Errors particle size, sampling

Equation (2.28), being statistical in nature, requires a large number of particles to be measured, especially if the spread of particle size is wide. The possibility of error from this source is stressed by Arnell and Henneberry who found that in a particular sample of finely ground quartz, two particles in a total of 335 had a diameter about twenty times the most probable diameter, and that if these were overlooked the calculated value of A would be nearly doubled. 

For determination of the aerodynamic diameters of particles, the most commonly apphcable methods for particle-size analysis are those based on inertia aerosol centrifuges, cyclones, and inertial impactors (Lundgren et al.. Aerosol Measurement, University of Florida, Gainesville, 1979 and Liu, Fine Paiiicles—Aerosol Generation, Measurement, Sampling, and Analysis, Academic, New York, 1976). Impactors are the most commonly used. Nevertheless, impactor measurements are subject to numerous errors [Rao and Whitby, Am. Ind. Hyg. A.s.soc.]., 38, 174 (1977) Marple and WiUeke, "Inertial Impactors, in Lundgren et al.. Aerosol Measurement and Fuchs, "Aerosol Impactors, in Shaw, Fundamentals of Aerosol Sci-... 

Approximate Values in Units of Percent The particle size distribution in a total dust sample is unknown and error in the XRD analysis may be greater than for respirable samples. Therefore, for total dust samples, an approximate result is given. (4) Nondetected A sample reported as nondetected indicates that the quantity of quartz (or cristobalite) present in the sample is not greater than the... 

A bulk sample is the last choice and the least desirable. It should be submitted "for laboratory use only" if there is a possibility of contamination by other matter. The type of bulk sample submitted to the laboratory should be cross-referenced to the appropriate air samples. A reported bulk sample analysis for quartz (or cristobalite) will be semi-quantitative in nature because (1) The XRD analysis procedure requires a thin layer deposition for an accurate analysis. (2) The error for bulk samples analyzed by XRD is unknown because the particle size of nonrespirable bulk samples varies from sample to sample. 

If weighing and grinding, which require as much time and labor as the actual analyses, could be eliminated and the samples analyzed on an as-received basis, one technician could be omitted. However, the accuracy would be decreased owing to the combined effects of greater matrix errors, no diluent, and nonuniform particle size. In some instances such a sacrifice in accuracy might be warranted by a gain in economy and rapidity of analysis. 

The mobile phase is usually seleeted by trial-and-error guided by prior experience or by performing preliminary analytieal separations of the sample in a saturated ehamber. PLC separations will be inferior to analytical TLC separations using the same mobile phase beeause of the thieker layer, larger particle size, and overloaded sample eonditions used for PLC. A good general rule is that analytical TLC should achieve separations with least 0.1 Rf value difference if the PLC separations are to be adequate with the transferred mobile phase. Isocratic development is usually used, but gradient development has been applied in certain situations for increased resolution. 

IAEA/AL/095 1996). Both methods were suitable, but needed to be repeated several times to produce the small particle size that was required. The particle size reduction, e.g. of IAEA-395 from a median size of 30 pm to 3.5 pm, improved the homogeneity of elements. Sampling constants (the minimum mass that can be used to achieve a random error of i % at the 65 % confidence level) improved from a factor of 1.2 for Sc, up to a factor of 800 for Au. The average improvement was about a factor of 2-10. (Ni Bangfa et al. 1996). From these initial experiences, it is dear that preparation of reference materials is critical with respect to the final particle size distribution, which should exhibit a low maximum (<50 pm) and a narrow range in particle sizes. Milling techniques to meet such criteria are available today, and materials that show intrinsic uniformity are particularly suitable to achieve the desired properties. 

Figure 2.1.6 shows the results of such a continuous synthesis process. It shows the variation of the mean particle size during the experiment. The error bars indicate the standard deviation of the particle size distribution of each sample based on the transmission electron micrographs (number distribution). The experiment was performed under the following conditions (A) ammonia, water, and TEOS concentrations were 0.8, 8.0, and 0.2 mol dm-3 7", = 273 K, T2 = 313 K total flow rate was 2.8 cm3 min-1 100 m reaction tube of 3 mm diameter residence time 4 h and (B) ammonia, water, and TEOS concentrations were 1.5,8.0, and 0.2 mol dm- 3 Tx = 273 K, T2 = 313 K total flow rate was 8 cm3 min-1 50 m reaction tube of 6 mm diameter, residence time 3 h. Further details and other examples are described elsewhere (38). Unger et al. (50) also described a slightly modified continuous reaction setup in another publication. 

The process of selecting n objects for a lot of size N is called sampling. The n objects are the sample and the total lot size N is the population from which the sample is taken. This lot may be in one mass or distributed among M separate containers. Randomness is critical to sampling because a random sample has minimum errors ideally all the error should be due to the intrinsic variability of the material and the variability due to sampling should be minimized. An attribute is the property of interest that is going to be measured, be it particle size, composition, etc. 

Figure 8 is a plot of r89M4 vs. particle size. We define rlt j as the ratio of equivalent fissions of i to equivalent fissions of j. These ratios have several advantages (1) they are independent of the percent active particles, (2) they are independent of errors in weighing the samples, and (3) in a truly representative sample of weapon debris, rltj = 1 for all i and j. Figure 8 presents two surprises (1) for Zuni, r89,144 is generally lower in the more distant sample, and (2) the Zuni 85-km. curve below 150/x and the Bravo curve above 50/x have positive slopes (and appear to... 

For the reasons described above, the droplet size distribution of the same emulsion measured on different laser diffraction instruments can be significantly different, depending on the precise design of the optical system and the mathematical theory used to interpret the diffraction pattern. It should be noted, however, that the most common source of error in particle size analysis is incorrect operation of the instrument by the user. Common sources of user error are introduction of air bubbles into the sample, use of the wrong refractive index, insufficient dilution of emulsion to prevent multiple scattering. and use of an unclean optical system. 

---