pp. 33-47 in: "Science of Whitewares"; eds. V.E. Henkes et al; The American Ceramic Society, Wesrerville, OH (1996) LIMITS IN PARTICLE SIZE CHARACTERIZATION Herbert Giesche, NYS College of Ceramics at Alfred University ABSTRACT The limitations and specific requirements of different particle size characterization methods are discussed. The theoretical background of each method is shortly summarized and limits in terms of minimum or maximum measurable particle size are discussed together with several experimental parameters which can influence the accuracy or which can cause systematic errors. The paper focuses on sedimentation and optical (light scattering) characterization methods. INTRODUCTION Particle size is one of the most important characteristics in ceramic processing. A precise analysis as well as control throughout all processing steps is essential for high quality production and low loss rates. A large number of publications dealt with specific aspects in sample preparation and the analysis itself. The precise and controlled preparation of a representative sample is the first and essential step in getting reliable analysis results. This sample preparation has to be as closely related as possible to the actual processing procedure or vice versa to the information and conclusions to be drawn from those measurements. Despite the importance of those steps, the present paper will not address those aspects but rather focus on general limitations and possible errors of the various groups of instruments or techniques. DEFINITION OF PARTICLE SIZE In most cases particle size is described as the size of a spherical particle, which would show the same behavior as the actual particle in the specific analysis. As shown below, a variety of size descriptions exist and one has to specify in each presentation of the results, which kind of particle size and how it was determined. Geometric properties: - Diameter, Chord Length (Si ), Feret- (XF ), or Martin- (XM ) Diameter - Surface Equal Sphere X = Surface S π - Projected Area X PM = 4 Area π (! non spherical particles are not always in a random position) - Volume Equal Sphere X V = 3 6 Volumeπ 18 η w g Sedimentation Equal Sphere (Stokes): XW = (ρ p − ρ fl ) g Si 10 0 10 0 80 80 60 60 40 40 20 20 0 1 10 10 0 m ass or volume % number % Moreover not only the precision or reliability of the analysis is important but also the way those results are presented. Some are quite obvious, like the difference between a cumulative and differential distribution, or using a linear or logarithmic scale on the size axis. However, there are some hidden differences which are not always indicated or realized. For example, the average particle size will change depending on whether a number-, area-, or volume- (mass-) distribution is shown. There is no general rule, which of those presentations to recommend, this very much depends on the preferences of the operator or the specific requirements of the process. As shown in fig. 1 the sample will have a vastly different mean particle size when plotted as a number- or volumedistribution. 0 10 0 0 Pa rt ic le S iz e / um Fig. 1: Micrograph (left) and cumulative size distribution (right) The situation gets even more complicated, since most real powders are not spherical. A shape factor can be introduced to account for this. One possible definition was provided by Wadell.1 the shape factor, Ψ, is defined as the surface-ratio of a volume equal sphere 2 XV ⎛ ⎞ over a surface equal sphere. Ψ = ⎜ X S ⎟⎠ ⎝ As can be shown easily the following rank of order is valid for non-spherical particles. XS = XPM ≥ XV ≥ XW IMAGE ANALYSIS Optical- or electron-micrographs can be used to determine the particle size of a sample according to the various definition as mentioned earlier. Even so this technique may provide much more information than just the particle size, there are a number of errors resulting from sample preparation or the analysis set-up. Fig. 2 demonstrates some of those errors, like the under-cut or the random or stable orientation of particles. Especially small particles may not be counted when they are covered by larger particles, or the measured particle size is larger for a sample with powder particles in a stable position compared to randomly oriented particles. In addition computerized image analysis can give faulty results depending on a correct alignment of the threshold value and/or different orientations of particles towards the screen pixels. The latter is especially true for needle like particles as demonstrated in Fig. 3. under cut, random, or stable Fig. 2 Errors due to particle position Influence of orientation: a) 22 pixels b) 13 pixels Fig. 3 Computer assisted image analysis SIEVE ANALYSIS Even with such a simple technique as the sieve analysis a number of errors have to be considered. Some of those errors are quite obvious, like damaged or uneven sieves, a too short analysis time, insufficient tapping, abrasion, breaking of particles, humidity, clogging, or agglomeration of the powder. Yet, in case the tapping frequency becomes too high, the passing-% of the sample likewise decreases. In the latter situation particles do not have enough time to pass through the sieve opening before the movement of the sieve will bounce them back. Thus for a fixed analysis time the amount of sample, remaining on the sieve, passes through a minimum, as shown in fig. 4. Further information about the correct sieve analysis is given in several reference papers.1-5 Fig. 4 Influence of time and vibration frequency on sieve analysis after F. G. Carpenter and V. R. Deitz2 SEDIMENTATION Using Stokes law,6 sedimentation experiments will provide information about the particle size distribution. Stokes law relates the gravitational, the buoyancy, and the drag force acting on the moving (settling) particle in a medium, as shown below. The greatest uncertainty in this equation is related to the drag force. The drag coefficient, Cw , the Reynolds number, Re, and the projected area of the settling particle can not be described sufficiently in all cases. The linear relation between Reynolds number and drag coefficient, Cw = 24/Re, is valid only in the laminar flow regime at Reynold numbers below 0.25. This relation is used for the basic Stokes equation as utilized in most sedimentation experiments. At higher Reynold numbers a transitional region and finally turbulent flow is observed as shown in fig. 5, where spherical particles were assumed. FG = VP ρ P g = π x3 ρP g buoyancy: FA = VP ρ fl g = π x 3 ρ fl g weight: drag force: FW = c W (Re) A P 6 6 ρ fl 2 with: VP : volume of particle ρ pl or ρ fl : density g: gravity constant x: particle diameter c W (Re): drag coefficient AP : projected area of particle w g : sedimentation velocity w 2g 1,000 100 transition range for: 0.25 < Cw < 2000 Cw 10 turbulent flow for Cw ≅ 0.45 laminar flow for: 1 Re < 0.25 Cw ≅ 24/Re 0.1 0.1 1 10 100 1000 10000 100000 Reynolds number Fig. 5 Drag coefficient as a function of Reynolds numbers for spherical particles The limitation in the range of Reynolds numbers thus limits the maximum particle size to be determined for a specific system, since ordinarily the basic Stokes equation, valid for the laminar regime, is used without further corrections. Table 1 summarizes those limitations for a variety of materials. Table 1: Laminar flow restriction, maximum particle diameter at Re=0.25 material X max = 3 4.5 η (ρ p − ρ fl ) ρ fl g 2 flour quartz alumina lead density [g/cm3] 1.5 2.65 3.96 11.4 Xmax water 97.2 65.3 53.5 35.3 in µm air 43.7 36.2 31.9 22.2 Yet the measurable range can be extended by using various equations to correct for the non linear relation between Cw and Re in the transition range. Table 2 provides a brief overview of equations, the recommended range, and the associated error for Cw for each equation over this range. In order to expand the measurable range a different dispersion media could also be used. Having a higher viscosity liquid expand the range to larger particle sizes. One example is shown in the following diagram (fig. 6). Table 2: Correction formulae for the transition range7 formula Cw = ... Re-range max. error % (24/Re) +2 < 10 -4/+4 (24/Re) +1 <102 +14/-40 (24/Re) + 0.5 < 105 +32/-40 (24/Re)+(2.8/Re1/4) 0.1 < Re < 4 103 +2/-8 (21/Re)+(6/Re1/2)+0.28 0.1 < Re < 4 103 +4/-4 (24/Re)+(3.6/Re0.313) 0.1 < Re < 103 +8/-5 ((4.8/Re)+0.63)2 0.1 < Re < 103 +20/-1 (24/Re)+(4/Re1/2)+0.4 < 2 105 +6/-6 Sieve Analysis % Passing 150 100 50 0 0 200 400 600 Particle size / um Volume % 100 80 60 40 20 0 0 160 320 480 640 size / um Fig. 6 Sieve analysis of a spray-dried powder compared with sedimentation analysis in ethylene glycol. The sieve analysis and particle size distributions determined from sedimentation analysis of a spray dried powder correlated well. Ethylene glycol was used as the dispersion medium, having a viscosity of 16.1 mPa s and a density of 1.1101 g cm-3. The powder immersed in the liquid had a density of 1.93 g cm-3, and thus particle sizes of up to about 0.5 mm could be measured in this way. In addition to limitations by the laminar flow regime or the maximum particle size, the shape of the settling particles has a noticeable influence on the sedimentation velocity. The following equation relates the sedimentation velocity with other particle parameters by using the previously defined shape factor, (XV /Xp )2. 4 (ρ p − ρ fl ) g Wg = 3 ρ fl 2 ⎛ XV ⎞ XV ⎜⎜ ⎟⎟ ⎝ X P ⎠ CW (Re( X P )) Several other factors are less easily described by correction formulae. The most prominent effects are temperature and concentration or density gradients in the system, or the increased Brownian motion of small particles. For example, the particle concentration not only decreases the actual sedimentation velocity at high solid loading (hindered sedimentation or particle particle interaction), at intermediate concentrations some models even predict a faster sedimentation due to particles following the draft of other particles and thus increased sedimentation velocity.8 In addition, several instrument specific factors have to be considered. In general the particle concentration and vice versa the particle size distribution is determined as a function of time and position by means of light or x-ray absorption, osmotic pressure, or the weight of the settled material. However, the detector signal is not always linear Fig. 7 Extinction coefficient of lime stone and quartz as a function of size for whithe light. proportional to the particle size at a given concentration. Especially light detectors require major correction at or particularly below particle sizes of 1 µm. Fig. 7 provides one example of the latter effect for quartz and lime stone. For example, 6 µm quartz particles will cause a 3 times higher light absorption compared to the same amount of 2 µm particles. As the particle size increases above 10 µm, those corrections become relative insignificant. However in the micron and submicron size range the latter effect can cause serious mistakes in the calculated particle size distribution. In order to avoid this effect some instruments use x-ray detectors which are independent of the particle size in this range. X-ray detectors, however will need higher particle concentrations than light detectors and this can influence the sedimentation velocity as demonstrated earlier. In case the analysis-material contains strong x-ray absorber, like elements with a high atomic number, the x-ray detection principle is ideally suited for those measurements down to particle sizes of 10 nm. Besides the fat that Brownian motion becomes significant in this range. INTERACTION OF LIGHT WITH PARTICLES Particles can interact with electromagnetic fields in various ways. The Coulter Counter is one of those principle, however, it will not be covered here in further detail. It is based on the change of electrical resistivity in a high frequency field when a particle passes through a narrow connection hole placed between both electrodes. From the change in resistivity the particle volume and thus the particle size can be calculated. The interaction of light with particles is another principle which can be used to determine the particle size.9,10 Depending on the size of the particle and the wavelength of the light, various effects can be observed. Large particles with a large size-wavelength ratio α = π X / λ >> 1 display Fraunhofer diffraction, whereas Rayleigh-Scattering dominates for small particles (α << 1). For the transition range of 0.1 < α < 10 the Mie theory11 will describe the light scattering behavior, Yet the latter theory is relative complicated and has been solved only for spherical particles. The scattering intensity varies largely with scattering angle, particle size, refractive index of the materials, and the polarization direction of the light-source / -detector. Fig. 8 shows one example indicating the complex behavior in the intensity distribution. Fig. 8 Mie Polarogram for λ = 0.4 µm, α =5, and refractive index m=1.46+0i For relative small particles the Rayleigh- or dipole-scattering is observed, also called Tyndall effect, and the corresponding formula is relative simple. 2 4 6 I =π X I0 8 λ4 ⎛ m2 −1 ⎞ X6 m ⎟⎟ 1 + cos 2 Θ ∝ 4 with m = p ⎜⎜ 2 m fl λ ⎝m + 2⎠ The angular dependence of the intensity is symmetric and independent (within the given limits) of the particle size. A particle size of 20 nm is about the upper limit when using visible light.12 For larger particles the following formula will describe the scattering behavior: I ( ) X4 X2 ⎛ π ⎞ 4 ⎛ J1 (α Θ ) ⎞ X ⎜2 = ⎟ ∝ 2 (for Θ = 0) or ∝ 2 (for Θ ≠ 0) I0 ⎜⎝ 4 λ ⎟⎠ λ λ ⎝ αΘ ⎠ with J1 (α Θ ) = Bessel function 2 2 Theory predicts very wavy curves and a larger opening of the aperture angle is to be used in order to smoothen the data. The smaller the particles the more it is recommended to measure in the forward direction since this will provide the highest scattering intensities. However, the X4 dependency will limit the measurable range in the forward direction. In addition the scattering intensity will become independent of the refractive index of the material for measurements in the forward direction. Fig. 9 shows the light scattering intensity as a function of particle size for monochromatic and white light.7 Two facts become evident from those graphs. First, a monochromatic light source and a small aperture size will cause a very noisy curve due to the many maxima and minima in the diffraction curve (left diagram). Using a broader wave length spectrum and a larger aperture size will smoothen the data, as shown in the right diagram, since the peaks of the different wavelength and for the different observation angles will cancel out those sharp maxima. Secondly it is apparent that the refractive index of the material has only a minor effect on the curves, however, absorption properties of the material, indicated by the imaginary part in the refractive index (curve 2), will cause a pronounced down shift of the scattering intensity curve. The extinction coefficient, which is the combination of light scattering and absorption has the same behavior. Thus in order to smoothen the data a large aperture and a broad wavelength range is recommended. An example for quartz and lime stone was shown earlier in fig. 6. Another effect which has to be considered seriously in light scattering experiments is the coincidence error. In principle the diffraction pattern and intensity distribution is determined for single particles. Thus, in case the particle concentration becomes too high, several particles will be measured at the same time and correspondingly critical errors can occur. Fig. 10 demonstrates this relation in a graph.7 The importance of small measuring volumes and/or low particle concentrations become evident. A coincidence error of 5% should be seen as an upper limit. Most light scattering instruments limit the actual measurement volume by focusing the light source and reducing the detector aperture as shown in the schematic diagram in fig. 10. 10 10 9 Intensity 0 deg 7 15 10 5 30 90 10 3 10 1 10 = 0.4 m m = 1.46 -1 0.1 1 10 100 particle size /um Fig. 9 Light scattering intensity for monochromatic- (left) and white-light (right) Fig. 10 Maximum number of particles, Cn , as a function of the coincidence error, fk , for various measuring volumes. Moreover each sample has to be prepared and analyzed carefully and the results should be critically evaluated. Fig. 11 gives an example of a spray dried powder. Since it disintegrated into the primary particles when dispersed in water, propanol or other organic liquids were used as dispersion medium. Experiments in test tubes indicated that the powder was well dispersed and that it did not disintegrate with time. However, when measured with a light scattering instrument, Microtrac FRA, Leeds Northrupp, a constant shift towards smaller particle sizes was observed, clearly an indication of powder disintegration. The pumping unit of the instrument caused such high shear forces that the particles broke apart. 100 3 min. 80 60 40 20 0 100 80 21 min. 60 40 20 0 100 80 37 min. 60 40 20 0 100 80 62 min. 60 40 20 0 0.13 1.06 8.5 68 543 Particle size / um Fig. 11 Disintegration of spray-dried powder under shear. Particle size as a function of time. CONCLUSIONS Various particle size characterization techniques have been described and critically analyzed. The limitations of each method were demonstrated. However a proper knowledge of those limitation can be used to modify the analysis conditions in such a way to accomplish a particle size analysis even so it is out side of the regular range. Examples also demonstrated that one should always try to characterize samples with different techniques, since systematic errors are not always obvious. Finally the operator has to judge, which sample preparation will provide the most reliable results. In addition it has to be stressed that different method will present results in different ways and especially the kind of distribution, e.g. a number-, area-, or volume-distribution should always be clearly labeled. Again, the operator has to decide, which presentation is the most useful to characterize or control the process. REFERENCES 1. 2. 3. H. Wadell; J. Geology, 40 (1932) 443-51 F. G. Carpenter, V. R. Deitz; J. Res. of NBS, 45 (1950) 328-45 K. T. Whitby; Symp. on Particle Size Measurement, Spec. Tech. Pub. No.234, ASTM 1958, 3/25 4. H. Rumpf; Staub 20 [8] (1960) 253-66 5. K. Leschonski; Proc. of 3rd Particle Size Analysis Conf., (1978) 205-17, HeydenVerlag, London 6. C. G. Stokes; Trans. Cam. Phil. Soc. 9 (1951) 7. K. Leschonski; Grundlagen und moderne Verfahren der Partikelmeßtechnik, GVT Kurs, Clausthal (1986) 8. B. Koglin; Chem.-Ing.-Techn. 44 [8] (1972) 515-21 9. M. Kerker, The scattering of light, Academic Press, New York (1969) 10. H.C. van de Hulst; Light scattering by small particles, Wiley, NY (1957) 11. G. Mie; Ann Phys 25 (1908) 377-445 12. R. Brossmann, Ph.D. Thesis, Karlsruhe (1966)