Technol., 2008, 42, 9005–9013.
42. S. Kwon and J. J. Pignatello, Environ. Sci. Technol., 2005, 39, 7932–7939.
43. J. Z. Wu, H. W. Sun, C. P. Wang and Y. H. Li, Chem. Res. Chin. Univ., 2012, 28, 624–630.
44. J. C. Crittenden, S. Sanongraj, J. L. Bulloch, D. W. Hand, T. N. Rogers, T. F. Speth and M. Ulmer, Environ. Sci. Technol., 1999, 33, 2926–2933.
45. X. L. Wang, S. Tao and B. S. Xing, Environ. Sci. Technol., 2009, 43, 6214–6219.

---