Analysis of colloids

JO URNAL O F C HRO MATO G RAPHY A ELSEVIER Journal of Chromatography zyxwvu A, 688 (1994) 97-105 Analysis of colloids VII. *Wide-bore hydrodynamic chromatography, a simple method for the determination of particle size in the nanometer size regime Ch.-H. Fischer*, M. Giersig zyxwvutsrqponmlkjihgfedcbaZYXWV Hahn-M eitner-Institut Berlin. Department CK, Glienicker Strasse 100, D- 14109 Berlin, Germany First zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA received 6 July 1994; revised manuscript received 6 October 1994 zyxwvutsrqponmlkjihgfedcbaZY Abstract Wide-bore hydrodynamic chromatography in a polyether ether ketone [PEEK) capillary (I.D. 0.7 mm, length 20 m) was used to determine the weight average diameter d, of colloidal particles. The method was applied to cadmium sulphide and gold sols in the diameter range between 3 nm and 27 nm. The method is based on the radial distribution of the analyte in the capillary due to the hydrodynamic flow profile in the capillary and due to the diffusion coefficient of the particles, which is dependent on their diameter. The diameter was calculated from the ratio of the heights of convection peak and diffusion peak. The size-quantization effect of small semiconductor particles made it possible to visualise the separation inside of the capillary. One important advantage of the applied method is the very much reduced adsorption, which often causes serious problems in the HPLC especially of inorganic colloids. The results of wide-bore hydrodynamic chromatography, size exclusion chromatography and transmission electron microscopy were compared. Introduction The study of colloidal semiconductor and metal particles in the nanometer size regime is a steadily growing field in chemistry. Because ultra small particles exhibit unusual physical and chemical properties, e.g. blue-shifted absorption and fluorescence spectra with decreasing diameter and because of their possible application in solar energy technology and microelectronic devices they have become the focus of much recent * Corresponding author. c For part VI see Ref. {Xl. for part V Ref. [7/ and for part IV Ref. 1261. 0021-9h73/94/$07.00 SSl?I 0021-9673(94)0096z-7 (Q I994 Elsevier Science B.V. All rights physicochemical research [ 1,2]. However the investigation of size dependent properties requires good and reliable size analysis. The classical technique is transmission electron microscopy (TEM), where the particle diameters are measured directly. Nevertheless, problems can arise from radiation damage due to the high energy applied to the material [3]. If no expensive image processing system is available, the measurement of the sizes on the micrographs is also rather tedious and often quite subjective. Size exclusion chromatography (SEC) has also recently been shown to be a very convenient method for the size determination for inorganic colloids [4,5]. In the case of HPLC-SEC, the system measures reserved C‘hromatogr A 68X (lY94) 97- 10- 5 size distributions within a few minutes. once the raphy. it is the peak shape, rather than the initial calibration has been carried out [6,7]. retention time, which was used for the calculaSpeed is important when kinetics of the growth tion of the molecular weight. The method is based on the dependence of the diffusion coeffiof very unstable colloids is investigated [6]. Further advantages of the chromatographic cient on the molecular weight. Conversely, diffumethod are the good statistics of the result, sion coefficients of species with known molecular on-line coupling with diode array detectors zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED for weight can be calculated from the peak width studies of size depending WVIVis spectra [6] and [20]. Vanderslice et. al. investigated the conthe possibility of scaling-up the system for precentration profiles in such flow systems and parative separations [8]. However, with some calculated them under various conditions [21]. kinds of colloids, e.g. PbS, adsorption or particle In the past the method in question has been growth during the passage through the column contributed to the group of flow injection analycauses difficulties even in the presence of the sis (FIA). However, most often in FIA constabilisers in the eluent. The hydrodynamic chrocentrations are measured by peak size either matography in packed columns showed the same after mixing with special reagents or without limitations ]9]. In the classical hydrodynamic mixing by the use of specific detectors. Though chromatography (HDC) in narrow capillaries or the described method has doubtless similarities with other name capillary hydrodynamic fracwith FIA, it is based on the hydrodynamic flow tionation (CHDF) organic particles could be profile in the capillary, which is generated by the separated according to the diameter with good flow resistance of the walls. The species to be resolution [lo-131. Silebi [ 111 described the analysed interact dynamically with the flow analysis of latex particles with diameters down to medium and indirectly with the walls. This is 88 nm (diameter range of the inorganic particles typical for chromatography, though not a specunder investigation: 3 nm-27 nm). Adsorption trum of various (particular) retention times is was less pronounced than in packed columns, measured. Therefore we prefer the name widebut still a problem. Therefore a chromatographic bore hydrodynamic chromatography (HDC) in technique could be helpful, where there is a order to stress similarities and differences to the smaller surface area than in a packed column or classical HDC. in a narrow capillary. In 1978 Mullins and Orr [14] and in 1979 Noel et al. [15] reported the fractionation of miExperimental crometer-sized particles in a capillary with an internal diameter of 250 pm. Submicrometer Chromatography particles could not be distinguished. In 1984 Kelleher and Trumbore [16] described W ide- bore HDC an easy method for the determination of the The experimental set-up consisted of a Merckmolar weight of biopolymers just by pumping a Hitachi L6000 HPLC pump, a Merck Autosamsample plug through a rather thick capillary. The pler A2000 (sample volume 100 ~1) and a Waters internal diameter was some tenths of a mil990 diode array detector, Autosampler and delimetre and therefore much thicker than in tector were connected via a 20 m long, 0.7 mm classical hydrodynamic chromatography (capilI.D. polyether ether ketone (PEEK) capillary. lary hydrodynamic fractionation) where the diThe flow rate was 0.8 mllmin. The eluent for the ameter is typically some microns. The expericadmium sulphide sols was lop3 M cadmium ment was carried out initially with a normal perchlorate (Ventron)/ +10e3 M sodium polyUV/Vis detector, but later with special RI detecphosphate (based on the phosphate units, Riedel tors which measured the radial concentration de Haen), and for gold sols lo-’ M sodium gradient [17-191. Unlike normal chromatogcitrate was used. Ch.-H. Fischer. M. Giersig : J. Chromntogr. SEC Two 125 x 4 mm columns (Knauer Stiulentechnik, Berlin, Germany) in series were used: For cadmium sulphide Nucleosil 5OOC4 (7 pm) and Nucleosil lOOOC4 (7 pm) and for gold Nucleosil 500 (15-25 pm) and Nucleosil lOOOC4 (15-25 pm). Eluents, pump and detector were the same as for HDC. A 68X (1994) size distributions micrographs. 99 97- 105 were measured from electron Electron microscopy .4 small drop of sample was adsorbed onto a 40%mesh copper grid coated with a 50 A thick carbon support film. After 10 seconds of contact time the fluid was blotted off. The grids were dried under argon and examined in a Philips CM Preparation zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA ofthe colloids 12 transmission electron microscope with an acceleration voltage of 120 kV. The microscope Cadmium sulphide sols was equipped with a supertwin lens and an Hydrogen sulphide gas or aqueous sodium EDAX detector. For imaging, axial illumination hydrogen sulphide solution was injected through was used as well as the “nanoprobe mode” with a septum into an aqueous solution of 10-j M a beam spot size of 1.5 nm, to enable diffraction Cd( ClO,), and 6 1 10ml M sodium polyphospatterns of the individual clusters to be obtained. phate, through which nitrogen had been bubbled All images were made under conditions of minifor ten minutes. The solution was shaken prior mum phase contrast and low electron dose with a to use. The particle size was controlled by the magnification of 120 000 and 430 000 X . initial pH value of the solution before sulphidc addition. A lower starting pH leads to smaller particles [22). Aged samples containing larger Results and discussion colloids were also used in some experiments. M ethod Gold sols Gold sols were prepared by using a mixture of trisodium citrate and tannic acid (Mallinckrodt product no. 8835) as reducing agent [23]. KAuCI, (85 ml, 0.1%) was heated to 60°C and stirred rapidly. A second reducing solution was prepared by mixing trisodium citrate (4 ml, 1% ) tannic acid (O-S ml, 1%) and an equivalent amount of K,CO, (O-5 ml. 10 2 M and making up to 25 ml.- This solution was also heated to 60°C and then added rapidly to the chloroaurate solution. The colour of these sols developed almost instantly. The solution was then boiled for several minutes and allowed to cool. Tannic acid increases the rate of nucleation, thereby generating smaller particles. The higher the tannic acidzcitrate ratio, the smaller the particle size. The lowest size limit achievable was found to be about 2-3 nm. The sols so prepared were stable for months, although sometimes a slow sedimentation was observed over time, which could be removed by centrifugation. The particle When a liquid passes through a capillary under laminar flow conditions, a parabolic flow profile is formed, i.e, fast flow in the centre and decreasing velocity towards the walls. Dissolved species are- transported forward with the liquid flow, but they can also move in other directions by diffusion. Here the motion perpendicular to the flow direction is of particular importance. It brings the solute from the centre to the walls and vice versa, i.e. from faster streams to slower flowing parts of the cross-section. However, this motion is dependent on the diffusion coefficient of the sample. When a sample of very big colloidal particles or very large macromolecules is injected into such a flow system with appropriate flow rate, their diffusion is low compared to the speed of the forward stream. Therefore the radial movement is negligible. In Fig. la the axial and radial concentration distribution of sample species in the capillary is shown schematically. based on the calculations of Van- loo Ch.-H. dif f usion al Fischer, M. Giersig I J. Chromatogr. A 688 (1994) 97-105 zyxwvutsrqponmlkjihgfedcbaZYXWVU relatively symmetric late-eluting diffusion zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA roe f f icient ioi peak diffusion coefficient. Materials with a medium diffusion coefficient show an elution profile with the elements of both kinds of peaks, the ratio depending on the size of the diffusion coefficient. Fig. lb shows the elution profiles corresponding to the upper situations. The diffusion coefficient of dissolved organic polymers depends on the size of their coils which is proportional to the molecular weight. In the case of colloidal particles their diameter is the important parameter. The experimental setup is very simple (Fig. 2). A pump delivers the eluent, and the sample is introduced by a sample valve and pumped through the capillary to the detector. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQ low for material with a high medium high - zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA -a 2 0 0 b1 iGj 0 axial distance I La-u.1 - . ;:%odof l -a 0 - time 1a.u.l Fig. 1. Behaviour of material with small, medium and high diffusion coefficient during laminar flow through a “widebore” capillary. (a) Axial and radial concentration distribution expressed in equiconcentration lines, based on the theoretical calculations of Vanderslice et al. 1211. (b) Corresponding elution profiles. derslice et al. [21]. The left part of Fig. 1 represents the situation just described for low diffusion coefficient. The analyte follows the laminar flow profile, apparently without any additional motion. On the other hand species with a very high diffusion coefficient exchange efficiently between slow and fast areas (Fig. 1, right part), and therefore the concentration is more uniform over the diameter and the average speed is slower than in the first case. Species with medium diffusion coefficient (Fig. 1, centre) show a behaviour between the extreme cases. These distributions cannot be easily visualised. But with a simple experimental set-up consisting of a pump, an injection valve with sample loop, the capillary and a detector with a through-flow cell, an elution profile can be obtained which is similar in principle to a chromatogram. The response gives the integral radial concentration at a certain axial distance from the injection point. Due to the flow, the whole distribution is pushed through the detector cell with time. The result is an early eluting, steeply increasing, but strongly tailing and therefore asymmetric convection peak for slowly diffusing species and a Application Cadmium sulphide sols For the first experiments cadmium sulphide sols were used. Much experience exists in SEC of these semiconductor colloids, so that the results could easily be compared. Stabilisers such as polyphosphates have to be added to these aqueous colloids in order to reduce particle growth. These molecules form complexes with the surface of the particles and protect it against direct contact with others and therefore against coagulation. On the other hand it also reduces adsorption on the surface of the column and the stationary phase. A series of CdS sols of different particle size, lo-’ M each, were prepared, whereby the size was controlled via the pH value before the sulphide addition. These sols were injected simultaneously in the HDC capillary and onto the SEC column. The eluent compositions of both methods were the same: lop3 M cadmium Fig. 2. Scheme of experimental set-up in wide-bore HDC. Ch.-H. Fischer, M. Ciersig 1. Chromatogr. A 688 (1994) 101 97- 105 shorter retention volumes, the HDC peak became less symmetric. In addition to the late diffusion peak, the early convection peak grew. Finally only the latter remained with a pronounced tailing. For proof of the separation inside a wide capillary the size quantization effect (Q effect) of the nm-sized semiconductor particles could be used. For these particles, the onset of absorption shifts to shorter wavelengths with decreasing particle size [24,25]. When a CdS sol with a broader size distribution was analysed, different 1 &,Yj$, 27.l)nm 1 chromatograms were obtained depending on the observation wavelength. At shorter wavelengths 3 4 6 14 22 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA (250 nm and below) all sizes are detected and time / min time / min have the same molar extinction coefficient, Fig. 3. Cadmium sulphide sols of different particle sizes whereas at longer wavelengths smaller particles analysed by SEC (left) and wide-bore HDC (right). The absorb less than larger ones. Consequently we samples are sorted, the smallest particles on top, largest at found in Fig. 4a at 250 nm a large diffusion peak the bottom. The weight average diameters d, are given on next to a small shoulder due to a convection the right hand side. peak. With increasing wavelength, the shoulder perchlorate/ * lop3 A4 sodium polyphosphate. grew and at 500 nm the convection peak was Fig. 3 shows the results. On the left, the size pronounced, because smalIer particles do not exclusion chromatograms and on the right the absorb any more in this wavelength range. The corresponding hydrodynamic chromatograms. separation is also evident from the spectra taken The samples are sorted with respect to increasing with the diode array spectrometer during the run particle diameter. As the SEC peak shifted to (Fig. 4b). The spectrum at 8.7 min had an onset SEC I WIDE BORE HOC 5UOnm \ --..-~ ~ I I I I I 6 10 14 18 22 time /min I h 250 . 300 . . 400 wavelength Fig. 4. Wide-bore HDC with diode array detection and the size quantization effect of a cadmium broad size distribution. Left: Chromatograms obtained at different wavelengths. Right: UV-Vis (dotted line), at 11.9 min (solid line) and the difference A between both spectra (dashed line). . . 500 / nm sulphide sol with a relatively spectra measured at 8.7 min 102 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Ch.-H. Fischer. M. Giersig i 1. Chromarogr. A 688 (1994) 97- 105 of absorption near 500 nm and no further fine weight from the obtained elution profiles. When structure, typical for rather large CdS particles. a gradient detector was used, the molecular mass The spectrum obtained at 11.9 min showed a M, was determined by means of the asymmetry similar onset, but also a small maximum at 330 ratio of the derivative signal [27]. Trumbore et nm and a shoulder at 350 nm. This spectrum is a al. [28] suggested for the normal elution profiles,” superposition of smaller and bigger particles, the ratio R of the height of the convection peak since the bigger particles of the sample are still h, to that of the diffusion peak h, (Eq. 1) for the eluting. As is evident from the chromatogram at determination of the molar weight M, of a 500 nm they are only of medium size with polymer by empirical correlation. When one of convection and diffusion peaks of about the the both peaks is not sufficiently pronounced, same height. Therefore a spectrum with a prothe height of the chromatogram at the position in nounced maximum typical for particles below 3 question is taken for the calculation. nm is obtained, when the first spectrum is subtracted from the second (spectrum A in Fig. R = h,ih, (1) 4b). The maxima in this region are due to socalled magic agglomeration numbers, i.e. energetically very stable agglomerates [ 1,2]. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED We also used the height ratio R, but as a function of the weight average diameter d,, since colloid chemists prefer the use of diameter, Gold sols Gold sols were prepared by the reduction of considering that the solid particles are rigid and tetrachloroaurate with citrate in the presence of non-swelling as opposed to the organic polymer tannic acid [23]. Increasing concentration of the coils. For the calibration, a number of colloids latter yields smaller particles. A series of these were prepared and their weight average diameter sols was also investigated by transmission elecwas determined by computer evaluation of the tron microscopy (TEM). In the micrographs, the SEC chromatograms or in the case of gold sols diameters of a sufficiently high number of pardirectly by TEM. In Fig. 6 the ratio R as a ticles, i.e. more than 1.50, was measured and the function of the diameter d, is given for both size distribution constructed (Fig. 5, left). Then colloids CdS and Au and there is a clear clepenSEC was carried out (Fig. 5, centre) on Nucleosil dence. From these calibration plots the average 500 and Nucleosil 1000 (15-25 Frn) [26]. When particle size of unknown samples of the same smaller silica material was utilised, the gold sol material can easily be determined. The slopes of was irreversibly adsorbed on the column. There both curves are quite different_ Two reasons can are some samples with bimodal distribution in be given to explain this. Firstly, the eluent the SEC and only a single size population in composition was different and secondly, the true TEM. It shows that sometimes in the TEM less size of the solid particles is different from the frequent populations can be overlooked. For the effective particle size due to the electrical double wide-bore HDC of gold SOIS. a 1 mM sodium layer, which is formed at the solid-liquid intercitrate solution was used as the eluent, the same face by electrolytes of the solution. The thickas that used with SEC. Again, the same trend in ness of this layer is dependent on the solid the chromatogram shape from pure diffusion to material but also on the particular electrolyte. pure convection peak was observed (sample Under the conditions of Fig. 6 the electrical a--, e in Fig. 5, right). These parallel experidouble layer of the gold particles seems to be ments allow a good comparison of all three thicker and therefore diffusion plays a less immethods. portant role than in the case of cadmium sulphide. Calculation Finally it should be mentioned that the concentration of the colloid itself also has an effect In the past different approaches have been on the diffusion rate and therefore on the result used in the determination of the molecular of size determination. The higher the concen- Ch.-H. Fischer, M. Giersig I .I. Chromatogr. A 688 (1994) --- _- .- i ---- -_- ---- --_-- -mm-em- j. -P!lJed % 103 97-105 1 4- 0 zyxwvutsrqponmlkjihgf 104 Ch.-H. Fischer. M. Giersig i J Chromalogr. A 688 (1994) 97-105 2 . z II ET 0 1 I I I I 2 3 4 5 6 [CdS] / mM Fig. 7. Effect of concentration on the result of wide-bore HDC. A 5 mM cadmium sulphide sol was diluted stepwise down to 0.1 mM. Height ratio R = h,ihz of convection peak to diffusion peak (0) and the particle diameter dual, calculated from R as a function of CdS concentration (0). 0 5 10 diameter 15 20 25 30 / nm Fig. 6. Calibration plot for the wide-bore HDC of gold (a) and cadmium sulphide (0) soIs. It shows the ratio R = h, /h, of the height of convection peak over diffusion peak as a function of weight average particle diameter d,. determined by SEC or electron microscopy. respectively. Experimental conditions are given in the text. Comparison zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQ of the methods and conclusion TEM shows directly the size and shape of particles, but radiation damage can occur [3] and alter the original particle size. Furthermore, less frequent populations could be overlooked [26] and sometimes during sample preparation, smaller and larger particles separate to some extent by tration of the colloidal particles, the faster is diffusion, so that the statistics in an observed their diffusion between slowly and fast flowing part of the w-hole sample is not perfect any more. parts of the cross-section of the capillary and the Without a digital analyser the analysis is time more pronounced is the diffusion peak. This is consuming and tedious. SEC gives a statistically demonstrated with a sol containing 5. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFE lop3 M more accurate view of the true size distribution CdS, which was diluted with water stepwise as long as all colloidal material elutes. The down to 1 - lOPa A4 (Fig. 7). From the particular analysis is very fast. However, the chromatogheight ratio R of the chromatogram and the raphy columns have a limited lifespan. A new calibration curve (Fig. 6) diameters between 89 column must then be calibrated again by TEM, nm and 104 nm were calculated. Although the because the colloids in question are not stable influence of the concentration is not dramatic over long periods. Further drawbacks are that one should try to work always with identical adsorption can occur with very active colloids colloid concentrations as those used for the and the equipment is relatively expensive. Widecalibration. For higher concentrations the effect bore HDC cannot give direct information about is smaller than for more dilute ones. On the the size distribution. However, in the case of other hand highly concentrated colloidal solusmall semiconductor or metal particles where the tions might become unstable, or particle growth size-quantization effects allow an independent may occur. A 10-j M solution is recommended measure of the particle size, it can be done as a good compromise. Then a fivefold higher or indirectly via the chromatograms at various dea tenfold lower concentration would cause an tection wavelengths (see Fig. 3). But the wideerror of only 7 or 8 percent. respectively. bore HDC has many advantages. Whereas in the Ch.-H. Fischer, M. Gierslg / J. classic HDC equipment and handling are quite sophisticated and detection is difficult due to the small dimensions, the wide-bore HDC uses cheap, empty, standardised capillaries. Therefore one calibration can be directly transferred to any other capillary of the same type. Although this work was done with HPLC pumps, these sophisticated instruments are not necessary for this technique. A precise flow rate is not required because only the ratio of the peak heights is measured, and not the retention time. The method is fast and less calculation is necessary than in SEC. But most important is the lack of any packing material, the large surface of which often causes problems of reversible and irreversible adsorption. In a relatively wide capillary of 0.8 mm I.D. the surface does not play a significant role. Therefore the method is especially recommended for colloids with high surface activity. Chramatogr. A 688 (lY94) tion Assessment [lo] ton, C.A. Sci., [ll] C.A. Sci.. (I21 J.G. [l] A. Henglein, Top. Curr. Chrm., 143 (1988) 115. [2] H. Weller, Advanced M aterials. 5 (1993) 88. [3] E. Zeitler (Editor), Cryoscqy and radiation damages. North-Holland Publ. Comp.. Amsterdam. 1982. [4] J.J. Kirkland, J. Chromatogr., 185 (1979) 273. [S] Ch.-H. Fischer, J. Lilie, H. Weller, L. Katsikas and A. Henglein, Ber. Bunsenges. Phys. Chem., 93 (1989) 61. (AC’S Symposium Sci.. American Chemical Society, WashingDC, 1987, p. 256. Silebi and J.G. DosRamos, J. Colloid Interface 130 (1989) 14. Silebi and J.G. DosRamos, J. CoIloid Interface 133 (1989) 302. DosRamos and C.A. Silebi. J. Colloid Znterface 135 (1989) 165. Silebi and J.G. DosRamos, AZChE J., 35 (1989) [13] C.A. 1351. [14] M.E. Mullins and C. Orr, ht. J. M ultiphase Flow, 5 (1979) 79. [15] R.J. Noel, K.M. Gooding, F.E. Regnier, D.M. Ball, C. Orr and M.E. Mullins, J. Chromafogr., 166 (1978) 373. [16] F.M. Kelleher and C.N. Trumbore, Anal. Biochem., 137 (1984) 20. Anal. Hancock (1988) Chem., and R.E. 58 (1986) 3207. Synovec, Anal. Chem., 60 2812. [ 191 D.O. Hancock, C.N. Renn and R.E. Synovec, Anal. Chem., 62 (1990) 2441. [ZO] G. Gerhardt and R.N. Adams, Anal. Chem., 54 (1982) 2618. [21] J.T. Vanderslice, A.G. Rosenfeld and G.R. Beecher, Anal. Chim. Acru, [22] L. Spanhel, Am. [23] J.W. [24] [ZS] References and Characterization Series, Vol. 332), [18] D.O. The authors wish to express their thanks to Dr. Paul Mulvaney, Melbourne, for the preparation of the gold colloids as well as for fruitful discussions. The help with the laboratory work of Ms. U. Michalczik and the assistance of Ms. U. Bloek with the electron microscopy is gratefully acknowledged. 105 [6] Ch.-H. Fischer, H. Weller, L. Katsikas and A. Henglein, Langrnuir, 5 (1989) 429. [71 Ch.-H. Fischer, M. Giersig and T. Siebrands, /. Chromarogr., 670 (1994) 89. [8] Ch.-H. Fischer, J. Liq. Chromarogr., 17 (1994) 3593. [9: A.W. Thornton, J.P. Olivier, C.G. Smart and L.B. Gilman, in T. Provder (Editor), Particle Size Distribu- [I71 J. Pawliszyn, Acknowledgement 97- 105 [26] [27] [28] Chem. 179 (1986) M. Haase. Sot., 119. H. Weller 109 (1987) and A. Henglein, J. 5649. Slot and H. J. Geuze, Eur. J. Cell. Chem., 38 (1985) 87-93. C.R. Berry, Phys. Rev., 161 (1967) 611. A. Fojtik, H. Weller, LJ. Koch and A. Henglein, Ber. Bunsenges. Phys. Chem., 88 (1984) 969. T. Siebrands, M. Giersig, P. Mulvaney and Ch.-H. Fischer, Langmuir, 9 (1993) 2297. V. Murugaiah and R.E. Synovec, Anal. Chim. Acta, 246 (1991) 241. C.N. Trumbore, M. Grehlinger. M. Stowe and F.M. Kelleher, J. Chromatogr., 322 (1985) 443.