Introduction

Fullerenes are a nanostructured class of carbon allotropes, discovered and structurally elucidated by Kroto et al. in 1985 [1]. Currently, the possibility to manipulate material at the nanoscale and the physicochemical properties exhibited by the fullerenes, such as thermal stability, conductivity, adsorption and catalytic capacity, have driven the exploration of their potential applications in different industrial fields [2]. In the USA, their industrial production was estimated at between 2 and 81 tons, in 2011 [3]. However, in the coming years, their use is expected to increase because of their inclusion in new drug delivery systems [4], medical imaging [5], antitumor agents [6], microelectronics [7], solar panels [8] and new materials [9].

Currently, however, fullerene risk assessment analysis is impaired by the lack of works dealing with their characterisation and analysis in complex environmental matrices [10]. Therefore, there are gaps of information with respect to their environmental occurrence, fate and transport. Moreover, as a consequence of the alteration of their physicochemical properties and aggregation states, toxicity tests appropriated for bulk substances may introduce artefacts during their study [11–13]. Therefore, specific and standardised toxicity tests, addressing these limitations and using new endpoints or new toxicity biomarkers for nanomaterials, continue being required in order to obtain more reliable data [11]. In contrast, while potential negative effects, such as antimicrobial activity [14], bioaccumulation [15, 16] and sub-lethal effects to aquatic organisms [17, 18], have been related to high concentrations of fullerenes in surface water, there is still a scarcity of available information about their occurrence and physicochemical states in real environmental scenarios. In addition, during recent years, the potential association of carbon-based nanomaterials with other organic contaminants, and how they may modulate their bioavailability and toxicity, have been subject to scientific attention [19].

In defining real environmental scenarios of exposure, one of the main limitations has been the development of analytical methods sensitive enough for the detection of fullerenes in complex samples. Currently, different sample preparation approaches have been presented, based on liquid–liquid extraction (LLE) [20–24], solid-phase extraction (SPE) [20–24] and filtration followed by ultrasonic-assisted extraction (UAE) with toluene [25, 26]. High-performance liquid chromatography coupled with tandem mass spectrometry (HPLC–MS) has been the technique of choice for their separation and detection [25–31]. Table 1 summarises the main analytical methodologies, which have been applied to water samples. Electrospray ionisation (ESI) [25, 27], atmospheric pressure ionisation (APCI) [32] and atmospheric pressure photoionisation (APPI) [31] have been used as ionisation techniques. However, it is noteworthy that the best performance in terms of sensitivity has been achieved by APPI. Furthermore, few studies have assessed the presence of fullerenes in the aquatic environment [25, 31, 33, 34], to date. In these studies, the concentrations of fullerenes in surface water ranged from the method limits of detection (MLOD) to 130 ng/l of C60 fullerene [34]. C60 and C70 fullerenes were also recently detected in particulates from the Mediterranean Sea atmosphere [28]. In addition, surface waters are natural receptors of fullerenes from wastewater disposal [25] and atmospheric deposition [28].

Table 1 Selected relevant methods for the HPLC–MS analysis of fullerenes in water samples
Full size table

In order to better understand the environmental behaviour of fullerenes, besides the quantitative analysis, their characterisation in the environment is required.

In this study, a novel combined approach is presented in order to characterise the nanoparticles’ content and size and to assess the occurrence of seven fullerenes (C60, C70, N-methylfulleropyrrolidine, [6,6]-phenyl C61 butyric acid methyl ester, [6,6]-thienyl C61 butyric acid methyl ester, C60 pyrrolidinetris-acid ethyl ester and [6,6]-phenyl C71 butyric acid methyl ester) in environmental water samples and wastewater. The size distribution and content of nanoparticles were assessed by nanoparticle tracking analysis (NTA). In parallel, a method based on UAE with toluene, followed by liquid chromatography coupled to high-resolution mass spectrometry, using an APPI ionisation source in negative conditions (LC–APPI–HRMS), was applied for the analysis of fullerenes. In comparison with the previous approaches, this analytical method presents mass accuracy, robustness and an improved sensitivity.

This combined approach was used to assess 48 real environmental samples, collected within the vicinities of Barcelona City (Spain): 12 river sediments (4 from the Besòs River and 8 from the Llobregat River) and 36 water samples (11 treated effluents from 6 different wastewater treatment plants (WWTPs), 25 river waters (10 from the Besòs River and 15 from the Llobregat River)).

This study contributes to augmenting data on fullerene occurrence in real surface water.

Experimental section

Chemicals and materials

Fullerene standards C60 fullerene (99.9 % purity, CAS: 99685-96-8), C70 (99 % purity, CAS: 115383-22-7), N-methylfulleropyrrolidine (abbrev. [60]NMFP, 99 % purity, CAS: 151872-44-5), [6,6]-phenyl C61 butyric acid methyl ester (abbrev. [60]PCBM, >99.5 % purity, CAS: 160848-22-6), [6,6]-thienyl C61 butyric acid methyl ester (abbrev. [60]ThPCBM, ≥99 % purity, CAS: 925673-03-6), C60 pyrrolidinetris-acid ethyl ester (abbrev. [60]PTAE, >97 % purity) and [6,6]-phenyl C71 butyric acid methyl ester (abbrev. [70]PCBM, 99 %, CAS: 609771-63-3) were purchased from Sigma-Aldrich (Steinheim, Germany). 13C-labelled C60 fullerene was purchased from MER Corporation (Tucson, AZ, USA).

Sodium chloride (≥99.5 % ACS reagent, reference 71380-500G) was purchased from Sigma-Aldrich (Steinheim, Germany). Toluene and methanol CHROMASOLV® were supplied by Merck (Darmstadt, Germany). Nitrogen, used as drying gas with 99.995 % purity, was acquired from Air Liquide (Spain).

Sampling

A total number of 48 samples (10 surface waters and 4 sediments from the Besòs River, 15 surface waters and 8 sediments from the Llobregat River and 11 wastewater effluents) were collected within the vicinities of Barcelona City (Spain). Details are summarised in Table 2.

Table 2 Characteristics of the sediment, wastewater and surface water samples
Full size table

River water samples were collected during periods December 2013 and February 2014, after a dry climatic period and shortly after a precipitation period, respectively. The river sediments were collected during the second sampling campaign at the same sampling sites as those of the surface waters (Table 2).

Wastewater effluents were taken from six different WWTPs in the municipalities of Sabadell and Montcada i Reixac, serving about 207,649 and 34,863 inhabitants, respectively, and discharging at the Besòs River, and also from the municipalities of Terrassa, Manresa, Rubí and Sant Feliu de Llobregat, serving about 215,055, 76,170, 74,468 and 43,769 inhabitants, respectively, and discharging at the Llobregat River.

Water samples were collected in amber glass bottles in duplicates, and a metal dredge was employed for sediment sampling. The samples were immediately transported to the laboratory at a temperature of 4 °C. The sample pre-treatment and extraction were carried out within the 24 h following arrival at the laboratory.

Sample preparation

Water samples were extracted according to a previous protocol [25] after some modifications and additions. In brief, 100 μl of 13C-labelled C60 fullerene at 1 pg/μl in toluene was dispersed in 2 ml of ethanol and spiked to 150 ml of surface water or to 100 ml of effluent wastewater samples. Samples were mixed thoroughly and filtered through 0.7 μm glass fibre and 0.45 μm nylon filters (Whatman, Maidstone, UK), previously tared. Particulate and filtered water were extracted and analysed in parallel for both surface water and wastewater samples.

The filters were dried at 60 °C for 24 h and were then weighed (the remaining humidity did not interfere in the analyses). Afterwards, the filters were extracted by UAE with 35 ml of toluene for 45 min. This process was repeated twice more. The combined extracts were collected in an amber spherical flask, and the excess of toluene was eliminated by rotary evaporation to 0.5 ml, approximately. Then, it was quantitatively transferred to a tared amber vial. The final volume was adjusted by weight to 1000 ml with toluene.

To assess the fullerenes in the filtrated samples, a large-volume liquid–liquid extraction (LLE), based on the method described by Bouchard and Ma [21], was performed. In brief, the filtered samples (100 ml of wastewater effluents or 150 ml of river waters) were placed in a separatory funnel and NaCl was added in order to obtain a 1 % saline solution. Then, the extraction was performed with toluene. After centrifugation, the toluene fraction was collected. This process was repeated twice more. Then, the process was continued, as already described.

Sediment samples were extracted after minor modifications, according to the procedure described by Sanchís et al. [26]. Wet sediments were centrifuged at 3000g for 5 min. After water decantation, sediments were dried overnight in a desiccator. Then, the samples were ground with an agate mortar and sieved (<125 μm). Ten grams of sieved sediment was placed into a glass vessel and fortified with 100 μl of 13C-labelled C60 fullerene suspension (1.0 pg/μl in toluene) and used as internal standard. Then, sediment samples were homogenised and left in the dark for 3 h at 4 °C in order to reach equilibrium. After this time, 40 ml of toluene was added and sediments were extracted by UAE during a period of 4 h. The suspension was centrifuged at 3000g for 10 min. This process was repeated twice more. Then, the combined extract (80 ml) was introduced into a spherical flask and concentrated by rotatory evaporation to 1 ml, approximately. Afterwards, it was transferred to a pre-tared vial and evaporated with a gentle nitrogen stream to 500 μl, approximately. The final volume was adjusted to 1.000 ml by weight.

Instrumental analysis

The separation of fullerenes was achieved by non-aqueous LC, using the ACQUITY UPLC System (Waters, Milford, MA, USA), equipped with a Buckyprep (150 × 2.0 mm, particle size 5 μm) analytical column from Nacalai Tesque (Kyoto, Japan). The mobile phase composition consisted of 100 % toluene, and the flow rate was kept at 0.400 ml/min. The separation of selected fullerenes was achieved in less than 4 min, throughout the total chromatographic run of 10 min. The sample injection volume was set at 10 μl. The LC system was coupled to a Q Exactive (Thermo Fisher Scientific, San Jose, CA, USA) hybrid quadrupole-Orbitrap mass spectrometer with an APPI source, acquiring data in APPI/APCI mode with negative polarity. The acquisition was performed in full scan (m/z 600–1200) with a resolution of 35,000 full width at half maximum (FWHM).

The method parameters were optimised by continuous infusion of a mixture of 1 mg/l standard solution of each fullerene, via syringe pump at a flow rate of 5 μl/min. The optimal APPI source conditions were as follows: sheath gas at 50 a.u., auxiliary gas at 35 a.u., spray voltage at 5.0 kV, spray current set at 10 μA, S-lens RF at 70, capillary T at 350 °C and probe T at 450 °C. In order to improve the sensitivity, the spare gas was closed.

For instrument control and data processing, Xcalibur 2.2 (Thermo Fisher Scientific) software was used. For the unambiguous determination of fullerenes, the molecular ion [M]˙− (in the case of pristine fullerenes, [12C x ]˙−, x = 60 or 70) was used for quantification and the isotopic signal [M+1]˙− (for pristine fullerenes, [13C1 12C(x − 1)]˙−, x = 60 or 70) was used for confirmation purposes. Overall, the number of identification points required for confirmation of organic contaminants in environmental analyses [35] was fulfilled.

In order to assess the potential carry-over contamination, in each injection batch, instrumental and procedural blanks and a standard mixture were injected. An instrumental blank was introduced after each five injections. In addition, after each batch of samples, a cleaning sequence was performed, consisting of the injection of toluene, using a toluene–acetonitrile gradient to remove polar interference substances from the system.

Quality control and quality assurance

The analytical method was validated with special attention to linearity, recovery, matrix effects, precision, limits of detection (LOD) and limits of quantification (LOQ). The instrumental quality parameters were calculated using a ten points calibration curve in toluene, ranging from the instrumental LOD to 500 ng/ml. The instrumental limits of detection (ILOD) were calculated by progressive dilution as the lowest concentration at which each compound could be detected. Instrumental linearity and sensitivity were estimated using the Pearson index (R 2) and the slopes of the calibration curves, respectively. Instrumental reproducibility (inter-day precision) was calculated as the average percentage of the relative standard deviation (RSD%) of standard solutions (ten replicates) at three concentration levels (50, 100 and 500 fg/μl) on three consecutive days.

The method limits of detection (MLOD) and quantification (MLOQ) were based on matrix-matched calibration curve points. The MLOD and MLOQ of each analyte were defined as the lowest concentrations which fulfilled the criteria: (1) signal-to-noise ratio, at least, 3 and 10, respectively; (2) relative standard deviation of three replicates, below 22 %; (3) Gaussian peak shapes; (4) less than 3 ppm of exact mass error; and (5) adequate molecular isotopic pattern.

The reproducibility (inter-day precision) was calculated as the average relative standard deviation (in percentage) of the analytes (n = 10) of blank wastewater effluents fortified with the seven analytes at three concentrations (10, 50 and 1000 ng/l). Recoveries were calculated by the analysis of the same samples.

Matrix effects were assessed by comparing the analytical signals of the standard suspension and the analytical signals of fortified extract at the same concentrations (0.2, 2.0 and 2.0 ng/g). Sediment extraction was previously validated [26].

Nanoparticle tracking analysis

Nanoparticle tracking analysis (NTA) utilises the properties of both light scattering and Brownian motion in order to obtain particle size distributions of samples in liquid suspension. A laser beam is passed through a prism-edged flat glass within the sample chamber. The angle of incidence and refractive index of the flat glass are designed to be such that when the laser reaches the interface between the glass and the liquid sample layer above it, the beam refracts, resulting in a compressed beam with a reduced profile and high power density. The particles in suspension in the path of this beam scatter light which can be detected [36]. NTA was performed in flow mode with a NanoSight LM10-HS instrument (NanoSight Amesbury, UK) equipped with an LM14 nanoparticle viewing unit with temperature control, a CCD monochrome camera and an automated syringe pump.

NTA calibration was carried out with stock suspensions of polystyrene latex nanoparticles with diameters of 100, 200 and 400 nm. A pre-scan was carried out in order to adjust the camera level and data threshold to the characteristics of each sample, prior to sample characterisation. In order to track a significant number of nanoparticles, measurements were carried out in quintuplicate, with recordings of time periods in the range of 60–90 s. The five size population distributions were normalised and averaged in order to obtain the final plot.

In order to avoid modification of the aggregation states of the samples, they were not frozen. Prior to performing the NTA analysis, to discard micro-sized suspended particles, the samples were filtered through a 450-nm pore size filter and then were shaken, prior to their analysis.

Precautions and safety considerations

In order to avoid degradation and oxidation processes, the samples and the extracts were stored in the dark, and exposure to air was minimised. Glass materials were heated to 400 °C, overnight, and rinsed with toluene, prior to their use. To circumvent cross-contamination, only glass was used. In order to minimise the potential losses by adsorption to the containers, the use of plastic materials was avoided during the whole analytical process. For the same reason, complete dryness was prevented during the evaporation of the extracts. Finally, in order to minimise any health risks, all manipulations and the solvent transferring procedures were performed under a fume hood. Since toluene was used as the mobile phase, the LC waste bottle was vented into the fume hood.

Results and discussion

Analytical method

The novel analytical approach for water matrices reported in this work combines UAE with toluene followed by LC, using a pyrenylpropyl group bonded silica based column, coupled to a high-resolution mass spectrometer (HRMS) using a dual-mode atmospheric pressure photoionisation source, in conjunction with atmospheric pressure chemical ionisation (APCI/APPI) ion source (Ion Max™, Thermo Scientific). The hybrid ionisation source allows working in APCI, APPI or combined APCI/APPI mode. In terms of sensitivity, the best performance was obtained with the APPI or the combined APCI/APPI mode. Since slightly better results were obtained using the hybrid source (Fig. 1), it was selected for the optimised method. In contrast, a compromise between mass accuracy and sensitivity was adopted and the acquisition was set at 35,000 FWHM.

Fig. 1
figure 1

Intensities (n = 3) obtained using APPI, APCI and a hybrid APCI/APPI ionisation mode for a pristine fullerenes and b functionalized fullerenes

Full size image

Validation was carried out for surface water and wastewater effluents. Good linearity was obtained (Table 3), with calibration curves (in standard solution and fortified blank extracts) having correlation coefficients of R 2 > 0.99 and no significant residual trends. A slight percentage of ion suppression, ranging from 2 to 30 % in wastewater, was observed. However, potential matrix effects were corrected by internal standard addition. The ILOD and instrumental limits of quantification (ILOQ) values were ranging from 1 to 100 fg and from 3 to 300 fg, respectively. In comparison to previous works for fullerene analysis in environmental waters, this method presents a great improvement in terms of sensitivity [26]. The MLOD ranged from 0.9 to 4.9 pg/l for surface waters, from 2.3 to 9.4 pg/l for wastewater effluents and from 19 to 160 pg/kg for sediments. Recovery yields were compressed between 67.7 and 106 %.

Table 3 Instrumental quality parameters for the analysis of fullerenes by HPLC–APPI(−)–Q-Orbitrap-MS. Method quality parameters for the analysis of fullerenes in surface waters and wastewaters
Full size table

Results in environmental river samples and WWTP effluents

C60 and C70 fullerenes were confirmed in all the samples, while functionalised fullerenes were not detected. These results can be related to the present low production and/or the degradation to more stable forms. In Fig. 2, the concentrations of fullerenes in wastewater samples are summarised. C76, C78 and C84 fullerenes, which were detected in other works [31], were not detected here. In Fig. 3, chromatograms of each matrix are presented.

Fig. 2
figure 2

Concentrations in wastewater effluent samples

Full size image
Fig. 3
figure 3

Example of extracted chromatograms of a real surface water, b a sediment sample, c a standard mixture, and d a blank sediment

Full size image

The sums of C60 and C70 varied from 0.5 to 9.3 ng/l with mean concentrations of 1.1 and 1.7 ng/l for C60 and C70, respectively. C70 was quantified at slightly higher concentrations with median concentrations of 0.4 and 1.7 ng/l for C60 and C70, respectively. These concentrations are consistent with the predictions by the Gottschalk [37] model. The highest levels of fullerenes were from the WWTPs located in Sant Feliu de Llobregat, Rubí and Manresa that are highly populated and industrialised areas. Nevertheless, high variability between WWTPs was observed. Fullerenes were mainly associated with the particulate phase. This is consistent with the DLVO model which, because of the suppression of the electric double barriers in high-ionic-strength media, predicts the supra-aggregation of nanoparticles.

In surface waters, C60 and C70 fullerenes were quantified at much lower concentrations than in wastewater (Table 4), as expected. C60 fullerene was the most ubiquitous compound, being detected in 100 % of the samples in concentrations from 31 pg/l to 4.5 ng/l, while C70 concentrations ranged from <MLOD to 1.5 ng/l. Fullerenes are hypothesised to reach the river water by WWTP discharge and by atmospheric deposition.

Table 4 Fullerenes concentrations in wastewater and freshwater samples
Full size table

A significant difference between the fullerene content in the surface water of the Besòs and Llobregat rivers was found. The median concentrations of C60 fullerene were 0.92 and 0.39 ng/l in the Llobregat and Besòs rivers, respectively. In the Besòs River, the mean concentration was slightly higher than the predictions by Gottschalk et al. [37], while in the Llobregat River, much higher values were detected, with mean concentrations of 0.498 and 0.697 ng/l for C60 and C70, respectively. In the European Union (EU) and the USA for the scenarios projected by the Gottschalk model, predicted values were 17 and 3 pg/l, and for the particular Switzerland scenario, 40 pg/l. The differences with the Gottschalk model can be attributed to the characteristics of the Mediterranean rivers, with intense differences in seasonal water flows [38, 39].

The presence of fullerenes in both the large particulate (diameter Ø > 450 nm) and the colloidal (diameter Ø < 450 nm) fractions of surface waters is noteworthy (Table 4). In general, fullerenes were mainly found associated to the particulate fraction. But, in some of the samples (e.g. S22 and S23), about 80–90 % of the fullerenes were found in the fraction smaller than 450 nm.

An inverse tendency was found between the levels of C60 fullerene associated with the particulate and in the aqueous phase (Fig. 4a, R 2 = 0.6320, p = 0.033). This correlation is not consistent with the predictions by Langmuir or Freundlich models, which would typically show a positive slope. Therefore, the fullerene sorption onto particulate surface is not the predominant process, as would be expected for hydrophobic organic pollutants. Fullerene aggregation with naturally occurring colloids is likely to occur in natural waters, contributing to the stabilisation of fullerene aggregates. In contrast, these findings support the tenet that fullerene aggregate suspensions can be stable in environmental conditions, similarly with those suspensions prepared at lab scale by long-term stirring in Oberdörster et al. [17] and further characterised in Zhu et al. [40].

Fig. 4
figure 4

Relationships between C60 concentrations in river water samples and other parameters. a C60 concentrations in the particulate versus C60 concentrations in the aqueous phase (p = 0.033). b C60 concentrations in the aqueous phase versus C70 concentrations in the aqueous phase (p < 0.00). c Relationship between C60 concentrations in the aqueous phase and concentrations of nanoparticles (p = 0.023)

Full size image

In the colloidal phase, a significant correlation (R 2 > 0.94, p < 0.000) was found between the concentrations of C60 and C70 fullerenes (Fig. 4b), suggesting a common origin for both fullerenes. In addition, the concentrations of C60 were, in general, about three times the concentration of C70 and this ratio is consistent with those from combustion flame emissions [41]. Therefore, atmospheric deposition can be considered the main source in river water. However, this ratio was not observed in WWTP effluents, where the influences of the organic matter content, aggregation and supra-aggregation formation, adsorption to suspended material and precipitation processes play an important role in the compartment of fullerenes from the water column to sediments. In addition, this is also an indication of the low influence of WWTP effluents in fullerene concentrations in river water.

NTA analysis (see Fig. 5) provided a wide characterisation of the number and size of nanoparticles and the aggregate sizes in the water samples. The total concentrations of nanoparticles in wastewater effluents were up to 20 times higher than those in river samples (Table 5). However, no significant differences in terms of size distribution or total nanoparticle concentrations were found between river samples, before and after WWTP discharges. Most probably, abrupt changes in salinity, pH and organic material contents when effluents are discharged into the rivers produce flocculation and precipitation of colloids and suspended material.

Fig. 5
figure 5

Captions from NTA recordings of a HPLC water blank (a), a surface water sample, (b) a WW effluent (c) and a WW influent (d)

Full size image
Table 5 Concentrations of nanoparticles in water samples obtained by NTA
Full size table

River samples collected near highways, under the influence high-intensity vehicle traffic, such as Manresa and Montcada i Reixach, presented concentrations of nanoparticles in the same order as wastewater effluents. But, it is noteworthy that samples with a high concentration of nanoparticles in the sub-450 nm phase exhibited lower concentrations of C60 (see Fig. 4c). In these cases, suspended nanoparticles are mainly inorganic (metal nanoparticles, metal oxides and nano-sized mineral grains from natural erosion), while when the total concentration of nanoparticles is higher, carbon-based nanoparticles, such as fullerenes, tend to be associated to humic and fulvic acids, forming large-sized hetero-aggregates that tend towards precipitation, which is consistent with the results obtained in the analyses of sediment samples of the same sites.

Figure 6 shows the concentration of fullerenes in sediments. The only detected fullerene in sediments was C60, occurring in 75 % of the samples. The concentrations of C60 ranged from <MLOD to 32.1 pg/g, with median and average concentrations of 10.0 and 11.0 pg/g, respectively.

Fig. 6
figure 6

Concentrations in river sediment samples

Full size image

A certain regression was found when the partitioning of C60 fullerene (sub-450 nm) between river water and the sediment, \( {k}_{{{\mathrm{C}}_{60}}_{\left(\mathrm{a}\mathrm{q}\right)},\kern0.62em {\mathrm{C}}_{60\left(\mathrm{s}\mathrm{e}\mathrm{d}\right)}} \), versus the weight of suspended material in water, was studied. As can be seen in Fig. 7a, for river waters with higher concentrations of particulates, C60 fullerene tends to partition into the sediment compartment (R 2 = 0.60, p = 0.023). In addition, the concentration of C60 fullerene in sediments tends to be higher when the particulate weight in water is much higher (R 2 = 0.41, Fig. 7b).

Fig. 7
figure 7

Relationships between C60 concentrations in river sediment samples and other parameters. a Amount of suspended matter versus the distribution of fullerenes and the sediment and the water compartments (p = 0.041). b Amount of suspended matter versus the C60 concentrations in sediment (p = 0.017). c Concentrations of C60 in sediment versus concentrations of C60 in water (p = 0.025)

Full size image

These two observations stress the major role played by hetero-aggregation processes in the distribution of fullerenes in natural waters and sediments, particularly in these scenarios where concentrations of natural colloids are high, as previously indicated in Quik et al. [42].

However, no correlation was found between the fullerene concentration and large-diameter particulates. Instead, the concentrations of C60 in the aqueous phase (Ø < 450 nm) showed a statistically significant correlation with the concentrations of C60 in the sediment (R 2 = 0.6265, p = 0.025, Fig. 7c). The preferential relationship between the sub-450 nm fraction and the sediment concentrations can be explained by the higher surface charge of the smallest aggregates, which allows them to interact further with humic acids, natural matter and clayey particulates.

Conclusions

A novel analytical approach combining NTA and UAE–LC–APCI/APPI–MS has been presented. The method based on UAE–LC–APCI/APPI–MS has presented an improved sensitivity in comparison with previous methods. The approach was used to study selected fullerenes in 48 samples of river waters, sediments and wastewater effluents.

The main results of this study corroborate that, as a consequence of the current low production of engineered fullerenes and the much higher stability of pristine fullerenes, solely pristine fullerenes were detected in all types of samples (river water, sediments and WWTP effluents) at picogram per litre and picogram per kilogram levels.

In the colloidal phase, a significant correlation was found between the concentrations of C60 and C70 fullerenes, suggesting a common origin. In addition, the ratio C60/C70 was consistent with ratios from combustion processes [41]. On the contrary, this ratio was not maintained in WWTPs, indicating multiple sources.

NTA analysis was used to provide a wide characterisation of the number and size of nanoparticles and the size of aggregates in the water samples, showing that the total concentrations of nanoparticles in wastewater effluents were about ten times higher than in river samples. However, no differences were observed between river samples, before and after WWTP discharging points, either in terms of size distribution or in terms of total nanoparticle concentrations.