Determination of several fullerenes in sewage water by LC HR-MS using atmospheric pressure photoionisation

Abstract

The main challenge in the mass spectrometric analysis of fullerenes in complex matrices is to prove unambiguously their presence or absence. Usually, this can be done by fragmentation, but due to the atmospheric interface and solvents commonly used, complex adduct formation will hinder quantification and qualification by distorting the isotopic cluster. In the present study, contrary to other methods in the literature, a 100% toluene isocratic mobile phase is used, employing an atmospheric pressure photoionisation (APPI) interface. In this manner, the adducts that are formed when methanol is present in the system are almost completely eliminated, thus avoiding higher than expected abundance for the isotopic cluster ions. The expected relative abundance of the isotopic clusters correlates well with the ones observed and was shown to be an effective tool for qualifying the presence of fullerenes even at low concentrations. For the analysis of suspended matter for the presence of fullerenes in sewage water, an optimised procedure was proposed. Filters with different mesh sizes (8.0, 0.7 and 0.45 μm) were used and the filtrate was analysed by means of solid phase extraction. For a 100 mL influent wastewater sample, limits of quantitation for the functionalised fullerenes were between 0.22 and 0.28 ng L⁻¹ and for the pristine fullerenes between 0.7 and 1.4 ng L⁻¹. The method used to analyse sewage water from six different wastewater treatment plants proved to be highly sensitive, able to unambiguously identify and quantify C₆₀ in sewage water.

---

Nano impact

Detection and characterisation of man-made nanoparticles is an important task. Therefore, analytical techniques that can determine these particles in complex environmental samples are urgently required. Up to now, analytical techniques have been struggling with matrices containing high amounts of dissolved organic matter. High resolution mass spectrometry for the detection of organic fullerenes can be a powerful tool to overcome this problem, not only to detect these particles but also to quantify them at low concentrations (ng L⁻¹). The application of an atmospheric pressure photoionisation device reduces the formation of complicated set of adducts. This enables an easy identification and quantification of several fullerenes. Due to the reduced complexity of adducts, the isotope ratio for ¹³C/¹²C is recommended to be used for identification.

Introduction

Since the discovery of C₆₀-fullerene by Kroto et al.,¹ as a very stable carbon allotrope, different analytical methods have been developed to determine the presence of fullerenes in the environment. The pristine fullerenes, without any functional group, are naturally present in the environment. These carbon-based nanoparticles have been detected, for example, in 10 000 year old ice cores.² Nowadays, the most prominent sources of fullerenes are incidental emissions by different combustion processes.^3,4 However, nanotechnology appears as a new potential source of fullerenes in the environment. Nowadays, fullerenes are applied in different materials and consumer products such as personal care products,⁵ drug delivery systems⁶ and in particular derivatised fullerenes – which are exclusively of anthropogenic origin – in solar⁷ and photovoltaic⁸ cells. In addition, their use and production is expected to increase substantially in the next coming years.⁹ Therefore, it can be expected that an ever increasing amount will leach into the environment through wastewater discharge, landfills, factories, storm water run-off, erosion from composites and deposition from the atmosphere.^10–13

However, up until now, a limited number of studies have determined the presence of fullerenes and in particular functionalised fullerenes in complex environmental samples. Farré et al. detected concentrations up to the μg L⁻¹ of C₆₀, C₇₀ and C₆₀-methylpyrrolidine in the suspended organic matter of wastewater treatment plants,¹⁴ and Chen and Ding reported concentrations of 98 and 130 ng L⁻¹ C₆₀ in surface waters and industrial effluent in Taiwan, respectively.¹⁵ Predicted environmental concentrations of fullerenes span from 0.003 ng L⁻¹ for European surface waters to 4 ng L⁻¹ for sewage treatment effluents.¹⁶ During a campaign in 2008, grab samples were taken from Dutch surface waters and no fullerenes were detected above a concentration of 5 ng L⁻¹ (LOQ).¹⁷

Due to their potential environmental and toxicological risks, the properties and behaviour of fullerenes have been thoroughly studied.^18–21 Because of their low water solubility (1.3 × 10⁻¹¹ for C₆₀ and 1.3 × 10⁻¹⁰ for C₇₀ ng mL⁻¹),²² in the aquatic media, one of their main characteristics is the tendency to form (meta)stable colloidal aggregates (nC₆₀).^17,18,21 These aggregates are responsible for their high apparent solubility in aqueous solution media which dwarfs the solubility of the single fullerene molecules.²² It should be pointed out that nC₆₀ aggregates thermodynamically have a metastable character in water solution depending on different environmental factors, such as salinity or organic material. Therefore, environmental changes can revert these aggregates to single molecules.¹⁹

The environmental behaviour of these analytes should be considered when designing sample preparation and extraction strategies, e.g. organic solvent extraction with toluene has been proposed to extract fullerenes from soils.^23,24 Recently, asymmetric flow field flow fractionation coupled to a mass spectrometer (MS)²⁵ has been reported for the direct analysis of environmental samples. This method allows the detection of fullerenes in different size fractions. However, so far the most common approach to extract fullerenes from aqueous matrices is solid phase extraction (SPE).^8,14,17,24

The separation of fullerenes is commonly achieved by high performance liquid chromatography (HPLC) using a C18 analytical column with toluene and methanol or acetonitrile as mobile phases.^14,17,26 In spite of that, the use of more specific stationary phases, e.g. with a pyrenyl-propyl group, has not been reported extensively.^8,23,27 The main disadvantage of using acetonitrile or methanol is that a complex set of by-products (adducts) is formed during the atmospheric pressure ionisation in the negative mode; mainly with electrospray but also to some extent with photoionisation,^8,17,28 the isotopic abundance pattern, mainly ¹³C, is distorted due to protonation.²⁸ Since pristine fullerenes hardly fragment, identification based solely on the presence of the accurate mass and retention time is not sufficient. For confirmation, the ratio between the ¹²C and ¹³C isotopes can be an alternative option, but ¹³C needs to be free from interferences (adducts).

Kolkman et al. have reported an analytical method for the detection of C₆₀, C₇₀ and six functionalised fullerenes in waters at environmentally relevant concentrations (LOD = 0.3 ng L⁻¹ and LOQ = 1 ng L⁻¹). This analytical method was based on the use of HPLC combined with an Orbitrap MS using electrospray ionisation (ESI) in negative mode. The main advantage of this method is the possibility of characterising other (unknown) fullerenes due to the application of full-scan high-resolution (HR) MS.

In the present work, we present a new analytical approach for the determination of various fullerenes in sewage water influents. The present method is based on the previous method developed by Kolkman et al.⁸ but with enhanced sensitivity and selectivity (10 times less sample volume is needed). The present approach makes use of a pyrenyl-propyl-based analytical column, with toluene as the mobile phase under isocratic conditions in combination with an APPI interface. This combination reduces the adduct formation and ion suppression. In addition, toluene acts as a dopant in the APPI process ion source enhancing the sensitivity.

Materials and methods

Chemicals and reagents

Standard compounds were purchased as shown in Table 1.

Table 1 Standard compounds

Name	Abbreviation used	Empirical formula	Standard purity	Supplier reference
Supplier and reference: A—Sigma-Aldrich (Steinheim, Germany), B—Solenne B.V. (Groningen, The Netherlands).
C60 fullerene	[60]	C60	99.90%	A: 572500
C70 fullerene	[70]	C70	99%	A: 482994
[6,6]-Phenyl-C61-butyric acid octyl ester	[60]PCBO	C79H28O2	≥99%	A: 684481
[6,6]-Phenyl-C61-butyric acid butyl ester	[60]PCBB	C75H20O2	>97%	B
[6,6]-Phenyl-C61-butyric acid methyl ester	[60]PCBM	C72H14O2	>99.5%	B
[6,6]-Bis-phenyl-C61-butyric acid methyl ester	Bis[60]PCBM	C84H28O4	99.50%	B
[6,6]-Phenyl-C71-butyric acid methyl ester	[70]PCBM	C82H14O2	99%	B
[6,6]-Thienyl-C61-butyric acid methyl ester	[60]ThCBM	C70H12O2S	>99%	B

---

Toluene (Ultra residue analysed), methanol and acetone (HPLC analysed) were supplied by Mallinckrodt Baker B.V. (Deventer, The Netherlands). Petroleum ether (puriss., boiling point 40–60 °C) was supplied by Sigma-Aldrich (Steinheim, Germany), and cellulose nitrate filters of 47 mm diameter with pore sizes of 8.0 μm (ref. 11301-50-AHN) and 0.45 μm (ref. 11306-50-ACN) were purchased from Sartorius Stedim Biotech (Göttingen, Germany). Glass fibre filters of 0.7 μm pore size were obtained from Whatman (Maidstone, Kent, UK). Solid phase extraction (SPE) cartridges with reversed phase end-capped octadecylsilane (C₁₈) material, bonded to silica gel (40 μm, 60 Å), were supplied by Mallinckrodt Baker B.V.

Sample collection

Influent (collected after the sand-trap) and effluent 24 h flow-dependent composite wastewater (WW) samples were obtained from six different locations. All samples were stored in high density polyethylene containers and immediately after collection kept at 4 °C until analysis.

Preparation of fullerene suspensions

Stock toluene suspensions of individual fullerenes were prepared at ~500 mg L⁻¹ and shaken overnight. A fullerene mix stock suspension was prepared by diluting the previous stock solutions to 40 mg L⁻¹ in toluene. Finally, the external calibration standard curve and the toluene spiking suspensions employed for method development were prepared from the fullerene mix stock suspension. A stock water dispersion of the fullerene mixture was prepared by solvent transfer method, following the procedure which is fully described in Kolkman et al. (2013).⁸ Briefly, a toluene stock solution was slowly dispersed under controlled temperature conditions in water under continuous ultrasonic sonication until the dispersion was clear and yellow, and finally the fullerene suspensions were filtered. The final concentration was confirmed by LLE⁸ (n = 7) and ranged between 80 and 140 μg l⁻¹ in water with a relative standard deviation of 2%.

Sample handling

Sample preparation and extraction was based on the method developed by Farré et al.¹⁴ with some modifications. Briefly, wastewater samples were filtered through a 0.45 μm filter, dried and extracted three times with 20 mL of toluene. To optimise the filtration and extraction of fullerenes from the particulate matter, different experiments were performed: (1) the amount of toluene to extract fullerenes from spiked filtered particulate matter is varied. 50 mL of blank wastewater were consecutively filtered through 8.0 μm, 0.7 μm and, finally, 0.45 μm filters, dried (3 h, 50 °C) and spiked with 100 ng of fullerene stock solution in toluene, dried and extracted by ultrasonic assisted extraction for 45 min by single extraction (10, 20 and 30 ml) and also through a double extraction with the same volumes, all in triplicate. (2) The extraction time needed with toluene under sonication is varied. Filters with particulate matter were spiked with 100 ng of fullerene stock solution in toluene, dried, and extracted with 10 mL of toluene at different times (30 to 240 min). (3) The distribution of the fullerenes associated with the particulate matter of sewage water was investigated. 50 mL of wastewater were spiked with the aqueous stock water suspension and were filtered consecutively through three filters (8.0 μm, 0.7 μm and, finally, 0.45 μm mesh filters), and the particulate matter fraction of each filter was analysed separately. The filtered aqueous phase was analysed with SPE.⁸ The experiment was carried out in triplicate at a concentration of 2 ng L⁻¹. The filtration of the wastewater samples was carried out using a custom-made stainless steel filtration unit in order to avoid adsorption of fullerenes to the glass frit. A 50 ml ultrapure water sample was spiked with the same volume of water dispersion and directly extracted via LLE and set to 100% recovery.

The optimised method is based on the combined analysis of the wastewater particulate matter and its aqueous phase. 100 mL of influent wastewater is filtered through 8.0 and 0.45 μm mesh size cellulose nitrate filters in that order. Each filter is placed in a 100 mL Erlenmeyer flask and dried at 50 °C overnight. 10 mL of toluene is added to each Erlenmeyer flask, fully covering the filter. Next the filters are extracted in an ultrasound bath for a period of 90 min. The toluene extracts of the particulate matter are combined with the extract of the aqueous phase (extraction according to Kolkman et al.⁸) in a glass tube. The mixture is evaporated under a heated nitrogen flow, by using an automated blow-down apparatus (Barkey, Germany) until ~0.5 mL of solvent is left. Finally, extracts are transferred to a vial and the volume is adjusted to 1000 μl by weight. For all optimization experiments with sewage water, influent grab wastewater samples from Utrecht WW treatment plant (Utrecht, The Netherlands) were used as field blanks. They were analysed and none of the targeted fullerenes were detected. For the procedural blanks, ultrapure water was used instead of sewage water. When the stock water dispersion was used, the same spiked amount was extracted by LLE and the concentration was set to 100%.

Safety precautions

The HPLC system setup was placed in a custom-built fume hood and the exhaust from the atmospheric pressure interface was directly vented into the exhaust system to prevent toluene vapours from entering the laboratory atmosphere.

High-resolution mass spectrometry measurements

A hybrid LTQ Orbitrap mass spectrometer (Thermo Electron, Bremen, Germany) was interfaced with a Surveyor HPLC system (Thermo Electron, Bremen, Germany) for chromatographic separation. The initial method used a HPLC gradient with toluene and acetonitrile with post column infusion of methanol⁸ and an ESI. The ESI interface was replaced by an Ion Max API source housing (Thermo Fisher Scientific) containing an APCI probe (used as a nebuliser/desolvation device) and a SyagenPhotoMate VUV light source (krypton discharge lamp, 10.0 eV) (Syagen Technology Inc., Tustin, CA, USA) forming the atmospheric pressure photoionisation interface (APPI) with a 100% isocratic gradient of toluene (20 min). 25 μL of extract was injected using the chromatographic system. Separation was achieved using a custom-made Cosmosil® Buckyprep column (250 mm length, 2.0 mm internal diameter, 5 μm particles) from Nacalai Tesque, Japan, with 100% of toluene as the mobile phase at 0.200 mL min⁻¹ (toluene also acts as the dopant). For comparison experiments, methanol was infused post-column.

The APPI source parameters were optimised by infusing a solution of C₆₀ in toluene and further fine-tuned with sequential injections of the mixture stock with the complete set of fullerenes. The vaporiser temperature was set at 350 °C. Sheath gas, auxiliary gas and swept gas were set at 50 arbitrary units (a.u.), 20 a.u. and 2 a.u., respectively. The tube lens was set to 200 V and the capillary temperature and voltage were maintained at 350 °C and −7.5 V, respectively; no discharge voltage was applied. Full-scan high accuracy mass profile spectra were acquired in the range of 100–1300 m/z with the resolution set at 30 000 (FWHM m/z 400). For confirmation purposes, accurate MS/MS spectra were recorded at a resolution of 7500 at normalised collision energy (NCE) of 40.

Quality assurance

In order to avoid sample contamination, all materials, in contact with fullerenes, were cleaned and rinsed (three times) with acetone, petroleum ether and finally toluene prior to use. Since fullerenes are prone to radical-oxidation in contact with UV light,²⁹ light exposure was avoided by using amber glassware or either covering with aluminium foil. All quantification was done by measuring external calibration standards between 10 and 500 ng L⁻¹ of all fullerenes in toluene. Instrumental interday and intraday repeatability were calculated at three concentration levels as relative standard deviation, which are obtained when analysing 5.0 μL of 2.0, 16 and 128 pg μL⁻¹ toluene standards. Instrumental limits of detection and quantification (ILOD and ILOQ, respectively) were defined as the concentration that provides signal-to-noise (S/N) values of 3 and 10 for the quantifier ion of each analyte and they were experimentally assessed by consecutive dilution of the same standard.

The validation of the method was carried out by analysing spiked wastewater. The method's limits of detection and quantification, recovery yields, linearity range and interday precision were assessed as follows: (1) recovery rates and interday repeatability were calculated at 10, 50 and 500 ng L⁻¹; (2) the limits of detection (MLODs) and quantification (MLOQs) for fullerenes were calculated for the quantifier ion of each analyte as 3 and 10 times its signal-to-noise (S/N) value.

Results and discussion

Extraction optimisation

The most common way of determining the amount of fullerenes associated with the particulate matter is to filter the sample. Since raw sewage is a highly complex matrix with high amounts of suspended matter, a filtration directly through 0.45 μm will result in clogging and therefore a prefilter (>0.45 μm) is needed. This is the reason why several different sized filters were used (8.0, 0.7, and 0.45 μm) to optimise the extraction. Starting point was a method developed by Farré et al. (2010). In addition to the optimisation, a larger number of fullerenes were included.

The main goal was to optimise the extraction of the set of (functionalised) fullerenes and to aim at a reduction of the amount of toluene to create a one-step extraction. First of all, the amount of toluene to extract fullerenes from spiked filtered particulate matter was varied. In order to obtain spiked samples of particulate matter, three aliquots (100 mL each) of a blank influent were passed over the three filters. The particulate matter collected on each of the filters was then spiked with 50 μL of fullerene mixture stock suspension at 2 ng μL⁻¹, dried and extracted with toluene in an ultrasonic bath. The extracts were evaporated and transferred to a vial. All experiments were performed in triplicate and the results are summarised in Fig. 1. As can be seen, 10 mL of toluene were enough for obtaining a good recovery in the case of 0.7 μm glass fiber filters (≥80% for all the fullerenes, >95% for C₆₀ and C₇₀) with only 45 min of extraction. Recoveries for all fullerenes did not improve significantly using 20 mL or 30 mL of toluene and decreased when applying two sequential extractions with toluene, probably due to an increase in evaporation steps and resulting manipulation of the sample. Similar trends were observed with 8.0 μm cellulose nitrate filters, although recoveries were slightly lower. The poorest recoveries were obtained with 0.45 μm cellulose nitrate filters and no significant differences between extraction volumes could be identified.

Fig. 1 Recoveries of (functionalised) fullerenes with different filters and extraction volumes (asterisk showing the chosen volume). The extraction time was kept constant.

Therefore, longer extraction times were tested for the 8.0 μm (Fig. 2) and 0.45 μm cellulose nitrate filters (Fig. 3), to which fullerenes seemed to attach more strongly. Optimum times were 60 min, 45 min and 90 min for 8.0 μm, 0.7 μm and 0.45 μm, respectively. Extraction times longer than these ones did not result in a significant improvement or even caused a slight decrease in the recovery. Extraction time was set to 60 minutes for all filters.

Fig. 2 Recoveries of fullerenes on 8.0 μm cellulose nitrate filter, obtained with different extraction times.

Fig. 3 Recoveries of fullerenes on 0.45 μm cellulose nitrate filter, obtained with different extraction times.

Fullerene distribution in wastewater samples

The distribution of the fullerenes in the wastewater was studied by separately analysing each of the filtrate fractions of a spiked influent wastewater free from fullerenes at 2 ng L⁻¹. Furthermore the distribution was analysed at different equilibration times allowed after spiking: from 1 min 180 min.

The results are summarised in Fig. 4.

Fig. 4 Fullerene distribution on 8, 0.7 and 0.45 μm filters and the aqueous phase as a function of time (min) allowed for equilibration after spiking.

Due to the complexity of the matrix, homoaggregates (solely one type of fullerenes) with a size above 450 nm (unpublished results) are unlikely to occur. In general heteroaggregate (fullerenes with other particles from the matrix) formation³⁰ appears to be quite fast in the samples except for [60]PCBO and [60]PCBB probably due to their longer butyl and octyl-ester groups. As can be seen, fullerenes distributed widely and they could be detected initially in almost all the fractions. When the equilibration time reached 180 minutes, the contribution of the aqueous fraction to the total mass balance was reduced to (almost) zero. The fullerenes aggregated with the respective particulate fractions. For example, after 2 h of equilibration time, 41 to 79% was recovered in the particulate fraction above 8.0 μm, and 7 to 25% of the fullerenes were recovered in the particulate fraction of 8.0 μm > x > 0.7 μm. The amounts of fullerenes were slightly lower in the particulate fraction of 0.7 μm > x > 0.45 μm, 8 to 22%. Finally, around 0 to 18% of the fullerenes passed through the three filters and were detected in the aqueous phase after 2 h. The present data suggest that most of the fullerene aggregates are likely to be captured by the filters due to heteroaggregate formation with the particulate matter. This implies that after some considerable time almost 100 percent of the fullerenes are likely to precipitate. Pakarinen³¹ demonstrated that a portion of nC₆₀ aggregates may remobilise from the sediment. Fig. 5 shows the method scheme that was used to quantify all possible fractions.

Fig. 5 Final method scheme.

Atmospheric pressure ionisation mass spectrometry

Using an ESI interface and an Orbitrap-MS in negative mode, methanol or acetonitrile is most often used to generate the electrons needed to produce a molecular ion. Depending on the HPLC stationary phase, they can be added to the mobile phase (reversed phase) or post-column (normal phase). The ESI process induces the complex adduct formation³² with the inconveniences previously stated.^8,17 This unfavourable behaviour hinders the process of qualifying and quantifying. For fullerene analysis, equal adduct patterns are obtained. Pristine fullerenes form adducts with oxygen, hydrogen, methanol and toluene (Fig. 6). For the functionalised fullerenes, a minor deprotonation product can be observed (<5% relative to the molecular ion peak [M-H]⁻˙) (Fig. 7).

Fig. 6 Mass spectra of C₆₀ with post column infusion of methanol generated by using an ESI-Orbitrap MS.

Fig. 7 Mass spectra of C₆₀PCBM with post column infusion of methanol generated by using an ESI-Orbitrap MS.

By using APPI instead of ESI, this complex adduct formation is reduced dramatically and hence the sensitivity is enhanced.^28,32,33 Still, adduct formation can occur by using methanol³² or acetonitrile in combination with toluene,²⁸ which is indispensable in a reversed phase HPLC system. Nuñez et al. (2012)²⁸ demonstrated that the ¹³C isotope is influenced by the addition of hydrogen to the fullerene double bonds. Switching to a normal phase system with a 100% isocratic toluene system eliminates most of the adduct formation and solely results in a minor oxygen adduct (0.23%) (Fig. 8).

Fig. 8 Mass spectra of C₆₀ generated by using a mobile phase consisting of 100% toluene and an APPI-Orbitrap-MS.

This results in an experimental ¹³C/¹²C ratio which correlates well with the theoretically predicted ratio (Fig. 9). By using a 100% isocratic elution with toluene and eliminating adduct formation (through ionisation of methanol or acetonitrile), the quantification becomes more straightforward compared to the common approach to sum up all adducts. For identification of pristine fullerenes, since they do not exhibit any fragmentation even at high collision energies, the most common approach is to use the oxidised ion as confirmation [C₆₀O]⁻˙. As demonstrated in Fig. 9, with 100% toluene APPI-Orbitrap, the ¹³C/¹²C ratio correlates well with the theoretical one and could be used for qualifying fullerenes instead of the very low abundant oxygen adduct. This would comply with Commission Decision 2002/657/EC³⁴ which states that the ratio of the intensity of a diagnostic ion from the isotopic pattern such as ¹³C_N-1 with a relative intensity higher than 10% of the ¹²C_N should be within ±20% of the theoretical ratio with the base peak. Fig. 9 demonstrates that if methanol is used in the mobile phase this criterion will often not be met. Hence identification was achieved for all analytes (4 points) by retention time agreement for the standard and sample (within 2.5%), monitoring the parent mass in high resolution (2 points) and confirmation of the ¹³C/¹²C ratio (2 points).

Fig. 9 Mass spectra of C₆₀ obtained using an APPI-Orbitrap. (A) Isotopic pattern with methanol/toluene (25/75 v%) mobile phase – APPI-Orbitrap. (B) Isotopic pattern, 100% toluene APPI-Orbitrap. (C) Theoretical isotopic pattern, C60. Resolution 30 000 FWHM 400 m/z.

The use of APPI-HRMS instrumentation results in better limits of detection and quantification than those obtained in previous studies. For instance, they are one order of magnitude lower than those described in Farré et al. (2010)¹⁴ using ESI-MS/MS, highlighting the suitability of the APPI-HRMS configuration for these compounds. Also, these parameters are of the same order (or slightly lower) than those obtained by Kolkman et al. (2013),⁸ analysing spiked ultrapure water by ESI-HRMS.

Moreover, the method presented here requires a sample volume that is 10 times smaller. The quality parameters were obtained with a spiked wastewater matrix. Similarly, Nuñez et al.²⁸ described a MLOD of 0.01 ng L⁻¹ for both C₆₀ and C₇₀, although in a freshwater matrix and using a triple quadrupole MS with enhanced mass resolution.

Method validation

Instrumental quality parameters are summarised in Table 2. As can be seen, instrumental limits of detection were in the fg range (between 7.9 and 31 fg). Interday repeatability was below 11% for all the compounds (between 4 and 11%) and it was significantly better when considering intraday runs (between 1.3 and 4.5%). Linearity ranges were acceptable through the studied range assessed between 128 fg and 1.28 ng. The sensitivity was slightly better for those fullerenes containing functional groups that allow an easier ionisation. Detection limits expressed as sample concentrations were in the pg L⁻¹ order (between 66 and 440 pg L⁻¹) which implies that the method can be applied to real samples. Recoveries for C₆₀ and C₇₀ for the filters and filtrate of the samples were near 100% at all the tested concentrations and they were acceptable for the functionalised fullerenes (49–104%) (Table 2). Interday repeatabilities were found to be between 1 and 15%.

Table 2 Instrumental and method quality parameters

			Bis[60]PCBM	[60]PCBO	[60]PCBB	[60]PCBM	[60]ThCBM	[70]PCBM	C60	C70
a Instrumental linearity was assessed between 125 fg and 1.28 ng. b Method linearity was assessed between 10 and 500 ng L^−1. c Method repeatability was assessed in quintuplicate and each of the replicates was extracted on a consecutive day.
Instrumental parameter	ILOD (fg)		15	11	22	16	31	21	7.9	10
	ILOQ (fg)		50	37	74	52	100	69	26	34
	Instrumental linearity (R^2)^a		0.9999	0.9996	0.9995	0.9999	0.9982	0.9997	0.9998	0.9979
	Sensitivity [a.u.]		15.8	14.0	15.4	16.9	15.1	10.4	8.85	8.18
	Intraday (n = 7) repeatability (%)	10 pg	2.34	3.40	3.27	2.62	4.17	3.22	2.60	4.58
		80 pg	1.34	1.49	1.54	1.36	1.81	1.83	1.65	1.31
		640 pg	3.56	3.73	3.48	2.74	2.79	1.51	4.44	3.03
	Interday repeatability (%)	10 pg	8.44	8.32	9.36	9.47	9.50	8.51	7.53	6.68
		80 pg	10.7	9.55	7.52	8.35	8.94	7.15	4.89	4.09
		640 pg	6.70	6.81	5.42	7.72	6.44	5.84	8.00	4.01
Method quality parameters	MLOD (pg L^−1)		77	68	75	83	74	66	210	440
	MLOQ (pg L^−1)		260	230	250	280	250	220	710	1400
	Method linearity (R^2)^b		0.997	0.998	0.995	0.994	0.995	0.991	0.991	0.991
	Method interday repeatability (%)^c	10 ng L^−1	8.51	11.3	14.9	9.98	14.9	11.3	13.5	12.2
		50 ng L^−1	3.91	4.24	5.20	10.9	8.60	8.24	11.5	7.90
		1000 ng L^−1	3.94	2.58	4.76	3.03	3.03	1.00	3.19	7.24
	Recovery (%)	10 ng L^−1	87.0	88.1	99.6	104	97.8	103	104	106
		50 ng L^−1	48.3	67.7	74.0	80.3	82.4	79.0	105	105
		1000 ng L^−1	50.1	72.1	101	97.6	98.9	88.9	98.6	102
	Accurate mass [M]˙^−		1100.19931	1008.20948	952.14688	910.09993	916.05635	1030.09993	720.00055	840.00055
	Theoretical ^13C/^12C ratio (%)		91	86	81	78	77	89	65	76
	Experimental ^13C/^12C ratio (%± RSD) [N = 14]		94.2 ± 0.4	88.1 ± 0.3	83.7 ± 0.6	80.8 ± 0.5	78.9 ± 0.3	91.9 ± 0.4	67.2 ± 0.4	78.6 ± 0.5

---

Analysis of field samples

To evaluate the applicability of the method to determine fullerenes by APPI-HRMS, 24 h flow-dependent composite influent samples were analysed from six different sewage water treatment plants. The results are shown in Table 3.

Table 3 Results showing concentrations of C₆₀ (ng l⁻¹) in six different sewage treatment plants. All other fullerenes were below the LOD

			^13C/^12C ratio
STP	Type	[C60] (ng L^−1)	%	Deviation from standard
a <: LOQ > LOD.
1	Influent	19	64	−2%
1	Effluent	< (0.1)^a	70	8%
1	Sediment receiving river	12 ng kg^−1	68	5%
2	Influent	5	61	−6%
2	Effluent	< (0.2)^a	73	12%
3	Influent	4	65	0%
3	Effluent	< (0.1)^a	75	15%
4	Influent	7	68	5%
4	Effluent	<	69	6%
5	Influent	9	65	0%
5	Effluent	< (0.1)^a	67	3%
6	Influent	2	66	2%
6	C-stage	1	72	11%
6	N-stage	<	n/a	—
6	Effluent Nereda™ reactor	<	n/a	—
6	Effluent	<	n/a	—

---

In order to establish the efficiency of the STPs, the total concentration of fullerenes in the effluents was determined according to Kolkman et al.⁸ disregarding possible fractionation (because of a much lower suspended particle concentration). The highest level of C₆₀ (19 ng L⁻¹) was observed in the influent of STP 1. In the effluents, no appreciable concentrations of fullerenes could be observed. A sediment sample was collected downstream from Treatment plant 1, extracted according to Carboni et al. (2013)²³ and analysed with the toluene-APPI-HRMS method. A level of 12 ng kg⁻¹ was observed.

Conclusions

In the present work, a simple isocratic HPLC method for analysing eight fullerenes in wastewater samples by using an APPI-Orbitrap MS has been developed and validated. The method allows the accurate mass identification of fullerenes aggregated with the particulate matter and those aggregated in the aqueous phase. The application of a normal phase column in combination with APPI avoids the use of methanol, thus minimising adduct formation in the interface, and allows a straightforward identification with unambiguous mass spectra never demonstrated before. Compared with ESI, the method is more sensitive and more robust, mass spectra are free from adducts, and experimental isotopic patterns match the theoretical ones. In all influent samples measured, only C₆₀ was encountered at levels up to 19 ng L⁻¹. The levels of C₆₀ observed in the effluents were all above the method detection limit but below the quantification limit.

Acknowledgements

This work is funded by the Joint Research Program of the Dutch Water Utilities (BTO) and NanoNextNL, a micro- and nanotechnology consortium of the Government of The Netherlands and 130 partners. The authors wish to thank the sewage treatment plants for providing wastewater samples. J. Sanchís is very grateful to the KWR Watercycle Research Institute for allowing him to perform an internship as a visiting scientist.

Notes and references

1. H. W. Kroto, J. R. Heath, S. C. O'Brien, R. F. Curl and R. E. Smalley, Nature, 1985, 318, 162–163.
2. R. D. Handy, R. Owen and E. Valsami-Jones, Ecotoxicology, 2008, 17, 315–325.
3. S. Utsunomiya, K. A. Jensen, G. J. Keeler and R. C. Ewing, Environ. Sci. Technol., 2002, 36, 4943–4947.
4. T. Laitinen, T. Petäjä, J. Backman, K. Hartonen, H. Junninen, J. Ruiz-Jiménez, D. Worsnop, M. Kulmala and M.-L. Riekkola, Int. J. Mass Spectrom., 2014, 358, 17–24.
5. T. M. Benn, P. Westerhoff and P. Herckes, Environ. Pollut., 2011, 159, 1334–1342.
6. Z. Chen, R. Mao and Y. Liu, Curr. Drug Metab., 2012, 13, 1035–1045.
7. C. J. Brabec, S. Gowrisanker, J. J. M. Halls, D. Laird, S. Jia and S. P. Williams, Adv. Mater., 2010, 22, 3839–3856.
8. A. Kolkman, E. Emke, P. S. Bäuerlein, A. Carboni, D. Truc Tran, T. L. Ter Laak, A. van Wezel and P. de Voogt, Anal. Chem., 2013, 12, 5867–5874.
9. F. Piccinno, F. Gottschalk, S. Seeger and B. Nowack, J. Nanopart. Res., 2012, 14, 1–11.
10. S. K. Brar, M. Verma, R. D. Tyagi and R. Y. Surampalli, Waste Manage., 2010, 30, 504–520.
11. T. Y. Sun, F. Gottschalk, K. Hungerbühler and B. Nowack, Environ. Pollut., 2014, 185, 69–76.
12. M. R. Wiesner, G. V. Lowry, P. Alvarez, D. Dionysiou and P. Biswas, Environ. Sci. Technol., 2006, 40, 4336–4345.
13. J. Sanchís, N. Berrojalbiz, G. Caballero, J. Dachs, M. Farré and D. Barceló, Environ. Sci. Technol., 2011, 46, 1335–1343.
14. M. Farré, S. Pérez, K. Gajda-Schrantz, V. Osorio, L. Kantiani, A. Ginebreda and D. Barceló, J. Appl. Hydrol., 2010, 383, 44–51.
15. H. C. Chen and W. H. Ding, J. Chromatogr. A, 2012, 1223, 15–23.
16. F. Gottschalk, T. Sonderer, R. W. Scholz and B. Nowack, Environ. Sci. Technol., 2009, 43, 9216–9222.
17. A. P. van Wezel, V. Morinière, E. Emke, T. ter Laak and A. C. Hogenboom, Environ. Int., 2011, 37, 1063–1067.
18. D. Bouchard, X. Ma and C. Isaacson, Environ. Sci. Technol., 2009, 43, 6597–6603.
19. J. D. Fortner, D. Y. Lyon, C. M. Sayes, A. M. Boyd, J. C. Falkner, E. M. Hotze, L. B. Alemany, Y. J. Tao, W. Guo, K. D. Ausman, V. L. Colvin and J. B. Hughes, Environ. Sci. Technol., 2005, 39, 4307–4316.
20. W.-C. Hou, L. Kong, K. A. Wepasnick, R. G. Zepp, D. H. Fairbrother and C. T. Jafvert, Environ. Sci. Technol., 2010, 44, 8121–8127.
21. Z. Wang, J. Chen, Q. Sun and W. J. G. M. Peijnenburg, Environ. Int., 2011, 37, 1078–1082.
22. D. Heymann, Fullerene Sci. Technol., 1996, 4, 509–515.
23. A. Carboni, E. Emke, J. R. Parsons, K. Kalbitz and P. de Voogt, Anal. Chim. Acta, 2014, 807, 159–165.
24. C. Wang, C. Shang and P. Westerhoff, Chemosphere, 2010, 80, 334–339.
25. P. Herrero, P. S. Bäuerlein, E. Emke, E. Pocurull and P. de Voogt, J. Chromatogr. A, 2014, 1356, 277–282.
26. C. W. Isaacson, C. Y. Usenko, R. L. Tanguay and J. A. Field, Anal. Chem., 2007, 79, 9091–9097.
27. J. F. Anacleto and M. A. Quilliam, Anal. Chem., 1993, 65, 2236–2242.
28. O. Núñez, H. Gallart-Ayala, C. P. B. Martins, E. Moyano and M. T. Galceran, Anal. Chem., 2012, 84, 5316–5326.
29. X. Qu, Y. S. Hwang, P. J. J. Alvarez, D. Bouchard and Q. Li, Environ. Sci. Technol., 2010, 44, 7821–7826.
30. J. T. K. Quik, I. Velzeboer, M. Wouterse, A. A. Koelmans and D. van de Meent, Water Res., 2014, 48, 269–279.
31. K. Pakarinen, J. Akkanen, M. T. Leppänen and J. V. K. Kukkonen, Chemosphere, 2014, 108, 320–325.
32. L. Li, S. Huhtala, M. Sillanpää and P. Sainio, Anal. Bioanal. Chem., 2012, 403, 1931–1938.
33. S.-I. Kawano, H. Murata, H. Mikami, K. Mukaibatake and H. Waki, Rapid Commun. Mass Spectrom., 2006, 20, 2783–2785.
34. European Comission, Off. J. Eur. Communities, 2002, 221, 8.

---

Footnotes

† Co-first authors.

‡ Visiting scientist at KWR.

---