Single Particle Inductively Coupled Plasma-Mass Spectrometry: A

Jul 10, 2012 - Single Particle Inductively Coupled Plasma-Mass Spectrometry: A Performance Evaluation and Method Comparison in the Determination of Na...
0 downloads 5 Views 2MB Size
Article pubs.acs.org/est

Single Particle Inductively Coupled Plasma-Mass Spectrometry: A Performance Evaluation and Method Comparison in the Determination of Nanoparticle Size Heather E. Pace,†,§ Nicola J. Rogers,† Chad Jarolimek,† Victoria A. Coleman,‡ Evan P. Gray,§ Christopher P. Higgins,§ and James F. Ranville*,∥ †

Nanosafety in the Environment, CSIRO Land and Water, Lucas Heights, NSW 2234 Australia Nanometrology Section, National Measurement Institute, West Lindfield, NSW 2070 Australia § Civil and Environmental Engineering, Colorado School of Mines, Golden, Colorado 80401, United States ∥ Chemistry and Geochemistry, Colorado School of Mines, Golden, Colorado 80401, United States ‡

S Supporting Information *

ABSTRACT: Sizing engineered nanoparticles in simple, laboratory systems is now a robust field of science; however, application of available techniques to more complex, natural systems is hindered by numerous challenges including low nanoparticle number concentrations, polydispersity from aggregation and/or dissolution, and interference from other incidental particulates. A new emerging technique, single particle inductively coupled plasma-mass spectrometry (spICPMS), has the potential to address many of these analytical challenges when sizing inorganic nanoparticles in environmental matrices. However, to date, there is little beyond the initial feasibility studies that investigates the performance characteristics and validation of spICPMS as a nanoparticle sizing technique. This study compares sizing of four silver nanoparticle dispersions (nominal diameters of 40, 60, 80, and 100 nm) by spICPMS to four established sizing techniques: dynamic light scattering, differential centrifugal sedimentation, nanoparticle tracking analysis, and TEM. Results show that spICPMS is able to size silver nanoparticles, across different sizes and particle number concentrations, with accuracy similar to the other commercially available techniques. Furthermore, a novel approach to evaluating particle coincidence is presented. In addition, spICPMS size measurements were successfully performed on nanoparticles suspended in algal growth media at low concentrations. Overall, while further development of the technique is needed, spICPMS yields important advantages over other techniques when sizing nanoparticles in environmentally relevant media.



INTRODUCTION

studies, as it is the core distinction between nanoscale materials and bulk materials. Unfortunately, characterization of nanoparticle size for toxicity testing is complicated by the fact that size is not constant among different matrices and concentrations, or even across time in the same matrix.11 As a result, understanding the relationship between nanosized particles and biological function requires not only characterizing the particles as they are received from the manufacturer, but also characterizing them across the same time frame and under the same set of experimental conditions and concentrations present during the biological exposure.12 Implicit in this is the requirement that

The potential risks of nanotechnology have received significant attention in recent years; however, linking observed toxic effects to specific nanoparticle characteristics has proven to be complex. Beyond composition, various other physicochemical characteristics have been shown to potentially influence how nanoparticles interact with and affect biological systems.1,2 Particle size, which includes size distribution and aggregation state, is argued to be one of the most important properties to quantify when conducting toxicity tests.1,3,4 For example, size appears to influence key processes such as uptake, distribution, and clearance from a cell or organism.5,6 There is also evidence that the toxicity of certain nanomaterials can potentially be sizedependent,7,8 although there are other studies that suggest otherwise.9,10 Regardless, particle size will continue to be an important property in the interpretation of nanotoxicology © 2012 American Chemical Society

Received: Revised: Accepted: Published: 12272

September 15, 2011 June 25, 2012 July 10, 2012 July 10, 2012 dx.doi.org/10.1021/es301787d | Environ. Sci. Technol. 2012, 46, 12272−12280

Environmental Science & Technology

Article

Specifically, the sizing capabilities of spICPMS were directly compared to those of four other commercially available nanoparticle sizing techniques. These techniques were dynamic light scattering (DLS), differential centrifugal sedimentation (DCS), nanoparticle tracking analysis (NTA), and transmission electron microscopy (TEM). This paper represents the first comparison of spICPMS to DSC, DLS, and NTA. In addition, the various performance characteristics of the method were explored across a range of particle sizes and concentrations. Special consideration was given to defining a new methodology for detecting particle coincidence based on both statistical and analytical approaches. Finally, the sizing capabilities of spICPMS were demonstrated on nanoparticles in algal growth media supplemented with dissolved organic matter. Ultimately, the goal of this study is to further establish spICPMS as an important new sizing technique in the area of nanotoxicology research.

any sizing technique employed must be capable of providing robust data for nanoparticles in complex matrices, at low, environmentally relevant concentrations. While there are a number of analytical techniques available for the determination of particle size at the nanometer scale, many of these techniques are not appropriate for environmental matrices. This is primarily because nanoparticles in environmental matrices are often at concentrations below the detection limits of many available sizing techniques. For instance, Domingos et al.13 conducted a method comparison of several sizing techniques used in nanotoxicology research and found that most of the techniques studied needed concentrations well above 1 mg/L to accurately measure size. Furthermore, in a review, Hasselöv et al.14 summarized the typical detection limits of the various available sizing techniques, showing that most sizing techniques had detection limits in the mg/L range, with few below the μg/L range. When compared to predicted environmental concentrations, which are in the subμg/L range,15 there is an obvious gap in our current ability to characterize nanoparticles at environmentally relevant concentrations. Beyond method sensitivity limitations, many current techniques have been shown to be ill suited to characterize polydisperse systems that contain aggregates and other incidental particles and debris. Polydisperse systems are known to be particularly debilitating to light scattering techniques, which are among the most popular techniques used to size nanoparticles in toxicology studies.13 In addition, heterogeneous samples present challenges for many techniques, such as light scattering and sedimentation, which assume homogeneous material properties. In such methods, particles of different compositions are not easily distinguishable. Thus, incidental and/or natural particulates can interfere with measurements of the target nanoparticle. Separation or fractionation can help remove interfering particles; however, without additional selectivity for the target nanoparticle, interference within environmental matrices is still highly likely. Another popular technique, transmission electron microscopy (TEM), is a highly selective method with little encumbrance from particle polydispersity. Yet, high operational costs, timeintensive sampling, extensive sample preparation needs, and the resulting sample preparation artifacts complicate the routine application of the technique. Consequently, the development of improved analytical techniques for characterization of nanoparticles in complex matrices typical of environmental and biological samples has become a recognized research priority in the area of nanotoxicology. Single particle inductively coupled plasma-mass spectrometry (spICPMS) is an emerging analytical technique that has the potential of addressing some of the major challenges faced by investigators when analyzing for particle size within toxicity tests. While the focus of this paper will be on the particle sizing aspect of the technique, it is important to point out that both particle number concentration and particle size can be determined simultaneously using spICPMS.16 Recently, the authors of this study presented a new protocol for finding particle size.17 This new protocol uses transport efficiency and dissolved standards to relate pulse height to particle size. This is in lieu of creating a size calibration curve from monodisperse, reference nanoparticles, a method first presented by Degueldre et al.18 The objective of this study was to assess the sizing capabilities of spICPMS using the newly developed protocol.



MATERIALS AND METHODS Stock Nanoparticle Dispersions. Silver nanoparticle dispersions of four different sizes (nominal sizes of 40, 60, 80, and 100 nm as reported by the manufacturer) were purchased from NanoComposix (USA). Stock particle dispersions came in an aqueous matrix with a 2 mM phosphate buffer. Transmission electron micrographs provided by the manufacturer showed near-spherical geometry for all particle sizes. Concentrations of the stock dispersions were reported by the manufacturer as 20 mg/L Ag for the 60 and 100 nm particle dispersions (NanoXact) and 1000 mg/L Ag for the 40 and 80 nm dispersions (Biopure). In addition, gold reference nanoparticle dispersions with a nominal diameter of 60 nm were purchased from the U.S. National Institute of Standards and Technology (NIST, USA RM8013). The NIST 60 nm gold reference nanoparticles have a certified stock concentration of 51.86 ± 0.64 mg/L Au and are measured by several sizing techniques with average diameters ranging from 53.2 to 56.6 nm. These particles were used to determine transport efficiency for the purposes of calculating particle size of the silver nanoparticles.17 Finally, a 100 nm gold nanoparticle (NanoComposix, USA) was used to examine the effects of particle concentration on particle coincidence. Particle stocks were dispersed in excess citrate and the reported concentration was 50 mg/L. Due to inherent sensitivities of each technique, all samples were analyzed at a concentration within the appropriate dynamic range for the technique. Single Particle ICP-MS. An Agilent 7500 CE ICP-MS with a quadrapole mass analyzer, a Micromist nebulizer, and a Scott Double Pass spray chamber (Agilent Technologies, CA, USA) was used in all analyses except the gold nanoparticle analyses performed for the investigation of particle coincidence (data discussed in the Supporting Information (SI)). For these methods a PE NexION 300 Q was used (operating conditions are described in the SI). The data acquisition for the Agilent 7500 was set to time-resolved analysis (TRA) mode, and the measurement duration of each run was 30 s with a dwell time of 10 ms (i.e., integration time for each reading), which equated to 2884 total readings. Using a 10 ms dwell time is a compromise between a longer dwell time, which increases the probability of coincidence, and a shorter dwell time, which could result in the partial capture of the particle ion cloud.19 The instrument was tuned daily using a multielement tune solution (1 ug/L Li, Co, Y, Tl, Ce, and Ba in 1% v/v hydrochloric acid) for optimal sensitivity and minimum oxide and doubly charged species 12273

dx.doi.org/10.1021/es301787d | Environ. Sci. Technol. 2012, 46, 12272−12280

Environmental Science & Technology

Article

particle size distribution, using Mie theory, are given in the SI (Table S-1). Dynamic Light Scattering. DLS analyses of nanoparticle suspensions were obtained using a Malvern Instruments Zetasizer Nano ZS (4 mW He−Ne laser at 633 nm, Malvern Instruments Ltd., Worcestershire, UK). The average particle diameter was reported from the z-average hydrodynamic diameter. The samples used for DLS measurements had nanoparticle concentration equivalent to 20 mg/L Ag. The standard deviation was calculated from triplicate measurements. More detail on the parameters used in DLS measurements is presented in the SI (Table S-1). Nanoparticle Tracking Analysis. NTA measurements were made using a NanoSight LM10 Instrument (NanoSight Ltd., Wiltshire, UK; operating software: NTA V2.1; Marlin F033B ASG camera), operating at 20 °C. Particles were visualized via scattered light from a red laser (633 nm), and sized based on their tracked path length. Similar to spICPMS, NTA is a single particle method where particle size distributions are built from many individually sized particles. Particle size distributions were presented as histograms with 1 nm bins and particle size was reported as both the arithmetic mean of all the nanoparticle sizes as well as the mode from the histogram. Standard deviations were calculated from three replicate measurements made using concentrations of 1 × 1011 to 1 × 1012 particles/L. spICPMS Analysis in Environmental Media. In a second set of experiments, a limited stability study in various environmental media was conducted. For these experiments, the size of the 100 nm Ag nanoparticles was assessed, using spICPMS analysis, at 0 and 144 h following dilution. Dilutions were made with various environmentally relevant media including algal growth medium (AGM) and AGM supplemented with two concentrations (3 and 10 mg/L) of Suwannee River humic acids (HA) (see SI for AGM details). In addition, dispersions in ultrapure water were included for a matrix effect control and 2% trace pure nitric acid dilutions were included as a positive control for nanoparticle dissolution. For each of the media types, nanoparticles were diluted, in triplicate, to three concentrations: 1000, 50, and 5 μg/L. Immediately before spICPMS analysis (1−2 min), all samples were diluted to a concentration of 0.5 μg/L, so that the concentration of the samples fell within the linear dynamic range of the method. All nanoparticle samples were kept in the dark at 20 °C and were undisturbed (i.e., no mixing) during the 144 h incubation period.

levels. A calibration curve, used in the sizing of silver nanoparticles, was produced using dissolved silver standards (AccuTrace, CT, USA) prepared in 0.2% trace pure nitric acid (Merck, Darmstadt, Germany). The peristaltic pump was set to 0.05 rps for all experiments, which translated to a sample flow rate of approximately 0.16−0.18 mL/min. The average flow rate was measured each day by weighing triplicate vials of ultrapure water before and after 1 min of sample aspiration. During spICPMS analysis the isotopes 107Ag and 197Au were monitored for the silver and gold samples, respectively. Data, in the form of counts/10 ms as a function of time, were exported to a spreadsheet for further processing. Particle size from spICPMS analysis was calculated using the protocol described in Pace et al.17 Likewise, transport efficiency was determined by the previously described “Particle Frequency Method” using the 60 nm gold reference nanoparticles from NIST. The particle number concentration of the 60 nm gold stock solution was calculated using the total gold concentration, the average diameter, and the density of gold. Since NIST reports average diameters ranging from 53.2 to 56.6 nm (depending on the method used), the median of the reported average values, 55 nm, was used for all calculations. As a single particle technique, the particle size distribution in spICPMS analysis is determined by sizing individual particles and then binning them to create a histogram. To compare across the different techniques, the diameter is reported as the mode, which was then averaged across three replicates. For spICPMS, the mode is determined from a particle size histogram, constructed using 1 nm diameter bins, for each individual run. This is similar to the process NTA uses to represent modal particle diameter. The standard deviation was then calculated from triplicate runs. Please see the SI for a brief overview of the sizing equations. As mentioned above, in the first set of experiments, the sizing capabilities by spICPMS analysis was compared to measurements made using three other established sizing methods. For these experiments, spICPMS analysis was done using nanoparticle stock dispersions diluted, using ultrapure (18.2 MΩ) water, to nominal concentrations ranging from 0.005 to 1 μg/L Ag, based on the manufacturer’s reported mass concentrations. Dilutions were performed in triplicate and carried out on the same day as ICP-MS analysis to minimize nanoparticle dissolution and/or aggregation in the samples. Details on the other three sizing methods are described below. Differential Centrifugal Sedimentation. DCS measurements were conducted on a disk centrifuge (CPS Instruments, model UHR24 000, operating software CPSV95; detection wavelength 405 nm). The instrument’s performance (i.e., accuracy of centrifugation speed, density of prepared gradient) was checked immediately before each measurement with a 100 nm Au nanoparticle (mean diameter, 0.0998 μm, BB International, UK) dispersed in water. The disk centrifuge was prepared with 17 mL of an 8−24% w/w density gradient solution of sucrose (Sigma Aldrich) in ultrapure water. Dodecane (0.5 mL) was used as a gradient evaporation barrier during measurement. The samples were measured at a rotation speed of 24 000 rpm. After reaching the requisite speed, the disk was allowed to equilibrate for 1 h before measurement. For each sample measurement, a 0.1 mL aliquot of 20 mg/L Ag nanoparticles was injected into the DCS. Triplicate measurements were performed on each sample. The physical properties used for input into the Stokes equation and for conversion of the measured absorption curves to the reported



RESULTS AND DISCUSSION The performance of spICPMS as a particle sizing technique was judged using three other commercially available sizing methods. Additionally the spICPMS data were compared to the TEM data provided by the manufacturer. The main criteria for comparison to the other techniques were the degree of agreement for the measured nanoparticle size and size distribution, the reproducibility of the measurement, and the breadth of the generated size distribution (i.e., peak width). Detection limit, sensitivity, and dynamic range of the spICPMS method were also explored and discussed. Comparability among Methods. It is recognized that the assorted particle sizing techniques rely on different fundamental principles in determining diameter. This has led to anticipated discrepancies between different techniques in the measured particle size.13,20,21 Because of this, a single true particle 12274

dx.doi.org/10.1021/es301787d | Environ. Sci. Technol. 2012, 46, 12272−12280

Environmental Science & Technology

Article

Table 1. Comparison of Average Particle Size (± std) Measured by spICPMS to Four Other Commercially Available Sizing Techniques method

nanoparticle diameter (nm)

nominal size single particle ICP-MS differential centrifugal sedimentation dynamic light scattering nanoparticle tracking analysis tem (Manufacturer)

mode peak z-avg mode mean

40

60

80

100

37.0 ± 2.7 39.4 ± 0.6 45.6 ± 0.7 50.3 ± 3.5 41.9

52.7 ± 2.1 57.9 ± 0.6 67.2 ± 0.9 67.5 ± 4.4 61.2

74.0 ± 2.7 72.8 ± 0.4 89.1 ± 1.6 87.3 ± 6.1 79.6

80.3 ± 0.8/102.3 ± 1.5a 91.3 ± 0.6/109.7 ± 0.8a 95.1 ± 1.0 89.0 ± 4.7 98.3

a

Bimodal size distribution observed. First diameter represents the mode of the taller, primary peak, and the second diameter represents the mode of the smaller, secondary peak.

Figure 1. Comparison of particle size distributions generated by single particle inductively coupled plasma-mass spectrometry (spICPMS) versus differential centrifugal sedimentation (DCS) and nanoparticle tracking analysis (NTA) for 40, 60, 80, and 100 nm silver nanoparticles. For presentation purposes spICPMS histogram is presented using 5 nm bin widths. All size distributions are scaled relative to their maxima on the y-axis to enable comparisons among the methods. The 40, 60, 80, and 100 nm spICPMS histograms were produced by combining three replicates runs at 0.005, 0.01, 0.1, and 0.5 μg/L Ag, respectively.

diameter is not easy to establish, even for monodisperse, reference nanoparticle dispersions such as those provided by NIST. Consequently, assessing the accuracy of spICPMS (i.e., comparison to a true, reference value) is impractical. Instead, the approach taken in this study is to compare sizing results across several different techniques, and provide a qualitative assessment of method “accuracy” by placing the sizing capabilities of spICPMS in context with other established sizing techniques. Furthermore, comparison to an alternative, established technique is a recognized form of method validation by many organizations.22 Average Size. A summary of the results from the four different particle sizing methods used in the present study is shown in Table 1. While general agreement of the average particle size was observed across all techniques, the average size measurements from DCS demonstrated better consensus with spICPMS measurements as compared to DLS and NTA. Likewise, DLS and NTA demonstrated greater agreement with each other compared to the other two techniques. Given the discussion above, the variation in the measured average particle size observed among the four techniques is not surprising, and can likely be explained by the differences in measurement being

made by each technique. For instance, DLS and NTA both measure diffusion coefficients to determine particle diameter. Diffusion coefficients correlate to the apparent size of dynamic hydrated/solvated particles, which typically increases the apparent diameter. On the other hand, spICPMS measures the mass corresponding to the inorganic core, with no reference to surface coating or shape effects. It can then be expected that DLS and NTA might consistently report larger average particle diameters than spICPMS. Although as stated previously, even for a standard material there is not a “true” size, but for monodisperse metal NPs characterization by TEM is generally considered a close approximation. The manufacturer-stated sizes obtained by TEM are also presented in Table 1. Data suggest that all methods produce an average size within a range of 10−20%. Precision. The intraday reproducibility (standard deviation in Table 1, n = 3) of the spICPMS method was less precise than either the DCS and DLS methods. However, spICPMS precision was similar to that of NTA. Both spICPMS and NTA are single particle detection techniques, and determine particle size by building histograms from individually sized nanoparticles. On the other hand, DCS and DLS are ensemble 12275

dx.doi.org/10.1021/es301787d | Environ. Sci. Technol. 2012, 46, 12272−12280

Environmental Science & Technology

Article

area and this approach is likely to become available in the next few years. Assessing the “accuracy” of the spICPMS generated size distributions is complicated by the same uncertainty as previously described for average size (i.e., there is no “true” size distribution). Rather we again rely on comparability to the manufacturers supplied TEM distributions (Figure S-4). Both spICPMS and DCS gave size distributions (Figure 1) that were most similar to the TEM data (Figure S-4) provided from the manufacturer with 5 nm bin sizes. Both methods captured the fact that the 40 and 80 nm particles had the narrowest and broadest distributions respectively. Both NTA and DLS size distributions were considerably broader. Size Detection Limit. For spICPMS analysis, the size detection limit is defined as the smallest detectable particle size. A particle can only be detected if the pulse from the particle can be distinguished from the noise of the background signal (i.e., the dissolved metal concentration in solution). In other words, the theoretical limit of detection can be determined by finding the value three standard deviations above the average background intensity and calculating the corresponding particle size for that intensity. In the present study, the background signal of the silver nanoparticle samples was typically near blank levels, with averages less than 1 count/10 ms and standard deviations below that. This allowed pulses as small as 3 counts/ 10 ms to be distinguished above the background signal. Using the same conditions and calibration curve as were used for the spICPMS data presented in Figure 1, a pulse of 3 counts/10 ms translates to a particle diameter of 32 nm. This explains why the 40 nm size distribution shown in Figure 1 is truncated at the 30−35 nm bin. The size detection limit depends on the instrument sensitivity as well as the signal-to-noise ratio. As shown in Figure S-2, the size distributions of the 40 nm particles differ slightly for two different days of spICPMS analysis. This difference can be attributed to an increase in the instrument sensitivity (i.e., number of counts generated per atom reaching the plasma) achieved on the second day. This is demonstrated by the inlaid calibration curves showing a 30% increase in the instrument response (i.e., a slope of 170 versus 130) on the second day. Given that the background signal still averaged around 1 count/10 ms, this increase in instrument response lead to a significantly lower detection limit. In this case, a detection limit of 20 nm for a pulse of 3 counts/10 ms was achieved, which made the difference between capturing only part of the 40 nm size distribution versus obtaining almost complete separation between the particle distribution and the background. A 20 nm detection limit for silver nanoparticles is close to the 18 nm detection limit recently reported by Laborda et al.29 The fluctuations seen in Figure S-2 are well within the acceptable operating conditions of the ICP-MS, which highlights the impact even small changes in instrument response can have on size detection limit when using spICPMS for nanoparticle sizing. Another factor that can affect the size detection limit is an increase in the background signal. This increase is usually the result of dissolved metals in solution. The higher background itself does not necessarily increase the size detection limit, as the average background signal is subtracted away. Instead, it is the increase in noise that is associated with an increase in signal that essentially drowns out the pulses from smaller nanoparticles (Figure S-3). For instance, if the background signal is around 0.01 μg/L Ag the size of the pulse 3σ above the average

techniques, where the measured parameters are derived from large numbers of particles. While single particle techniques can avoid many artifacts through the benefit of analyzing individual particles, they are also subject to reduced statistical coverage. This reduced statistical coverage likely explains the poorer precision for both spICPMS and NTA, where the sample size for both techniques was around 100−300 particles/run. For single particle techniques, precision similar to DLS and DCS can be achieved by increasing the sample size to more than 1000 particles/run.20,23 Peak Width. Figure 1 shows the particle size distributions generated from spICPMS, DCS and NTA analysis (raw intensity versus time data for spICPMS are shown in Figure S-1). DLS size distributions were not included in Figure 1, as they were exceptionally broad, and thus uninformative for comparative purposes. As shown in Table 1, all three methods demonstrate general agreement in the overall average particle size. However, comparison of the width of the distributions produced by each method in Figure 1 shows that peak resolution from spICPMS analysis is distinctly greater than NTA and has nearly the same resolution as the DCS (an instrument known for its superior resolution capabilities24). Furthermore, spICPMS analysis of the 100 nm particle solution showed evidence of a multimodal size distribution with a primary peak around 80−90 nm and a smaller, secondary peak at 100−110 nm, which was further confirmed by the DCS. NTA was not able to resolve this second peak, which is likely the result of its reduced resolution, when compared to spICPMS. In comparing the distribution widths, one must recognize that each method will be subject to various artifacts that can artificially broaden the distribution. For DSC, the major contributors to peak broadening are diffusion and convective flow due to temperature gradients.25 The DCS methods employed in this study minimized these effects. For NTA, particle size is computed from the distance that the NPs move as a result of Brownian motion (which is the same physical phenomena used for DLS analysis). The breadth of the size distributions arises from uncertainties in measurement of the distance traveled.26 In the case of spICPMS, the partial capture of the ion cloud during a given dwell time will result in a smaller pulse and thus a smaller computed size. Conversely, the capture of an intact particle and a partial ion cloud will result in a larger signal. The length of time this transient cloud of ions occurs depends on a number of instrumental configuration and operational factors.27,28 For the instrument used in this study, this is likely about 0.3 ms.19 Comparing this short transient time to the 10 ms dwell time used in this work suggests probability of less than about a one in sixteen chance of capturing a partial cloud of ions. Given this low probability, combined with that fact that a change in the mass of ions translates to a change in 1/3 power in diameter, this effect contributes minimally to peak broadening, as evidenced by comparison of the peak widths to the TEM data. A number of workers,19,27,28 including our group, are investigating the use of very short dwell times (