Size Discrimination and Detection Capabilities of Single-Particle

Article pubs.acs.org/ac

Size Discrimination and Detection Capabilities of Single-Particle ICPMS for Environmental Analysis of Silver Nanoparticles Jani Tuoriniemi, Geert Cornelis, and Martin Hassellöv* Department of Chemistry and Molecular Biology, University of Gothenburg, SE-412 96 Gothenburg, Sweden S Supporting Information *

ABSTRACT: The detection capabilities of single particle inductively coupled plasma-mass spectrometry (spICPMS) with respect to particle size and number concentrations are investigated for the case of silver nanoparticles (ca. 20−80 nm). An iterative algorithm was developed where particle measurement events were distinguished as outliers from the more continuous dissolved ion signal if the measured intensity was more than five times the standard deviation of the whole data set. The optimal dwell time for 40−80 nm particles, limiting both incomplete and multiple particle events, was 5 ms. The smallest detectable particle size (ca. 20 nm) is mainly limited by the overlap of particle events and dissolved signal that increases with noise on both signals. The lowest measurable number concentration is limited by the relative frequency of erroneously identified particle events, a limit that can be reduced by acquiring more data points. Finally, the potential of spICPMS for environmental detection of nanoparticles is demonstrated for a wastewater treatment plant effluent sample.

T

concentrations of ENP in aqueous samples. If a sufficiently dilute suspension of metal-rich particles is nebulized into the plasma, then a burst of ions will be generated when each discrete particle is vaporized, atomized, and ionized. The ions of the mass of interest are subsequently counted with a very high frequency similar to the time scale of a particle event. The theory of spICPMS for liquid analysis was outlined by Degueldre et al.,12−16 where they showed that the concentration of particles can be derived from the time-resolved signals. They also experimentally demonstrated the feasibility of the method for a range of particles of different materials with sizes larger than 80 nm.12−16 The spICPMS approach is further explored in this study for the case of Ag nanoparticles (AgNP). AgNP can be found in many consumer products today, particularly because of their antibacterial properties, and concerns have been raised about emissions into the environment.17−21 spICPMS has already been applied for characterization of the effluent of a washing machine in which AgNP are used.22 Laborda et al.23 estimated both AgNP particle number concentration and dissolved silver concentration from spICPMS measurements. A 3*σ threshold limit is usually adopted to discriminate the ion bursts of a particle from the background signal. While 3*σ is a commonly used definition for detection limits (DL), this concept cannot be used for spICPMS where the intensity of particle signals are outliers within a data set of which the mean

he increasing application of engineered nanoparticles (ENP) in consumer products is forecasted to have a major impact on most segments of the future society, and ENP are currently being used in a large variety of consumer products.1 Following indications that they are potentially harmful for organisms in the environment,2 the effect and exposure assessment of ENP need to be supported by in situ analysis and physicochemical characterization, which is difficult with currently available methods. For instance, models predicting the environmental concentrations of ENP in different compartments such as surface waters, soils, and sediments3,4 need to be validated with environmental measurements,5 but the quantification of ENP in complex environmental matrixes is very challenging due to the likely low environmental concentrations3,4 ( n*σdiss for a particle event to be recognized as such (derivation in Supporting Information). The choice of detection threshold, n*σ, in the algorithm is thus critical for the analysis, and n will be varied in this study. Counting Particle Events and the Particle Number Concentration Detection Limit. The average number of particles λ entering the plasma during a dwell, based on the first part of eq 1, is given by:

is mainly determined by the continuous signal originating from dissolved analytes. Moreover, it can be calculated that 0.13% of a Gaussian distributed dissolved signal is erroneously counted as particles if a 3*σ detection limit is assumed. A Poisson distribution is probably more appropriate for mean dissolved signals approaching zero,23 and in this case, ca. 0.5% of the data points are erroneously counted as particles. Previous work on spICPMS12−16,23 did not elucidate how many particles need to be counted in order to reach a certain level of confidence in the results. In addition, no method for determining the particle size without the use of standard particles of known sizes, which are currently not available for most nanoparticles, has hitherto been presented. In addition, the previous studies by Degueldre et al.12−16 only showed detection of particles of 80 nm and above. More recent work showed that 40 nm AgNP can be discriminated from the dissolved signal,23 but enhanced reactivities and toxicity have been demonstrated for sizes below approximately 30 nm.24,25 Consequently, there is a need to further develop and optimize spICPMS to reduce the lowest detectable size limit and to improve the methods for discriminating between the particle events and the background signal. The current study investigates the detection capabilities for AgNP using a sector-field ICPMS (Thermo-Finnigan Element 2). The software of this instrument allows one to reduce dwell times (also called sampling time), i.e., the detector acquisition time for each data point below 5 ms and the minimum dwell time used in previous work.12−16,23 It is discussed how this affects the size detection limits of spICPMS. A method for determining particle size using calibration with dissolved standards is also presented, and data analysis is improved. In addition, the number concentration detection limit of spICPMS and how it may be improved is discussed. Finally, as a feasibility check, an environmental sample from a wastewater treatment plant (WWTP) effluent is analyzed.

■

THEORY Signals of Dissolved and Particle Bound Analytes. The number of ions of dissolved analytes, Idiss, arriving at the detector in each data point (dwell) during the dwell time, tdwell (s), is given by:23 Idiss = tdwellfneb cdissAq × fis fion fplasma ftrans

(1)

Equation 1 contains two parts that account for dilution and efficiency factors in the sample introduction system, and plasma and mass spectrometer, respectively. In the first part giving the number of analyte atoms introduced to the plasma, f neb is the fraction of the analyte that passes through the nebulizer and spray chamber, the nebulization efficiency, cdiss, is the concentration of ions in the sample in mol L−1, A is Avogadro’s number, and q is the flow rate of sample into the nebulizer in Ls−1. The second part is the counting yield that determines the fraction of analyte atoms entering the plasma that are detected. f is is the abundance of the monitored isotope, f ion is the fraction of analyte that becomes singly charged positive ions in the plasma, f plasma is the fraction of analyte exiting the plasma through the sampler cone, and f trans is the transport efficiency or the fraction of ions that is transported from the sampler cone to the detector. The terms in the second part of eq 1 are most likely equal for dissolved or particulate signals because several studies have shown that there is no significant difference between conventional ICPMS analysis of acid digested or

λ = qfneb c ptdwell

(2)

where cp is the particle number concentration in the sample. The probability, pk, for obtaining single (k = 1) and multiple (k > 1) particle events and the fraction of dwells containing particle events, pp, can be calculated using Poisson statistics (further discussion in Supporting Information). The total number of particle events counted in a sequence of D dwells is D*pp. At low λ, the chance for multiple particle events, pk>1, is negligible relative to pk=1 and the approximation pp ∼ λ is valid. The particle count at the detector increases linearly with cp in this case but, at higher λ, i.e., higher cp, nebulization efficiency, flow rate, and/or dwell time, a lower particle count than predicted from the initial linear trend is obtained. The particle 3966

dx.doi.org/10.1021/ac203005r | Anal. Chem. 2012, 84, 3965−3972

Analytical Chemistry

Article

Figure 1. Time resolved signals of spICPMS and the corresponding signal intensity frequency histograms for samples of ultrapure water, a 1 μg L−1 dissolved Ag standard, a fresh 40 nm AgNP suspension, and an aged 40 nm AgNP suspension. The dashed lines indicate the 5*σ detection threshold.

for this instrument is found in Supporting Information. Dissolved standards were included in each analysis sequence, and the nebulization efficiency was measured in triplicate using the waste collection method.30,31 A detailed description of the measurement of nebulization efficiency and calibration of particle size using dissolved standards is found in Supporting Information. The nanoparticle stock suspensions were diluted in series up to 109 times in ultrapure water. The concentrations are listed in Table SI-2, Supporting Information. The nominal dwell time was set at 1, 5, or 10 ms, and 10 000 or 40 000 dwells of 107Ag measurements were collected for each dilution. Calibration curves, frequency histograms, and particle counts were calculated from the exported raw data using spICPMS theory in MATLAB (version 6). For each dilution series, the particle concentrations were determined using eq 2 and the measured nebulization efficiency on a sample with a high enough concentration to obtain sufficient counting statistics (details in Table SI-2, Supporting Information). These measured values were used to calculate the concentrations of the other samples in the dilution series using the dilution factor. WWTP Samples. WWTP effluents were sampled at the GRYAAB WWTP (Gothenburg, Sweden) the 19th of May 2009 and analyzed the same day. The sample was a daily flow composite sample from the final treated water from which three subsamples were filtered through an acid rinsed and neutralized 5.0 μm Millipore Durapore Millex (PVDF syringe filter with 3.9 cm2 filter area) followed by a 0.45 μm Millipore Durapore Sterivex-HV (PVDF filter with 10 cm2 filter area). The first 2 mL were discarded, and 10 mL of filtrate was collected into acid washed and neutralized polycarbonate vials. The dwell time was not optimized for these samples, and the minimum possible

number concentration can be determined either on the basis of measured nebulization efficiency and using eq 2 or from calibration curves prepared from particle standards (see below). The particle number detection limit can be related to the relative error in particle counting Er (derivation in Supporting Information): Er = 2z

1 N

(3)

where z is the z-score of the normal distribution which assumes the value of 1.96 for a 95% confidence interval.29 Given a desired accuracy, the particle number detection limit thus depends only on the number of particles counted, which increases with D.

■

EXPERIMENTAL SECTION Chemicals. AgNP suspensions were obtained from British biocell international (BBI, UK) (20, 40, 60, and 80 nm in nominal diameter). The particles were citrate-coated and consisted of elemental silver. Unless otherwise stated, the samples were measured briefly after the first opening of the container. Reference Au nanoparticle suspensions (30 and 60 nm nominal diameter) were obtained from NIST. Dissolved silver standards (ultra scientific) for ICPMS calibration were diluted in ultrapure water from a 1000 ppm solution in 2% HNO3. AgNP spICPMS Measurements. A Thermo Element 2 ICPMS (settings given in Table SI-2, Supporting Information) was used in all experiments after sensitivity and short-term stability optimization using 1 μg L−1 In in 2% HNO3. A discussion regarding issues with the dwell times that are specific 3967

dx.doi.org/10.1021/ac203005r | Anal. Chem. 2012, 84, 3965−3972

Analytical Chemistry

Article

dwell time, 0.1 ms, was used together with a relatively large number of dwells (D = 65 500) to maximize the chance of detecting a significant number of nanoparticle events in this first feasibility study that did not involve an elaborate sample characterization. To calibrate for particle number concentration dilution series of the NIST, Au standard particles were measured (calibration curves in Figure SI-8, Supporting Information). Concentrations in the WWTP samples were obtained from the calibration curves containing the frequency of particle events plotted against the number concentration provided by the manufacturer. To ensure that sensitivity was not size-dependent to any significant extent, the calibration curves were measured for both 30 and 60 nm Au nanoparticle sizes.

■

RESULTS AND DISCUSSION Dissolved and Particulate Signals. Typical spICPMS raw data and frequency histograms using 5 ms dwell time are shown in Figure 1. The low intensity dissolved signal and the particle related spike in ultrapure water is due to both impurities and carry-over from previous samples. The signal of the 1 μg L−1 dissolved Ag standard has a RSD of 34% and is relatively symmetrically distributed around a mean close to 100 ion counts. Even in the “fresh” 40 nm AgNP sample (analyzed within one week after delivery), one can recognize a nonzero dissolved signal resembling that of ultrapure water and particle events that appear as intense spikes. The intensities of these particle spikes are lower for the “aged” particle suspension (analyzed 3 months after opening) compared to the fresh suspension, suggesting a particle size decrease while the dissolved signal intensity is increased, probably due to partial dissolution upon exposure of the fresh standard to oxygen. The overlap of nanoparticle and dissolved signal is more pronounced in the aged standard. Figure 1 also shows that a 5*σ threshold value indeed identifies most measured intensities of ultrapure water and 1 μg L−1 dissolved Ag as below the particle detection threshold, whereas the nanoparticle events of the fresh 40 nm AgNP suspension are identified as outliers and are thus retained as nanoparticle events. This approach also demonstrates the difficulty in distinguishing nanoparticles from dissolved signals in the case of overlapping signals as is the case for the aged 40 nm AgNP suspension. Both the fresh and aged 40 nm standard may contain false positives, i.e., signal intensities of dissolved Ag that are above the 5*σ threshold and that are thus erroneously regarded as particle signals, but the proportion of false positives to total detected particle events is likely higher in the aged sample than in the fresh sample. Analogous data for 1 ms dwell time is shown in Figure SI-3, Supporting Information. Effect of Dwell Time. Figure 2 shows frequency−intensity histograms for 80 nm Ag nanoparticles measured using 1, 5, and 10 ms dwell times. While the intensity of the dissolved signal increases with tdwell, the particle signal does not. The overlap between dissolved and nanoparticle signals thus reduces as tdwell is reduced (equations found in Supporting Information). However, the standard deviation of the dissolved signal and the probability that a particle event is only partially measured increase with reducing tdwell, especially when the dwell time is comparable to or lower than the duration of particle event, (ca. 0.1−0.5 ms).32 Shortening tdwell can thus also increase the overlap of dissolved and nanoparticle signals, and an optimum tdwell with a minimal signal overlap therefore exists. Figure 2 shows that 5 ms is in most cases an optimum dwell

Figure 2. Frequency−intensity histograms of 80 nm AgNP measured with dwell times of 1, 5, and 10 ms. The dashed lines indicate the particle detection threshold.

time because increasing tdwell to 10 ms does not increase resolution of the particle signal from the dissolved signal. The broad tail in this sample can be explained by multiple particle events. The overlap in the data with tdwell = 1 ms is high because of a large proportion of incompletely measured particle events. Similar trends were observed for 40 and 60 nm particles (Figures SI-4 and SI-5, Supporting Information). Measurement of Particle Size. The particle sizes were determined by calibrating the ion count with respect to mass of Ag introduced into the plasma using dissolved standards. Provided that the chemical composition is known, this allows estimation of the particle size distribution by assuming a spherical shape for all particles. The calculated diameters are thus corresponding volumetric spherical diameters. Figure 3 shows the calculated size distributions of 20, 40, 60, and 80 nm particles calculated from data obtained using a 5 ms dwell time. The calculated diameters of the 40, 60, and 80 nm AgNP standards are in relatively close agreement with the nominal diameters. Diameter calculation requires the nebulization efficiency, a value that was determined in this study using an indirect determination of which the accuracy is disputed, especially when the nebulization efficiency is relatively low (i.e., < 5%).30 However, it can be shown (Supporting Information) that any error in nebulization efficiency is reduced to its third root when propagated to an error in diameter. Figure 3 shows that a 50% variation of the nebulization efficiency results in ca. ± 10 nm variation of the peak maximum of 80 nm particles. The indirectly measured nebulization efficiencies were 15−20% where the discrepancy with, e.g., directly measured nebulization efficiency, a value that is generally regarded as more accurate, is likely lower than 50%.30 Moreover, direct methods cannot be run at the same time as the spICPMS determination, whereas the nebulization efficiency is highly dependent on sample introduction settings33 that may differ from the conditions during a direct nebulization efficiency determination where the plasma is not ignited. However, the accuracy of the nominal particle sizes was not investigated in this study that was not 3968

dx.doi.org/10.1021/ac203005r | Anal. Chem. 2012, 84, 3965−3972

Analytical Chemistry

Article

deviation rather than the sensitivity of the ICPMS. Further improvement of the detection capabilities with respect to size should thus focus on reducing the dissolved signal intensity and its standard deviation. Since the samples are diluted up to 109 times to avoid measuring multiple particle events, the dissolved background predominantly consists of sample carry-over and not dissolved silver coming from the sample itself. However, if a detectable concentration of dissolved silver from the sample remains, the magnitude of its signal can be reduced by decreasing the sample uptake rate. Reducing the noise of the dissolved signal can be achieved by, e.g., monodisperse droplet delivery32,37 or a Peltier cooled desolvation system37 by which it is possible to partially resolve 20 nm AgNP from the dissolved background, whereas this is not possible with conventional sample introduction.37 Developing algorithms capable of deconvoluting overlapping signals can also contribute to detection capabilities. The theoretical smallest detectable equivalent spherical diameter is 6.1 nm, because this value corresponds to the approximate minimum silver mass needed to deliver at least one 107Ag ion to the detector; given that knowing the mass of silver introduced to the plasma during dwell, it can be calculated that only ca. 1/30 000 of the 107Ag atoms reaches the detector in a typical measurement. Choice of Detection Threshold. The number of particles counted per 10 000 dwells are plotted as a function of the parameter n for 20 and 40 nm AgNP and for a 2 μg L−1 dissolved standard using a 5 ms dwell time (Figure 4; data for 1

Figure 3. The size distributions of the 20, 40, 60, and 80 nm nominal diameter Ag particles determined by spICP-MS. The error bars demonstrate the effect of possible 50% errors on the determination of nebulization efficiency on the calculation of size. A tdwell of 5 ms was used except for the 20 nm AgNP for which it was 1 ms.

aimed at demonstrating the accuracy of particle size determination of spICPMS. Since spICPMS measures the mass of particles that varies with the third power of the diameter, this technique can potentially achieve a very high resolution in diameter. However, studies on single droplets introduced into plasma indicate that, e.g., variation in the point of vaporization and trajectories that lie off the sampler cone axis may introduce variation in the intensity of particle signals.34−36 Since there is no information about the polydispersity of the samples, the size resolution capabilities could not be assessed in this study. As for particle sizing by microscopy, a large number of particles are required for the estimation of the mean diameter with sufficient accuracy, and the sizing capabilities will ultimately be determined by the available sample volume. The distributions of particle event signals of the 20 nm particles were to a large extent overlapping with the dissolved signal although a dwell time of 1 ms was used. Only the high intensity tail corresponding to 20−25 nm volume equivalent spherical diameters could be discriminated. Figure 3 thus confirms that 20 nm AgNP are possible to detect and to some extent possible to discriminate. The size detection limit is for these samples set by the dissolved background and its standard

Figure 4. The number of particle events counted by the iterative algorithm as a function of threshold value (n) for data obtained using a 5 ms dwell time. The following data sets are included: 2 μg L−1 dissolved Ag and 40 and 20 nm AgNP suspensions.

ms dwell time is shown in Figure SI-6, Supporting Information), suggesting that an n value of at least 4 must be used to reduce the number of false positives to less than 0.1% of the total count in the case of the dissolved standard. The particle count should thus reach a value that is independent of n if n > 4, in the absence of false positives. This behavior is approached for the 40 nm particles measured at 5 ms dwell time where the particle events are completely resolved from the dissolved signal (Figure 3). In the case of 40 nm particles measured at tdwell = 1 ms and for all 20 nm data, the particle events are not resolved from the background and it can be seen that the particle count continues to decrease upon increasing n. Increasingly more true particle counts are removed together 3969

dx.doi.org/10.1021/ac203005r | Anal. Chem. 2012, 84, 3965−3972

Analytical Chemistry

Article

depends on the determination of nebulization efficiency, and the ability to discriminate particle events from the dissolved signal. Although the number concentrations were not confirmed in this study, Figure 5 demonstrates the dynamic range and linear concentration dependent response of spICPMS particle counting. The contribution of false positives is significant when the particle concentration is low and the particle number is consequently overestimated. The frequency of false positives thus sets the limit of detection with respect to concentration when present. At higher concentrations, the counting yield decreases due to the occurrence of multiple particle events that are each counted as one particle event and a downward slope can be observed. This behavior is observed regardless of size or dwell time for the particles investigated here. A plateau corresponding to the linear response to concentration is obtained over ca. 2 orders of magnitude of particle number concentrations, even if particle events are not resolved from the background, as is the case for 20 nm AgNP and measurements using 1 ms dwell time. This suggests that the iterative algorithm may also be successful if the nanoparticle and dissolved signal intensities partly overlap. However, the particle count is less accurate in this case because a fraction of the true particle events are not counted as such. Moreover, any error in nebulization efficiency is also propagated to the particle count, and the effective dwell time in eq 2 is not necessarily equal to the nominal tdwell (further discussion and illustration in Figure SI-1, Supporting Information).38 Effect of Number of Dwells. Figure 6 shows how the linear range of particle number determination depends on

with false positives in this case because of the overlap of particle and dissolved signals. It is evident that to reduce the number of false positives to a predefined acceptable level, e.g., 0.1% of the total count, n values considerably higher than 3 must be used to reach this level over a broad concentration range. However, when very high values for n are used, particles are omitted from counting. For the remainder of the discussion, n = 5 was therefore chosen as a compromise between retaining a sufficiently high particle count over a broad concentration range and ensuring that contribution of false positives is less than 0.1% of the actual particle events. (Particle count over a range of concentrations are compared for n values of 3, 5, and 8 in Figure SI-7, Supporting Information.) Concentration Calibration Curves. The logarithm of particle count divided by concentration is plotted as a function of the logarithm of concentration for the 20, 40, 60, and 80 nm AgNP measured using 1 and 5 ms dwell time in Figure 5. The

Figure 6. The logarithm of counted particle events divided by the concentration as a function of the logarithm of concentration for 40 nm AgNP measured using a 5 ms dwell and 5*σ detection threshold. The number of data points used was varied by randomly selecting from the 40 000 measured data points for each sample.

Figure 5. The logarithm of counted particle events divided by the concentration as a function of the logarithm of concentration. The data obtained for 20, 40, 60, and 80 nm Ag particles using 1 ms (left) and 5 ms (right) dwell time and 5*σ detection threshold is shown. 10 000 dwells were used for each sample. The boxes indicate the region where the relative frequency of false positives (solid line) and multiple particle events (dashed line) becomes significant.

randomly selecting smaller data point sets among the 40 000 measured dwells. When D = 1000, only one or no particles are counted for the first 7 data points and the true particle number is increasingly overestimated at lower concentrations, because the number of false positives becomes more significant relative to the total count. Moreover, there is more scattering in data obtained from smaller data sets, because of higher statistical fluctuations. Increasing D to 5000 reduces this scatter and a linear response is obtained over at least 2 orders of magnitude, but the scatter in data is still significantly higher than for D >

particles were counted with a 5*σ detection threshold. The data points in such plots are on horizontal lines in the particle concentration range where true particle counts dominate over false positives and a linear response of the experimental particle count therefore occurs. The level of these lines depends on the measured particle number concentration, whose accuracy 3970

dx.doi.org/10.1021/ac203005r | Anal. Chem. 2012, 84, 3965−3972

Analytical Chemistry

Article

applications, e.g., additive to diesel fuel to improve combustion efficiency, support in the automobile catalysts, and as chemicalmechanical polishing in certain applications, e.g., electronics industry. The Gothenburg wastewater system receives road runoff and industrial and domestic wastewater, so the Ce particles may originate from man-made sources. TiO2 nanoparticles are used in a large number of applications from sunscreens and other cosmetics to paint.43 The predicted environmental concentration in wastewater effluents for TiO2 in Europe is 3.5 μgL−139 leading to 1 975 300, 246 920, and 73 160 particles mL−1, assuming spherical 20, 40, or 60 nm sizes, respectively; values that are again comparable to the measured 32 656 mL−1, considering the large uncertainty on modeled values. However, natural sources cannot be ruled out for both Ce- and Ti-containing nanoparticles. Both brookite TiO2 and Ce bearing allanite, for instance, have been found in river sediments.11

10 000. Figure 6 thus shows that the concentration detection limit can be decreased by increasing D. However, the extent by which this limit is reduced is lower at higher D values as Figure 6 shows that increasing D from 20 000 to 40 000 provides much less reduction in scatter and the proportion of false positives. The results agree with eq 3. Wastewater Effluent Samples. Of the screened elements (Ag, Ce, Ti, Si, Zn, Cr, Cu, Mo, Pt, Sb, W, Y, and Zr), only the Ag, Ce, and Ti analyses demonstrated particle events (Figure 7). The filtration through the 450 nm nominal cutoff filters sets

■

CONCLUSIONS AND OUTLOOK AgNP spICPMS signal intensities can be resolved from dissolved signal intensities when the size exceeds ca. 20 nm and the dwell time exceeds 1 ms. When these signal intensities are not fully resolved, the particle number determination is less accurate. Further improvement of the detection limit can be achieved by reducing the dissolved signal intensity and its noise, e.g., using monodisperse drop delivery.32 The choice of a proper detection threshold is important for the quantification of the particle number concentration. Using n values of ca. 5 compromises between a relatively low frequency of false positives and a relatively high effective particle count. The lowest detectable concentration is determined by the number of acquired dwells but also by the relative frequency of false positives. The volume equivalent spherical diameters calculated from the signal intensities using calibration with dissolved standards agree with the nominal diameters of the particles, suggesting that spICPMS can be used as a fast and sensitive particle sizing method for environmental samples. The detection of particles in an environmental sample was demonstrated. The shortest possible tdwell of 0.1 ms was used in order to minimize the size detection limit. Increasing the tdwell increases sensitivity with respect to concentration and allows more accurate determination of the mass of analyzed element in the particle and thus its size. With the instrumentation used here, it is only possible to measure a single mass within the duration of a particle event which does not allow determination of the composition of individual particles. This limits the applicability of the method for unknown samples, where it would be interesting to adapt the concepts developed in this paper for instruments capable of simultaneous multielement detection.44 Such a method could, for instance, detect whether the detected particles are in fact AgNP or whether these are transformed into Ag2S nanoparticles.

Figure 7. Time resolved signals of spICPMS for wastewater effluent samples. The monitored elements were Ti, Ce, and Ag.

an upper limit on the particle sizes. Gold particle reference materials were used to estimate the particle number concentrations (Figure SI-8, Supporting Information). The dwell time of 0.1 ms does not allow accurate sizing of the particles. The numbers of particle events were low and therefore a 8*σ detection threshold was chosen to obtain a sufficiently low frequency of false positives. The concentrations were 9568 particles mL−1 for Ag, 2312 particles mL−1 for Ce, and 32 656 particles mL−1 for Ti. The predicted environmental nano-Ag concentration in wastewater effluents for nano-Ag in Europe is 42 ng L−1.39 Assuming a spherical particle size of 20, 40, and 60 nm, this predicted concentration corresponds to 95 493, 11 937, 354 particles mL−1, suggesting that the measured particle number concentration in WWTP is in the order of magnitude of the predicted concentration. It is unknown to what extent AgNP are being emitted as particles or whether they in fact occur as dissolved ions in applications such as food packaging plastics, kitchen utensils and surfaces, textiles, washing processes, etc., but recent investigations suggest that both silver forms are reprecipitated as nanoscale silver sulfide (Ag2S) particles in wastewater treatment plants.40−42 spICPMS analysis does not allow one to verify whether the detected silver particles are elemental silver nanoparticles or silver sulfide, which in this case would also impede size estimation. Manufactured CeO2 nanoparticles are used in a number of

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel.: +46-31-786 9050. Fax: +46-31-77227856. 3971

dx.doi.org/10.1021/ac203005r | Anal. Chem. 2012, 84, 3965−3972

Analytical Chemistry

Article

Notes

(26) Mitrano, D. M.; Lesher, E. K.; Bednar, A.; Monserud, J.; Higgins, C. P.; Ranville, J. F. Environ. Toxicol. Chem. 2011, 31, 115− 121. (27) Allabashi, R.; Stach, W.; de la Escosura-Muñiz, A.; Liste-Calleja, L.; Merkoçi, A. J. Nanopart. Res. 2009, 11, 2003−2011. (28) Krystek, P.; Ulrich, A.; Garcia, C. C.; Manohar, S.; Ritsema, R. J. Anal. At. Spectrom. 2011, 26, 1701−1721. (29) Levine, D. M.; Ramsey, P. P.; Smidt, R. K. Applied Statistics for Engineers and Scientists; Prentice Hall: Upper Saddle River, NJ, 2001. (30) Smith, D.; Browner, R. Anal. Chem. 1982, 54, 533−537. (31) Gustavsson, A. Spectrochim. Acta 1986, 41B, 291−294. (32) Gschwind, S.; Flamigni, L.; Koch, J.; Borovinskaya, O.; Groh, S.; Niemax, K.; Gunther, D. J. Anal. At. Spectrom. 2011, 26, 1166−1174. (33) Sharp, B. L. J. Anal. At. Spectrom. 1988, 3, 939−963. (34) Dziewatkoski, M. P.; Daniels, L. B.; Olesik, J. W. Anal. Chem. 1996, 68, 1101−1109. (35) Olesik, J. W.; Kinzer, J. A.; McGowan, G. J. Appl. Spectrosc. 1997, 51, 607−616. (36) Stewart, I. I.; Hensman, C. E.; Olesik., J. W. Appl. Spectrosc. 2000, 54, 164−174. (37) Franze, B.; Stenge, I.; Engelhard, C. J. Anal. At. Spectrom. 2012, DOI: 10.1039/C2JA00003B. (38) Heithmar, E. M. Abstracts of papers of the American Chemical Society, 237th ACS National Meeting, Salt Lake City, UT USA, 2009; Vol. 237, pp 58−COLL. (39) Gottschalk, F.; Sonderer, T.; Scholz, R. W.; Nowack, B. Environ. Sci. Technol. 2009, 43, 9216−9222. (40) Choi, O.; Deng, K. K.; Kim, N. J.; Ross, L.; Surampalli, R. Y.; Hu, Z. Q. Water Res. 2008, 42, 3066−3074. (41) Kaegi, R.; Voegelin, A.; Sinnet, B.; Zuleeg, S.; Hagendorfer, H.; Burkhardt, M.; Siegrist, H. Environ. Sci. Technol. 2011, 45, 3902−3908. (42) Kim, B.; Park, C. S.; Murayama, M.; Hochella, M. F. Environ. Sci. Technol. 2010, 44, 7509−7514. (43) Kaegi, R.; Ulrich, A.; Sinnet, B.; Vonbank, R.; Wichser, A.; Zuleeg, S.; Simmler, H.; Brunner, S.; Vonmont, H.; Burkhardt, M.; Boller, M. Environ. Pollut. 2008, 156, 233−239. (44) Schilling, G. D; Andrade, F. J.; Barnes, J. H.; Sperline, R. P.; Denton, M. B.; Barinaga, C. J.; koppennaal, D. W.; Hieftje, G. M. Anal. Chem. 2007, 79, 7662−7668.

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank the following funding agencies for support: CEFIC Long range research initiative, European Commission FP7 projects NANOFATE and MARINA, the Swedish Environmental Research Council FORMAS, and the Gothenburg graduate school Environment and Health. We thank Dr. Nicklas Paxeus at GRYAAB for help with the wastewater sampling.

■

REFERENCES

(1) Roco, M. C. J. Nanopart. Res. 2011, 13, 427−445. (2) Klaine, S. J.; Alvarez, P. J. J.; Batley, G. E.; Fernandes, T. F.; Handy, R. D.; Lyon, D. Y.; Mahendra, S.; McLaughlin, M. J.; Lead, J. R. Environ. Toxicol. Chem. 2008, 27, 1825−1851. (3) Gottschalk, F.; Sonderer, T.; Scholz, R. W.; Nowack, B. Environ. Sci. Technol. 2010, 43, 9216−9222. (4) Arvidsson, R.; Molander, S.; Sanden, B. A.; Hassellov, M. Hum. Ecol. Risk Assess. 2011, 17, 245−262. (5) Paterson, G.; Macken, A.; Thomas, K. V. Anal. Methods 2011, 3, 1461−1467. (6) Gallego-Urrea, J. A.; Tuoriniemi, J.; Pallander, T.; Hassellov, M. Environ. Chem. 2010, 7, 67−81. (7) Howard, A. G. J. Environ. Monit. 2010, 12, 135−142. (8) Hassellöv, H.; Kaegi, R. In Environmental and Human Health Impacts of Nanotechnology; Lead, J. R., Smith, E., Eds.; John Wiley & Sons, Ltd: Chichester, U.K., 2009. (9) Von der Kammer, F.; Ferguson, P. L.; Holden, P. A.; Masion, A.; Roger, K. R.; Klaine, S. J.; Koelmans, A. A.; Horne, N.; Unrine, J. M. Environ. Toxicol. Chem. 2011, 31, 32−49. (10) Kaegi, R.; Wagner, T.; Hetzer, B.; Sinnet, B.; Tzuetkov, G.; Boller, M. Water Res. 2008, 42, 2778−2786. (11) Plathe, K. L.; von der Kammer, F.; Hassellov, M.; Moore, J.; Murayama, M.; Hofmann, T.; Hochella, M. F. Environ. Chem. 2010, 7, 82−93. (12) Degueldre, C.; Favarger, P. Y. Colloids Surf., A 2003, 217, 137− 142. (13) Degueldre, C.; Favarger, P. Y. Talanta 2004, 62, 1051−1054. (14) Degueldre, C.; Favarger, P. Y.; Bitea, C. Anal. Chim. Acta 2004, 518, 137−142. (15) Degueldre, C.; Favarger, P. Y.; Wold, S. Anal. Chim. Acta 2006, 555, 263−268. (16) Degueldre, C.; Favarger, P. Y.; Wold, S. Talanta 2006, 68, 623− 628. (17) Luoma, S. In Project on emerging nanotechnologies; Woodrow Wilson international centre for scholars, The PEW charitable trusts, 2008. (18) Benn, T. M.; Westerhoff, P. Environ. Sci. Technol. 2008, 42, 4133−4139. (19) Geranio, L.; Heuberger, M.; Nowack, B. Environ. Sci. Technol. 2009, 43, 8113−8118. (20) Kaegi, R.; Sinnet, B.; Zuleeg, S.; Hagendorfer, H.; Mueller, E.; Vonbank, R.; Boller, M.; Burkhardt, M. Environ. Pollut. 2010, 158, 2900−2905. (21) Benn, T.; Cavanagh, B.; Hristovski, K.; Posner, J. D.; Westerhoff, P. J. Environ. Qual. 2010, 39, 1875−1882. (22) Farkas, J.; Peter, H.; Christian, P.; Gallego-Urrea, J. A.; Hassellov, M.; Tuoriniemi, J.; Gustafsson, S.; Olsson, E.; Hylland, K. Environ. Int. 2011, 37, 1057−1062. (23) Laborda, F.; Jimenez-Lamana, J.; Bolea, E.; Castillo, J. R. J. Anal. At. Spectrom. 2011, 26, 1362−1371. (24) Aruguete, D. M.; Liu, J.; Hochella, M. F., Jr. In Environmental and Human Health Impacts of Nanotechnology; Blackwell Publishing, Ltd.: West Sussex, United Kingdom, 2009; pp 79−108. (25) Auffan, M.; Rose, J.; Bottero, J.-Y.; Lowry, G. V.; Jolivet, J.-P.; Wiesner, M. R. Nat. Nanotechnol. 2009, 3, 634−641. 3972

dx.doi.org/10.1021/ac203005r | Anal. Chem. 2012, 84, 3965−3972