Langmuir 2009, 25, 3743-3747
3743
Size-Selective Separation of Polydisperse Gold Nanoparticles in Supercritical Ethane Dylan P. Williams and John Satherley* Centre for Nanoscale Science, Department of Chemistry, The UniVersity of LiVerpool, Crown Street, LiVerpool, L69 7ZD, United Kingdom ReceiVed July 7, 2008. ReVised Manuscript ReceiVed January 13, 2009 The aim of this study was to use supercritical ethane to selectively disperse alkanethiol-stabilized gold nanoparticles of one size from a polydisperse sample in order to recover a monodisperse fraction of the nanoparticles. A disperse sample of metal nanoparticles with diameters in the range of 1-5 nm was prepared using established techniques then further purified by Soxhlet extraction. The purified sample was subjected to supercritical ethane at a temperature of 318 K in the pressure range 50-276 bar. Particles were characterized by UV-vis absorption spectroscopy, TEM, and MALDI-TOF mass spectroscopy. The results show that with increasing pressure the dispersibility of the nanoparticles increases, this effect is most pronounced for smaller nanoparticles. At the highest pressure investigated a sample of the particles was effectively stripped of all the smaller particles leaving a monodisperse sample. The relationship between dispersibility and supercritical fluid density for two different size samples of alkanethiol-stabilized gold nanoparticles was considered using the Chrastil chemical equilibrium model.
1. Introduction The development of a simple technique to produce monodisperse samples of ligand-stabilized metal nanoparticles is one of the most active areas of research in nanoscale science.1 In a previous study2 it was shown that alkane thiol-capped gold nanoparticles disperse in supercritical ethane and that the dispersibility was dependent on the core diameter. That study2 indicated the possibility of using a supercritical fluid to disperse selectively gold nanoparticles with a very small range of core diameters. The principal aim of the present study was to expand on the previous work2 by exploring whether gold nanoparticles, as prepared, with a small degree of polydispersity can potentially be split, using a supercritical fluid, into fractions with each containing particles of the same core diameter and to quantify the results. Additional aims were to fully characterize the particles prior to and following treatment with supercritical ethane and to examine the pressure dependence of the particle dispersibility for different sizes of nanoparticles and interpret the results using the Chrastil chemical equilibrium model.3 Due to their unique physical properties gold nanoparticles are widely studied and have a variety of practical applications.4 In order for these particles to be useful to computer and life science applications a near-monodisperse sample is required. Several techniques have been developed for separating a polydisperse sample into monodisperse fractions including controlling the synthesis of particles by using polymeric stabilizers,5 the use of size exclusion chromatographic techniques,6 HPLC,7 gas expanded liquids,8 and * To whom correspondence should be addressed. E-mail: js1@ liverpool.ac.uk. Phone: +44151 794 3530. (1) Brust, M.; Kiely, C. J. Colloids Surf., A 2002, 202, 175–186. (2) (a) Clarke, N. Z.; Waters, C.; Johnson, K. A.; Satherley, J.; Schiffrin, D. J. Langmuir 2001, 17, 6048–6050. (b) Shah, P. S.; Holmes, J. D.; Johnston, K. P.; Korgel, B. A. J. Phys. Chem. B 2002, 106, 2545–2551. (c) Shah, P. S.; Hanrath, T.; Johnston, K. P.; Korgel, B. A. J Phys. Chem. B 2004, 108, 9574–9587. (3) Chrastil, J. J. Phys. Chem. 1982, 86, 3016. (4) Daniel, M.-C.; Astruc, D. Chem ReV 2004, 104, 293–346. (5) Hussain, I.; Graham, S.; Wang, Z.; Tan, B.; Sherringham, D. C.; Rannard, S. P.; Cooper, A. I.; Brust, M. J. Am. Chem. Soc. 2005, 127, 16398–16399. (6) (a) Al-Somali, A. M.; Kreuger, K. M.; Falkner, J. C.; Colvin, V. L. Anal. Chem. 2004, 76, 5903–5910. (b) Siebrands, T.; Giersig, M.; Mulvaney, P.; Fischer, C. H. Langmuir 1993, 9, 2297–2300.
recently the chemical recognition properties of DNA.9 The use of supercritical fluids offers a simple, clean, and tuneable medium for dispersing metal particles. The size range of particles dispersed may be controlled by simply varying the pressure of the fluid.
2. Experimental Section Dodecanethiol-stabilized gold nanoparticles were prepared by the Brust synthesis technique with a 3:1 gold-to-thiol ratio.10 Following purification by washing in ethanol (to remove excess alkanethiol) and the removal of the solvent by evaporation, the product was further purified by Soxhlet extraction.11 This latter technique is a continuous extraction process which uses a flask, a condenser, and an extraction chamber between the two which fills with hot, condensed solvent. The material to be purified is transferred to an extraction thimble (typically made out of cellulose) by drying from solution. The thimble is placed inside the condenser; when the process is started the solvent vapor is condensed and runs into the extraction chamber, the chamber fills with hot solvent. Once the chamber is full it is emptied by a siphon arm, this returns the solvent to the flask. The solvent is once again evaporated and condensed, which provides a continuous supply of hot, clean solvent for the extraction process. This process suits a solid material with an impurity which has limited solubility in a given solvent.11 This extraction process has been shown12 to reduce the amount of quaternary ammonium salt that remains electrostatically bound to the particles after synthesis, thereby potentially increasing the dispersibility of the particles in nonpolar supercritical fluids, such as ethane. Following preparation, the nanoparticles were placed into an openended glass vial by evaporation from a solution of toluene. The vial was placed into a stainless steel pressure vessel; the vessel was then pressurized with ethane and heated so that the ethane was above its (7) (a) Jimenez, V. L.; Leopold, M. C.; Mazzitelli, C.; Jorgenson, J. W.; Murray, R. W. Anal. Chem. 2003, 75, 199–206. (b) Wilcoxon, J. P.; Martin, J. E.; Provencio, P. Langmuir 2000, 16, 9912–9920. (8) McLeod, M. C.; Anand, M.; Kitchens, C. L.; Roberts, C. B. Nano Lett. 2005, 5, 461–465. (9) Lee, J.-S.; Stoeva, S. I.; Mirkin, C. A. J. Am. Chem. Soc. 2006, 128, 8899–8903. (10) Brust, M.; Walker, M.; Bethell, D.; Schiffrin, D. J.; Whyman, R. Chem Commun. 1994, 801–802. (11) Von Soxhlet, F. Polytechnisches J. (Digler’s). 1879, 232, 461. (12) Waters, C. A.; Mills, A. J.; Johnson, K. A.; Schiffrin, D. J. Chem Commun. 2003, 540–541.
10.1021/la802133r CCC: $40.75 2009 American Chemical Society Published on Web 02/10/2009
3744 Langmuir, Vol. 25, No. 6, 2009
Williams and Satherley
Figure 2. Comparison of the colors of the undispersed (A) and dispersed (B) samples at 276 bar.
Figure 1 shows the absorbance of undispersed particles (A) and dispersed particles (B) as a function of wavelength at the pressures indicated on the figure. The trend observed in the comparison of the UV-vis absorption spectra for the undispersed
samples (Figure 1A) (these were recorded in an identical volume of toluene as the dispersed samples to allow direct comparison) is consistent with that observed from the spectra of the dispersed samples. The total amount of sample that remained undispersed decreased as the pressure of the system was increased, additionally, a more intense plasmon absorption band is observed in these spectra at all pressures. This observation corresponds to the presence of a higher proportion of larger nanoparticles in these samples compared to the dispersed samples at corresponding pressures. The comparison of the UV-vis spectra (each measured in 5 mL of toluene) recorded for the dispersed particles (in the range of pressures 50-250 bar) shows an increase in dispersibility with increasing pressure (Figure 1B). The observed increasing dispersibility is consistent with the increasing solvent strength of the supercritical fluid with increasing pressure.2 The dependence of dispersibility on particle size was considered in earlier publications.2 It was shown that larger particles require higher pressures to become dispersible and become progressively more dispersible as the pressure increases still further. This is largely due to the increasing particle-solvent attractive interactions at higher pressures which eventually become more significant than the strong particle-particle interactions between large molecules at higher pressures. The increasingly pronounced plasmon band absorbance seen in the UV-vis spectra with increasing pressure illustrates the increasing dispersibility of the larger particles (responsible for this band) at higher pressures. This trend supports the findings of the UV-vis absorption spectra recorded on the undispersed particles. A comparison of the colors of the solutions (Figure 2) of dispersed and undispersed nanoparticles at 276 bar supports this interpretation. The dispersed sample (B) at this pressure was brown, which suggested the average core diameter of these particles was significantly below that of the undispersed particles. At core diameters of ∼2 nm and below the plasmon band resonance becomes undetectable as the electronic structure of nanoparticles is most accurately described by discrete energy levels at these sizes so the free electron model ceases to accurately describe the electronic structure of the nanoparticles.2a,14 The undispersed sample (A) at the same pressure was red, indicating the presence of a plasmon band resonance and therefore nanoparticles with an average core diameter larger than those which dispersed. The color of the dispersed particles remains brown over the full range of pressures investigated. The TEM data (Figure 3) for the samples removed from the apparatus at a pressure of 276 bar support the evidence from the UV-vis spectra; a clear distinction between the average core
(13) Lide, D. R.; Kehiaian, H. V. CRC Handbook of Thermophysical and Thermochemical Data; CRC Press: Boca Raton, FL, 1994.
(14) Schaaf, T. G.; Shafigullin, M. N.; Khoury, J. T.; Vezmar, I.; Whetten, R. L.; Cullen, W. G.; First, P. N. J. Phys. Chem. B 1997, 101, 7885.
Figure 1. UV-visible absorption spectra for the undispersed samples (A) and dispersed samples (B) of alkanethiol-stabilized gold nanoparticles in supercritical ethane at the isobars shown.
critical point (pc ) 48.8 bar, Tc ) 305.2 K)13 using the technique described in ref 2a. The required pressure (in the range 50-276 bar) was maintained for a period of 18 h at a temperature of 318 K. At the end of the experiment the pressurized fluid was slowly released through a layer of acetone in a conical flask; nanoparticles which dispersed in the ethane precipitated in this layer and were collected by filtration. In order to extract the dispersed particles which precipitated on the inside of the high pressure apparatus during depressurization, the cell, all connecting pipework, and valves were removed and thoroughly washed with toluene. The two samples of dispersed nanoparticles were then combined in a standard volume (5 mL) of toluene for analysis. The undispersed nanoparticles remained bound to the inside of the glass vial; these particles were washed off in 5 mL of toluene (taking care to wash only the inside of the vial) at the end of each experiment. It was possible that some of the dispersed particles precipitated on the inner surface of the vial and, therefore, contaminated the sample of particles that did not disperse. However, two points need to be made regarding this possible contamination. First, the inner surface area of the vial was only 4% of the total surface area inside the pressure vessel, pipework, and valves. Also, since the vial was positioned at an angle of about 80° in the vertically mounted pressure vessel this meant that the effective surface area within the vial that could be contaminated with precipitating particles during decompression was significantly less than 4%, possibly less than 1%. Second, analysis by mass spectrometry, see below, did not detect any contamination. Consequently, contamination in these experiments did not appear to be an issue. If this had been an issue then much more complex apparatus would have been required to guarantee a high quality separation. The samples of dispersed and undispersed nanoparticles were analyzed by UV-vis spectroscopy, MALDI-TOF (matrix-assisted laser desorption/ionization-time of flight) mass spectroscopy and TEM (transmission electron microscopy).
3. Analysis of Nanoparticle Fractions
Separation of Polydisperse Gold Nanoparticles
Langmuir, Vol. 25, No. 6, 2009 3745
Figure 3. Transmission electron micrographs of the undispersed particles (A) (magnified in image D) and the dispersed particles (B) (magnified in image E) recovered from supercritical ethane at 276 bar and 318 K. Image C shows the TEM of the nanoparticles as prepared.
diameters of the two samples can be seen. The undispersed particles (A) have a larger average core diameter and more uniform size distribution than the particles seen in the micrograph of the dispersed particles (B). The TEM image of the as-prepared nanoparticles shows a wider range of nanoparticles sizes (C) These micrographs have been processed with ImageJ image software,15 and it has been found that the particles which remained undispersed have an average core diameter of 4.4 nm and the dispersed particles have an average core diameter of 2.3 nm (Figure 4). The histograms show the almost complete absence of particles greater than 4 nm in the dispersed sample (B). MALDI-TOF mass spectra (Figure 5) recorded for undispersed (A) and dispersed (B) particles at 276 bar and 318 K also support the above findings. The peak at 30 kDa is seen in both spectra and corresponds to the magic number cluster Au147. It is a relatively sharp peak due to the stability of this cluster. The peak centered on 16 kDa is due to Au79 clusters.16 Due to the stability of a number of structures similar to Au79 a broader peak is observed than for Au147. These structures may take the form of clusters with extra or missing vertex atoms. Comparing the two images (15) Abramoff, M. D., Magelhaes, P. J., Ram, S. J. Image Processing with ImageJ. Biophotonics International, Vol. 11, issue 7, pp 36-42, 2004. (16) Whetten, R. L.; Khoury, J. T.; Alvarez, M. M.; Murthy, S.; Vezmar, I.; Wang, Z. L.; Stephens, P. W.; Cleveland, C. L.; Luedtke, W. D.; Landman, U. AdV. Mater. 1996, 8, 428–433.
it can be seen that at 276 bar the undispersed sample that remains in the glass vial is free of the 16 kDa peak; this shows that the clusters based on Au79 have been completely removed at this pressure, leaving only the larger Au147 clusters. The spectrum of the dispersed particles shows the presence of both of these peaks, indicating that at this pressure all of the smaller (Au79) size gold clusters have been dispersed and some of the larger (Au147) clusters. At 100 bar (Figure 5C) both of the peaks are also observed but at this pressure the 16 kDa peak is very significantly larger in comparison to the one at 30 kDa. Comparing (C) with (B) shows that the smaller particles are much more dispersible than the larger ones and that the increase in dispersibility with increasing pressure is greater for the smaller particles. The UV-vis spectrum of dispersed particles at 100 bar (Figure 1A) shows no evidence of plasmon band character, indicating that the MALDI-TOF results are much more sensitive at detecting the presence of the larger particles. At some pressure less than 100 bar only the smaller size should be dispersible. The reason that there is a relatively large pressure range where the two sizes are dispersible is due to the two clusters being very similar in size. The undispersed Au147 clusters obtained from the supercritical separation process are close to monodisperse (as seen in the TEM images), highlighting the quality of the separation process from a sample of polydisperse starting material. The MALDI-TOF spectra suggest that this technique can be used to
3746 Langmuir, Vol. 25, No. 6, 2009
Williams and Satherley
Figure 4. Core diameter distribution histograms obtained using ImageJ image analysis software14 on the TEM images of the undispersed particles (A) and the dispersed particles (B) recovered from supercritical ethane at 276 bar and 318 K and the sample of nanoparticles as prepared (C).
obtain a sample of nanoparticles which is enriched with a certain magic number size (30 kDa in this case). An approximate dispersibility of the nanoparticles in supercritical ethane at 276 bar and 313 K can been obtained using the Beer-Lambert law and the equation proposed by Huo et al.17 to estimate the extinction coefficient of gold nanoparticles in any solvent. By assuming the dispersed nanoparticles are essentially monodisperse and have an average core diameter of 3.7 nm (from the TEM data) an extinction coefficient of 3.3 × 10-5 M-1 cm-1 is obtained which results in a concentration of 1.2 × 10-6 mol/dm3 for the dispersed nanoparticles at saturation conditions recovered in 5 mL of toluene. Further, the volume of the vessel and pipework is ∼20.4 cm3 and, at the above conditions, the density of supercritical ethane is 0.334 g/cm3. This results in a dispersibility of 3.6 × 10-9 mol/g, expressed as moles of gold dispersed per gram of supercritical ethane. The dispersibility of dodecanethiol-stabilized gold nanoparticles as a function of pressure was investigated and considered in terms of the Chrastil model3 for solutes at saturation conditions (17) Liu, X.; Atawar, M.; Wang, J.; Huo, Q. Colloids Surf., B 2007, 58(1), 3–7.
in supercritical fluids. This is a chemical equilibrium model which arrives at a simple expression:
ln C ) n ln δ + m
(1)
where C is the concentration of the solute and δ is the density of the supercritical fluid. n is a constant and is interpreted as the number of solvent molecules surrounding the solute, and m, also a constant, is related to the enthalpy of solvation and volatility of the solute. This model has previously been used to relate the dispersibility of metal nanoparticles in supercritical fluids with the density of the solvent at different isothermal conditions.18 A more detailed explanation of the theory behind this model and its application to metal nanoparticles in supercritical fluids is provided in an earlier study by Fernandez et al.18 Here we investigate this model for dodecanethiol-stablized gold nanoparticles as the solute and supercritical ethane as the solvent at a constant temperature of 318 K and as a function of pressure (or solvent density) and for two different particle sizes. (18) Fernandez, C. A.; Hoppes, E. M.; Bekhazi, J. G.; Wang, C.; Wiacek, R. J.; Warner, M. G.; Fryxell, G. E.; Bays, J. T.; Addleman, R.S. J. Phys. Chem. C 2008, 13957–13957.
Separation of Polydisperse Gold Nanoparticles
Langmuir, Vol. 25, No. 6, 2009 3747
Figure 6. Dispersibility plots of dodecanethiol-passivated gold nanoparticles (squares correspond to 4.2 nm particles and diamonds to 2.5 nm particles) at 318 K as a function of the solvent density in a ln-ln form in supercritical ethane. Solids lines were obtained by linear regression. Table 1. Chrastil Parameters of Dodecanethiol-Stabilized Gold Nanoparticles in Supercritical Ethane N M
2.5 nm particles
4.2 nm particles
2.1 -22.5
10.6 -46.1
where a plasmon band is only observed at the highest pressures measured. It is worth noting that the value of n varies significantly for the two samples of nanoparticles. The higher value of n for the 4.2 nm nanoparticles can be interpreted as corresponding to an increase in the number of solvent molecules associated with the nanoparticles dispersed in the supercritical fluid and is consistent with the increased size of these nanoparticles. These results are similar to those obtained by Fernandez et al.18 for octanethiol stabilized gold nanoparticles in supercritical ethane at pressures up to 500 atm.
4. Summary and Conclusions Figure 5. MALDI-TOF mass spectra of the undispersed particles (A) and the dispersed particles (B) recovered from supercritical ethane at 276 bar and 318 K. (C) Spectrum for dispersed particles under the same conditions expect at a pressure of 100 bar.
In order to investigate the pressure dependence of dispersibility a number of experiments were carried out in which two different monodisperse sample sizes with core diameters of 2.5 ( 1.1 and 4.2 ( 0.8 nm (obtained using the technique described by Cooper, et al.4) were processed in the pressure vessel in the way already described. The results were obtained by recording the UV-visible absorption spectra of the dispersed particles in constant volumes of toluene and calculating the extinction coefficients of these nanoparticles.15,19 The dispersibility was then determined using the method described and the equation proposed by Huo et al.16 The results are shown in Figure 6 as a plot of ln c vs ln δ. As can be seen two linear plots result for the two different sizes confirming the validity of this approach for the conditions of these experiments. Table 1 shows the values of the slope (n) and intercept (m) for each sample size. The dispersibility of the larger nanoparticles is significantly lower than that of the 2.5 nm particles over the measured pressure range. At the highest pressure (i.e., the highest density) measured the dispersibility of the two samples is approaching the same value. This is in agreement with the absorption measurements of the polydisperse sample (Figure 1) (19) Williams, D. P. Ph.D. Thesis: A Study of the Solubility and Separation of Alkanethiol Stabilized Gold Nanoparticles in Supercritical Fluids, 2008, Chapter 6, 160.
In conclusion, this study has shown that supercritical ethane was effective at isolating some of the larger diameter nanoparticles from the original sample. Clearly, different size fractions are dispersed by changing the pressure. At the highest pressures nanoparticles with a wider range of core diameters were dispersed. The application of the Chrastil model has shown the large difference in dispersibility of 2.5 and 4.2 nm size nanoparticles at low densities. However, the dispersibility values of both samples appear to approach a single value at the highest densities measured. Although this study has shown the potential of supercritical fluids for the separation of polydisperse nanoparticle samples into individual size fractions, it has also shown that if the distribution of core diameters of nanoparticles is small, separation is very difficult. It is possible that selective precipitation could be demonstrated more effectively using larger nanoparticles with a larger degree of polydispersity, but the aim of this paper was to demonstrate the application of this technique to samples of nanoparticles with small levels of polydispersity in order to obtain highly monodisperse fractions. Acknowledgment. We thank EPSRC for financial support, Alan J. Mills for providing MALDI-TOF mass spectra, and Dr. Ian Prior and Miss Cornelia Muncke at the University of Liverpool EM Unit for assistance in obtaining TEM images. Supporting Information Available: Experimental details, additional UV-visible absorption spectra, TEM images, and MALDI-TOF mass spectra. This material is available free of charge via the Internet at http://pubs.acs.org. LA802133R
