Letter pubs.acs.org/ac
Hydrodynamic Chromatography Coupled with Single ParticleInductively Coupled Plasma Mass Spectrometry for Investigating Nanoparticles Agglomerates Denis Rakcheev,‡ Allan Philippe,‡ and Gabriele E. Schaumann* Institute for Environmental Sciences, Department of Environmental and Soil Chemistry, University Koblenz-Landau, Landau, D-76829 Germany S Supporting Information *
ABSTRACT: Studying the environmental fate of engineered or natural colloids requires efficient methods for measuring their size and quantifying them in the environment. For example, an ideal method should maintain its correctness, accuracy, reproducibility, and robustness when applied to samples contained in complex matrixes and distinguish the target particles from the natural colloidal background signals. Since it is expected that a large portion of nanoparticles will form homo- or heteroagglomerates when released into environmental media, it is necessary to differentiate agglomerates from primary particles. At present, most sizing techniques do not fulfill these requirements. In this study, we used online coupling of two promising complementary sizing techniques: hydrodynamic chromatography (HDC) and singleparticle ICPMS analysis to analyze gold nanoparticles agglomerated under controlled conditions. We used the single-particle mode of the ICPMS detector to detect single particles eluted from an HDC-column and determine a mass and an effective diameter for each particle using a double calibration approach. The average agglomerate relative density and fractal dimension were calculated using these data and used to follow the morphological evolution of agglomerates over time during the agglomeration process. The results demonstrate the ability of HDC coupled to single-particle analysis to identify and characterize nanoparticle homoagglomerates and is a very promising technique for the analysis of colloids in complex media.
E
results alone.9,12 Thus, it is not possible to distinguish between large primary particles and between homo- or heteroagglomerates of the same elemental mass without additional information on the system.7 Another powerful method for characterizing NPs is hydrodynamic chromatography (HDC).13,14 Starting from the commonly accepted mechanism, the separation mechanism relies solely on size and does not depend on the coating or the surface charge of the particle.14 Tiede et al.13 recently showed an application of HDC for separating a mixture of TiO2, SiO2, Al2O3, and Fe2O3 NPs in sewage sludge. These results suggested that this versatile method offers separation from ionic/molecular background signals, simultaneous sizing, and quantification in a short time frame. However, HDC-ICPMS has a detection limit in the μg L−1 range and does not provide information on the structure or the mass of the particles as it measures only the effective diameter (DE, maximum cross section).14,15 Therefore, a method is
ngineered nanoparticles (NPs) are of great concern1,2 because of their increasing use in ordinary products and their size-related properties. The environmental concern of nanomaterial is still little understood although many scientists are working in this field. One of the greatest difficulties in this field is detecting, quantifying, and analyzing NPs in complex matrixes such as soil or surface water,2 because of their low concentration,3 high background signal of natural colloids,1 and various possible modifications to the physical states (homo- or heteroagglomeration, formation of salts, and coating4). Following published recommendations, we use the term “agglomerate” exclusively.5 In an effort to achieve NPs characterization, that is fulfills these prerequisites, early works demonstrated the feasibility of single-particle inductively coupled plasma mass spectrometry (sp-ICPMS).6 In the last years, this technique was applied successfully to several colloidal systems3,7−9 and showed the following advantages: a very low detection limit (ng L−1 range), with simultaneous sizing and quantification, generally no sample preparation, element specificity, and rapid measurement.10 Although an agglomeration process can be monitored by measuring the increase of the particle mass,11 only the particle elemental mass can be calculated from the sp-ICPMS © 2013 American Chemical Society
Received: June 28, 2013 Accepted: October 24, 2013 Published: October 25, 2013 10643
dx.doi.org/10.1021/ac4019395 | Anal. Chem. 2013, 85, 10643−10647
Analytical Chemistry
Letter
Agglomeration Experiments. In order to obtain agglomerates of different sizes and at different states of agglomeration, 4 μL of 10 nm Au NPs suspension was diluted in a 5 mM CaCl2 solution to obtain a final gold concentration of 0.1 mg L−1. After adding the nanoparticle suspension, the mixture was shaken for 5 s and was allowed to stand before sampling. The suspension was shaken three times overhead before each sampling in order to homogenize the suspension, and 50 μL aliquots were sampled from the middle of the vial and diluted in 1 mL MQW in order to slow the agglomeration process. Sampling was carried out at several time points up until 2 h of agglomeration had elapsed: 0, 10, 20, 40, 60, and 120 min. These aliquots were analyzed using HDC-sp-ICPMS. A control sample was prepared using MQW instead of CaCl2 solution. The 10 nm gold NPs from the NIST were chosen although their primary particles were not detectable by sp-ICPMS, because they readily agglomerated upon addition of CaCl2. The larger gold NPs from Aldrich remained stable even at high ionic strength and at various pH levels for over 12 h. Agglomerate relative density and fractal dimension values were obtained using 3 replicates, where each replicate contained approximately 2500 data points. Agglomerate relative density and fractal dimension values were obtained using three replicates, for each sample, to estimate the maximum errors of DE and DC (considered here as ΔDC and ΔDE, respectively). Each replicate contained 1000 to 3000 data points. From the errors, we calculated the maximal errors in the relative density and fractal dimension using the following equation:
required that is able to determine the mass and geometric parameters in order to analyze the structure of agglomerates. A combination of sp-ICPMS and HDC (HDC-sp-ICPMS) should give information on two complementary agglomerate characteristics, which are both mass and DE. HDC-sp-ICPMS was first used to determine the size, number concentration, and mass fraction of gold NPs added to bottled drinking water.16 However, this technique was not used to characterize homoagglomerates, heteroagglomerates, or mixed particles (e.g., AgCl, Ag2S, and Ag2SO4). The objectives of this study were to obtain information on the morphology of primary gold NPs and agglomerates, determining the mass and D E of individual particles simultaneously. Using a methodology adapted from the pioneering study mentioned above,16 for these agglomerates, we calculated the average mass and effective diameter and, using these values, the relative density and the fractal dimension. We defined the relative density parameter of a particle (agglomerate or primary particle) by the ratio of the particle mass and the volume of an imaginary sphere with the same DE as the analyzed particle (particle volumetric mass) divided by the bulk density of gold. For the calculation of the fractal dimension, we follow the commonly accepted definition:17
( ) =3 ln( ) ln
dF
DC D1 DE D1
(1)
⎛ ⎜ ΔD 1 ΔdF = dF⎜ C D ⎜ C ln DC D1 ⎝
where D1 is the average diameter of the primary particles and Dc =
6mp 3
πρAu
( )
(2)
where mp is the mass of the particle and ρAu is the bulk density of gold. DC corresponds to a gold sphere having the same mass as the particle.
■
■
MATERIAL AND METHODS Chemicals. All chemicals obtained from the suppliers were used without further purification. Milli-Q water (MQW) was used for all dilutions and the sample preparations. Gold Nanoparticles. Standard citrate-stabilized gold NPs (Aldrich, nominal diameters: 30, 50, 100, 150, and 250 nm) were used as size calibrants. The sizes of these calibrants were verified using scanning electron microscopy and nanoparticletracking analysis (see Supporting Information). Ten nm citratestabilized NPs (nominal diameter, NIST, certified analytical standards, characteristics from the supplier are available in the Table S-1 in the Supporting Information) were used for the agglomeration experiment. The size of these particles was verified (9.9 ± 1.2 nm) using normal mode ICPMS coupled to HDC (column: PL-PSDA type 1, four replicates). The elemental concentration of these suspensions was determined using ICP-OES after being dissolved in aqua regia and conformed to those provided by the suppliers. Descriptions of the HPLC and ICPMS systems can be found in the Supporting Information. Data Analysis. Data processing is described in the Supporting Information. We observed artifacts due to unavoidable remnants from suspensions injected prior to the experiments as reported also by Pergantis et al.16 The data processing was designed to remove these artifacts.
⎞ ΔDE 1 ⎟ + ⎟ DE ln DE ⎟ D1 ⎠
( )
⎛ ΔD ΔDE ⎞ ΔdP = 3dP⎜ C + ⎟ DE ⎠ ⎝ DC
RESULTS AND DISCUSSION Calibration. Carrying out a calibration in single-particle mode requires the consideration of each spike signal obtained. The average intensities and average retention times obtained were plotted against particle size. The mass of the calibrants was calculated using the core diameter measured by scanning electron microscopy (SEM, at least 150 particles) assuming a spherical geometry of the respective standard. We calibrated DC and DE using spike intensity and retention time, respectively, by plotting a calibration curve using the gold standards described above as calibrants. The particle relative density was calculated using geometric relationships: dp =
mp VEρAu
=
DC3 DE 3
(3)
where DC is defined in eq 2 and VE is the volume of a sphere with a diameter equal to DE. Calibration curves for spike intensity (see Figure S-2 a, Supporting Information) and retention time (see Figure S-2 b, Supporting Information) were obtained using gold standards over 10 repetitions. A linear increase in the spike intensity with particle mass was observed in accordance with the fact that spike intensity is proportional to the number of detected ions 10644
dx.doi.org/10.1021/ac4019395 | Anal. Chem. 2013, 85, 10643−10647
Analytical Chemistry
Letter
Figure 1. (a) 2D-plot of the DE calculated from the retention time and the diameter calculated from the spike intensity (DC) measured by HDC-spICPMS of standard citrate-stabilized gold nanoparticles with nominal diameters from 30 to 250 nm. Straight line represents DE = DC function. (b) 2D-plot of the DE over the DC of 10 nm gold NPs after 2 h of agglomeration. Dashed line represents the DE (9.9 ± 1.2 nm), measured by HDCICPMS in normal mode, at time zero for the gold primary particles.
deviations for DE were 2−4 times higher than for DC. This suggests a higher uncertainty in the DE measurement compared to the measurement of the mass. The distribution of DE reflects the polydispersity of the sample as observed using other methods such as NTA, DLS, and SEM. For 30 nm Au NPs, the lowest DC values were discrete because they are close to the background signal limit and thus correspond to a low number of ions reaching the photomultiplier. The detection limit was determined as DC for which the distance between two consecutive discrete values reaches 5% of the corresponding highest measured DC. Following this definition, we were able to achieve a detection limit of 20 nm for DC. The distribution of DC for the 250 nm particles is due to a technical feature of the specific detector used for this study, which switches to “analog” mode when the amount of ions entering the detector exceeds the upper limit of the photomultiplier. When the amount of ions reach a certain value (in our system around 900 000 count s−1), the ions detector reaches the saturation. In this case, the photomultiplier switches to a less sensitive mode, where another ratio of counts per seconds to number of detected ions is used. This technical limitation thus leads to an upper cut off and to a DC which is apparently much lower than expected for heavy particles. This is the reason why in Figure 1a two clouds of points for the 250
and thus to the mass of each particle. The retention time decreases as the particle diameter increases and a linear relationship was observed between the retention time and the square root of the diameter as expected from HDC theory and previous publications.1,2 These two calibration curves were used to determine DE and DC for each single particle on the basis of retention time and mass, respectively. Figure 1 shows the DC obtained from eq 2, as a function of the DE calculated for each individual spike signal (particle) of the gold standard calibrants using the calibration curves shown in Figure S-2, Supporting Information. As expected, DC and DE for this data set are distributed symmetrically around the 1/1 line. Table S-2, Supporting Information, shows comparisons of core diameters obtained using SEM and HDC-sp-ICPMS. Related distribution histograms are given in the Supporting Information. There is a shift in the modal size distribution of 0−11% toward larger particles for the sizes obtained using HDC-sp-ICPMS compared to those obtained using SEM. The standard deviation for HDC-sp-ICPMS was also 15−25% greater than for SEM. Thus, HDC-sp-ICPMS reveals that an intrinsic size estimation error of 15−25% exists for HDC-spICPMS. Therefore, we considered that the distribution of the data points along the DC axis observed in Figure 1 is mainly due to the width of the size distribution of the dispersions. Standard 10645
dx.doi.org/10.1021/ac4019395 | Anal. Chem. 2013, 85, 10643−10647
Analytical Chemistry
Letter
nm particles can be observed. The first cloud actually corresponds to incorrectly attributed DC of particles with a mass exceeding the upper cutoff. Due to the specific features of the instrument software, it is not possible to trace when the detector changes mode for each spike. Thus, this is a technical limitation of the sp-ICPMS mode due to the specific equipment in the authors’ laboratory, but it could be overcome by optimizing the detection technique. However, this problem can be quickly observed by the abnormal distribution. In such a case, the heaviest colloids should be removed prior to measurement (e.g., via sedimentation, centrifugation, or filtration) providing such treatments do not disturb the structure of the agglomerates. The upper limit around 210 nm measured in this article corresponds to gold particles. It could be significantly higher for particles with a higher detection limit or for particles containing elements where several isotopes can be measured at a time (e.g., silver particles). Thus, each system needs to be validated individually regarding the upper limit of the sp-ICPMS detector. Agglomeration Experiments. During agglomeration, DC and DE gradually increased over time with the fastest changes occurring during the first 25 min (figure 2) as expected from the theory of electrostatically induced agglomeration.17 These results were confirmed by DLS experiments (see Figure S-3, Supporting Information) and are in accordance with the literature18,19 where agglomeration of citrate-stabilized gold nanoparticles was observed at high ionic strength. Stankus et al.18 observed that gold nanoparticle agglomeration occurs upon CaCl2 addition for positively and negatively charged particles stabilized with different capping agents. They observed that macromolecular stabilizers only were able to prevent agglomeration at pH 5−6. The DE became 4−5 times larger than the DC, indicating the formation of loose agglomerates. The agglomerates were measured as being up to 120 nm in size and therefore remained well below the upper mass limit of the detector. The cutoff for the greatest masses was thus not a factor in the analysis of the agglomerates in this study. Figure 1b shows DE against DC for the suspension sampled after 2 h of agglomeration. The distribution of the cloud of the agglomerated particles above the line of the DE = DC function shows that DE is much greater than DC, indicating that the agglomerated particles have a different geometry to the primary particles.17 The average agglomerate relative density and fractal dimension remained 0.0055 ± 0.0043 and 1.65 ± 0.20, respectively, throughout the entire agglomeration period. A significant decrease in the relative density was observed between 10 and 20 min, indicating the beginning of the formation of loose agglomerates. From the constant relative density until at least 60 min, we conclude that the morphology of the agglomerates is maintained during this time period. As shown above, the agglomerates were still growing during this time frame, indicating that the growth mechanism is maintained over time. The dF value of 1.65 ± 0.20, obtained in our experiment, corresponds to lose agglomerates.17 A combination of sp-ICPMS and HDC (HDC-sp-ICPMS) can, therefore, give information on both the mass and DE of each single particle and, therefore, on agglomerate characteristics via the elemental mass/volume ratio of each detected particle. Homo- and heteroagglomerates have an elemental mass/hydrodynamic volume ratio that is lower than that of primary particles. For heteroagglomerates, the elemental distribution is expected to be even less dense than for homoagglomerates. Analyzing more than one element using
Figure 2. Variation of different parameters over the time measured for 10 nm gold nanoparticles during an agglomeration experiment (2 h) using HDC-sp-ICPMS. (a) DC measured from the spike intensity curve over time; (b) efficient diameter measured using the spike retention time over time; (c) modal agglomerate relative density distribution; (d) modal fractal dimension distribution. The data were calculated using 3 replicates corresponding together to a total of around 6000 spike signals. The error bars in c and d were calculated as described in the Material and Methods section.
HDC-ICPMS in a nonsingle particle mode could even provide the opportunity to clearly distinguish between homo- and heteroagglomerates and to study the composition of each particle. In addition, HDC-sp-ICPMS allows for the estimation of the fractal dimension for each individual particle detected if the primary particle size is known. It is thus possible, starting from the analysis of spike signals, to determine if the measured particles are primary particles or if they are agglomerates. Primary particles should have an average fractal dimension of close to 3 and an average relative density of close to 1 (2.7 ± 0.2 and 0.600 ± 0.005, respectively, for our calibrants). Furthermore, considering the uncertainties on the d F 10646
dx.doi.org/10.1021/ac4019395 | Anal. Chem. 2013, 85, 10643−10647
Analytical Chemistry
Letter
(6) Degueldre, C.; Favarger, P. Y. Colloids Surf., A: Physicochem. Eng. Aspects 2003, 217, 137−142. Degueldre, C.; Favarger, P. Y.; Rossé, R.; Wold, S. Talanta 2006, 68, 623−628. (7) Pace, H. E.; Rogers, N. J.; Jarolimek, C.; Coleman, V. A.; Higgins, C. P.; Ranville, J. F. Anal. Chem. 2011, 83, 9361−9369. Tuoriniemi, J.; Cornelis, G.; Hassellov, M. Anal. Chem. 2012, 84, 3965−3972. (8) Franze, B.; Strenge, I.; Engelhard, C. J. Anal. At. Spectrom. 2012, 27, 1074−1083. Mitrano, D. M.; Lesher, E. K.; Bednar, A.; Monserud, J.; Higgins, C. P.; Ranville, J. F. Environ. Toxicol. Chem. 2012, 31, 115− 121. (9) Pace, H. E.; Rogers, N. J.; Jarolimek, C.; Coleman, V. A.; Gray, E. P.; Higgins, C. P.; Ranville, J. F. Environ. Sci. Technol. 2012, 46, 12272−12280. (10) Hassellov, M.; Readman, J. W.; Ranville, J. F.; Tiede, K. Ecotoxicology 2008, 17, 344−361. (11) Han, G. J.; Xing, Z.; Dong, Y. H.; Zhang, S. C.; Zhang, X. R. Angew. Chem.,Int. Ed. 2011, 50, 3462−3465. (12) Heithmar, E. M. Abstracts of Papers of the American Chemical Society, 237th ACS National Meeting, Salt Lake City, UT, March 22− 26, 2009; American Chemical Society: Washington, DC, 2009. Heithmar, E. M.; Pergantis, S. A. Characterizing Concentrations and Size Distributions of Metal-Containing Nanoparticles in Waste Water, EPA/600/R-10/117; U.S. Environmental Protection Agency: Washington, DC, 2010. (13) Tiede, K.; Boxall, A. B. A.; Tiede, D.; Tear, S. P.; David, H.; Lewis, J. J. Anal. At. Spectrom. 2009, 24, 964−972. (14) Striegel, A. M.; Brewer, A. K. In Annual Review of Analytical Chemistry; Cooks, R. G., Yeung, E. S., Eds.; Annual Reviews: Palo Alto, 2012; Vol 5, pp 15−34. (15) Striegel, A. M. Anal. Bioanal. Chem. 2012, 402, 77−81. (16) Pergantis, S. A.; Jones-Lepp, T. L.; Heithmar, E. M. Anal. Chem. 2012, 84, 6454−6462. (17) Hunter, R. J. Foundations of colloid science; Oxford University Press: Oxford, 2001. (18) Stankus, D. P.; Lohse, S. E.; Hutchison, J. E.; Nason, J. A. Environ. Sci. Technol. 2011, 45, 3238−3244. (19) Zook, J. M.; MacCuspie, R. I.; Locascio, L. E.; Halter, M. D.; Elliott, J. T. Nanotoxicology 2011, 5, 517−530. (20) Amal, R.; Raper, J. A.; Waite, T. D. J. Colloid Interface Sci. 1990, 140, 158−168. Amal, R.; Raper, J. A.; Waite, T. D. J. Colloid Interface Sci. 1992, 151, 244−257.
determination, this method has the potential to distinguish between reaction- and diffusion-limited agglomerates (fractal dimensions of around 2.5 and 1.7, respectively).20 HDC-spICPMS thus shows a great potential for the analysis of agglomerates at very low concentrations in environmental media.
■
CONCLUSION HDC-sp-ICPMS shows great potential for resolving agglomerated particles from their building blocks, primary particles. Agglomeration processes, involved in the formation of agglomerates, are characterized here with a fractal dimension of ∼1.65 and a particle relative density of ∼0.0055. These values correspond to loosely agglomerated particles. A combination of HDC and sp-ICPMS allows the fractal dimension and agglomerate relative density parameters to be determined, which are characteristic for agglomerated particles, making this method a novel approach for the characterization of agglomerated particle morphology. Improvements to the mass detectors and the development of software for analyses in single-particle mode should allow investigation of heteroagglomerates of different sizes in complex media and thus allow the application of this method in more complex systems, e.g., where chemical changes to the particles (oxidation and the formation of salts) and organic matter embedment are occurring. The feasibility of multielemental analysis should also be addressed in the future.
■
ASSOCIATED CONTENT
S Supporting Information *
Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Author Contributions ‡
D.R. and A.P. contributed equally to this work.
Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS We would like to thank Lotto Rheinland-Pfalz-Stiftung SozialStipendien for partially financing the work and the DFG SP MASK (SCHA849/16-1) for financial support of the experiments within the INTERNANO research unit (FOR1536).
■
REFERENCES
(1) Wigginton, N. S.; Haus, K. L.; Hochella, M. F. J. Environ. Monit. 2007, 9, 1306−1316. (2) Christian, P.; Von der Kammer, F.; Baalousha, M.; Hofmann, T. Ecotoxicology 2008, 17, 326−343. (3) Nowack, B.; Ranville, J. F.; Diamond, S.; Gallego-Urrea, J. A.; Metcalfe, C.; Rose, J.; Horne, N.; Koelmans, A. A.; Klaine, S. J. Environ. Toxicol. Chem. 2012, 31, 50−59. (4) Hotze, E. M.; Phenrat, T.; Lowry, G. V. J. Environ. Qual. 2010, 39, 1909−1924. Levard, C.; Hotze, E. M.; Lowry, G. V.; Brown, G. E. Environ. Sci. Technol. 2012, 46, 6900−6914. Lowry, G. V.; Gregory, K. B.; Apte, S. C.; Lead, J. R. Environ. Sci. Technol. 2012, 46, 6893−6899. (5) Nichols, G.; Byard, S.; Bloxham, M. J.; Botterill, J.; Dawson, N. J.; Dennis, A.; Diart, V.; North, N. C.; Sherwood, J. D. J. Pharm. Sci. 2002, 91, 2103−2109. 10647
dx.doi.org/10.1021/ac4019395 | Anal. Chem. 2013, 85, 10643−10647
