Chapter 3 Quality Aspects in Particulate Analysis

Transcription

1 Chapter 3 Quality Aspects in Particulate Analysis Contents 3.1 Introduction Relevant parameters for Product Quality Quality Aspects in Particulate Analysis Design of a PSD Analysis Procedure Definitions, Abbreviations and Symbols Bibliography Annex 3A: Protocol and Standard Operating Procedures for Adequate Characterization of Particulate Products Annex 3B: Typical Properties of Particulate Materials Abstract : The quality of particle size analysis for the purpose of control of product quality consists of several parts. First, the most relevant parameter(s) have to be chosen to represent the quality of a specific product in a specific application. Some background is given, but an optimum answer to any new product with new properties can only be found by thorough investigation, knowledge and experience. Secondly, the quality aspects of both technique and total procedure for measurement have to be considered in relation to the requirements set for product performance and an adequate procedure selected. The quality aspects of a technique include measurement range, analysis time, precision, noise, drift, bias, accuracy, resolution, sensitivity, lower detection limit, traceability, and investment and labor costs. A quantitative description for these aspects is given together with the major error sources. Altogether, it is advised to set up a protocol for each PSD measurement method in relation to the given application(s), in which all choices made are clearly described. Good practice requires: adequate quality for selected PSD parameter(s) realistic evaluation of quality versus costs adequate design of the total procedure use of existing written standards statement on method validation Standard Operating Procedures and Protocol adequate reporting of results H.G. Merkus, Particle Size Measurements, DOI / _3, Springer Science+Business Media B.V

2 44 3 Quality Aspects in Particulate Analysis 3.1 Introduction In the previous chapter it has been shown that many equivalent particle sizes are available to represent the size of a non-spherical particle, which in turn depends on the measurement principle. Only for spheres, the same diameter should result from measurements, regardless of measurement principle. Most often, however, particulate materials contain non-spherical particles as well as show a particle size distribution (PSD). For each application, the most relevant parameters have to be selected in order to represent product performance. In addition to particle size, often the quantity of particles in a given size range or particle shape or particle porosity is relevant for the performance of a product. Always, quality assessment of a process stream or product requires determination of the relevant properties within given limits of confidence with respect to precision. In turn, this requires procedures and instruments that have adequate precision and sensitivity for the relevant properties as well as competent operators. These aspects will be treated in this chapter. 3.2 Relevant Parameters for Product Performance Very often, one or a few parameters of PSD, particulate concentration, surface area or particle shape are used to describe the characteristics of a process stream or the performance of a product in a given application, i.e., its property function. In Chap. 1 already some examples have been given for relationships between product performance and particle size. Sometimes, literature and theory can be used as such or to set up a series of measurements in order to identify this relationship in a quantitative way [1, 2]. Sometimes, such experimental studies have to be based on experience or intuition and the best parameters selected, preferably based on robust regression analysis. Always, acceptance limits for these parameters should be determined in relation to product performance and, thus, the precision for their determination has to be established. Many candidates for these parameters exist, for example: particle size: a weighted mean size or moment of the PSD, or D 10, D50, D, or 90 maximum size (for a given type of particle size) PSD width: D90 /D, (D -D )/D, or a characteristic width parameter of a modeled distribution (See Chap. 2) absolute or relative amount in a given size interval or in between a given size and the lower/upper size of specified technique: e.g., PM10 or 63 + μm sieve residue volume- or number-based particulate concentration specific BET surface area per unit mass characteristics of the pore size distribution: e.g., size(s) and quantities of dominant pores pore volume characteristics: e.g., pore volume in a given pore size interval per unit mass particle shape: qualitative, e.g., (near-) spherical, cubic, angular, fiber, flake; or quantitative, e.g., aspect ratio, fractal dimension, number percentage of spheres.

3 3.2 Relevant Parameters for Product Performance 45 Below, some background is given to supplement intuition (See also Annex. 3B). Generally, the behavior and performance of particulate products depend on a complex mixture of forces, one of which may be dominant. In dry powders, usually the attractive interparticle forces are of the Van der Waals type. These forces are relatively strong for particles smaller than about 10 μm and for particles having flat surfaces. Cohesivity of fine and ultrafine powders and adsorption of such small particles to the surface of large ones are mainly caused by these attractive forces. Van der Waals forces become insignificant for diameters larger than about 50 μm. Then, the particles show up as individual entities and the powder can flow and segregate easily during vibration under influence of gravity by inertial effects. Presence of very small amounts of liquid at the surface of particles, for example due to hygroscopicity or capillary forces, may lead to liquid bridges in between the particles, which also hold them strongly together. This is also true for particulate pastes, which contain more liquid but still hold a high concentration of solids. Even for large particle sizes, such mixtures are very cohesive and have no tendency to segregate. In particulate/liquid mixtures having lower solids concentration, several phenomena can be observed. One is sedimentation due to gravity, which is dominant and fairly fast for particles larger than about 10 μm and a density larger than about 2,000 kg/m 3. Up to about 100 μm, sedimentation is following laminar flow rules; for larger particles, it becomes turbulent and different rules are followed. For particles smaller than about 1 μm, Brownian motion (diffusion) becomes important. This movement may bring them in close contact and, in absence of repulsive forces, the particles will be attracted to each other again by Van der Waals forces and they will agglomerate or flocculate. This can be counteracted by giving the particles some electric charge by ionization or by adsorption of highly charged ions or of molecules with long chains. Then, electrostatic and/or steric repulsive forces can be effective to stabilize the particulate dispersion. Such stabilization is often applied for colloidal dispersions (See Chap. 5). Drying of particulate/liquid mixtures by evaporation of the liquid will lead to very strong bonds between the particles, especially if the liquid contains some dissolved material, which leads to solid bridges after evaporation of the liquid. Separation of particles from gaseous process streams is often based on centrifugal forces (in cyclones), impaction/sieving/diffusion/interception (for filtration) or electrostatic forces (in electrostatic precipitators). For liquid process streams most often cyclones and filters are used. Particle size together with mass and electrical charge determines the ease of particle separation in these processes in relation to force strength and width of separator channels or filter openings. But other process parameters, such as pressure drop over a filter, may be important as well. There are two main causes for the presence of aerosols in ambient air. First, there are natural causes, e.g., condensation of water vapor leading to fog, wind erosion of the sea and of deserts, eruption of pollens and seeds from plants and eruptions of ash and smoke by volcanoes and forest fires. The second group of causes is anthropogenic, e.g., automobile exhausts, particles blown away from roads or unprotected

4 46 3 Quality Aspects in Particulate Analysis storage heaps of particulate material by transport and handling activities. Emission is often caused by local (wind) turbulence and strongly related to particle size. Small particles are emitted and cause dusting more easily than large particles. Deposition most often results from gravity sedimentation or with the aid of rain. Large, heavy particles are usually only transported by wind over small distances (less than 100 m) in contrast to small, light particles, for which aerosols may be very stable. Inhalation and deposition of aerosols is controlled by particle size (aerodynamic diameter) through mechanisms of sedimentation, impaction, diffusion, and interception. The ratio of surface area to volume of particles plays an important role in their reactivity, flammability, explosivity, dissolution rate, and drying rate. This ratio is inversely related to particle size, since it involves the diameter squared over the diameter cubed. The ease of wetting of the surface or diffusion limitation to and from the surface may also play a role and, thus, limit the effect of the area to volume ratio. The color of pigments is primarily governed by their chemical and crystal structure; they determine the absorption of the different wavelengths of visible light. But also particle size and shape are important for pigment particles. For, maximum hiding power, gloss and efficiency require a particle size that is about equal to the wavelength of the visible light. Much larger particles may stick out of the paint layer (nibbing) and, thus, ruin its gloss. Relatively few large particles will cause this effect. Much smaller particles show much less scattering and absorption, which leads to increased transparency and low pigment efficiency. Similarly, visibility of particles in air or liquids is influenced by particle size. Here too, maximum visibility per volume of particles is reached when the particles have a diameter of about 0.5 μm, in the middle of the visible light region. This is not only important for limited vision in fog but also for beer filtration. Table 3.1 gives some typical concentration data for different products as background. Note the difference in dimensions in air (per m 3 ) and liquids (per liter), which is related to the typical length (volume) of observation. Many production processes produce materials having a medium to very broad PSD. Examples are comminution, crystallization, polymerization, and spray drying. Only at some special, well-controlled conditions (ultra)narrow PSD s or even monosized parti- Table 3.1 Typical concentrations of particles in liquids and air [3, 4] Particle size about m for aerosols and m for liquids Clean room 10-6 μg/m 3 (< 1 particle/m 3 ) Rural area 20 μg/m 3 Indoor 40 μg/m 3 Urban area 100 μg/m 3 Just visible dust emission 1,000 μg/m 3 Uncontrolled stack emission 10 4 μg/m 3 Parenteral fluid 100 μg/l Typical ESZ medium 1000 μg/l Pilsner beer 1,000 μg/l Just visible water turbidity ~ 10 4 μg/l White beer ~ 10 5 μg/l

5 3.2 Relevant Parameters for Product Performance 47 cles can be produced (e.g., emulsion polymerization). Thus, normally the raw products have to be accommodated to the required PSD width by some classification process, for example by means of sieves or cyclones. But even then, a narrow PSD width of the product (D 90 /D 10 ratio of ) is usually the best that can be attained. Furthermore, broadening of the PSD may occur during transport of concentrated streams, e.g., during conveying on belts or in air streams, where high wear on the particles may result in attrition and breakage. Coarse and very coarse particles with sizes larger than about 100 μm are more easily comminuted than finer particles. For medium or coarse particles with sharp edges, the edges are fairly easily removed resulting on the one hand in rounded particles of about the same size as the original ones and on the other hand in (ultra)fine particles with sizes smaller than a few micrometers. Flakes and crystalline needles and other brittle particles will usually break down to smaller particles, having a (very) broad range of about 10 90% of the mass or 2 95% of the size of the original particles. The resulting fine particles from attrition or breakage are often the cause of dusting problems of the final product. If dry powders are to be used in liquid suspensions or pastes, then also good wettability and dispersion are essential. Here, the surface chemistry of the particles plays a dominant role. For example, products with a non-polar, hydrophobic surface (e.g., polyethene and polypropene) can only be wetted by low surface tension liquids. Products with a polar, hydrophylic surface (e.g., with hydroxyl groups) can also be wetted by water, despite its high surface tension. Surfactants can be used to lower the surface tension of water and, thus, facilitate wetting of non-polar compounds. Typically, it is not desirable that liquids used in the production of suspensions cause the particles to dissolve or swell. In general, the rule is like dissolves like. It means that the chances for dissolution or swelling are greatest if the liquid and the solid have a similar chemical nature. For example, sugar will dissolve in water, starch will swell in (hot) water and polystyrene particles will swell in toluene. High molecular weight and cross-linking in polymers may reduce the degree of swelling. More information on wetting and dispersion of powders in liquids is given in Chap. 5. One problem with product performance is that often a compromise is required between different properties. For example, detergent powder should be easily dosed, transported and dissolved, but show no dusting in view of inhalation hazards. Toner quality is determined not only by good adsorption to paper and sharp lines during copying, but also by absence of dust. Concrete and ceramic strength is on the one hand dependent on easy particle movement during processing and on the other hand on high particulate density before drying. Pharmaceutical ingredients should be dosed and pelletized easily and show rapid dissolution behavior but neither segregation nor dusting. In general, production, classification and transport have to be optimized. The limitation is set here by the costs of the product. Consequently, for optimum product performance and costs, compromises have to be sought in particle size, PSD width, shape, porosity and/or concentration. This usually results in more than one specified parameter for particulate mixtures. Whenever possible, the relationship between product performance and PSD parameters should be quantified in a mathematical model. In view of costs, use of a large number of specified parameters should be avoided. It is advised to use robust regres-

6 48 3 Quality Aspects in Particulate Analysis sion analysis to find an appropriate mathematical model for this property function of a particulate material [4]. Principal Component Analysis (PCA) offers a good possibility (See Chap. 20). 3.3 Quality Aspects in Particulate Analysis In the selection of relevant size (distribution) parameters, adequate quality of their measurement is also an issue. In general, the following aspects of quality can be discriminated (See also Chaps. 17 and 20): Measurement Range for Particle Size The measurement range of a method or technique describes the range of particle or pore sizes for its application. It is essential that it covers the sizes that influence product performance. Typical overall measurement ranges for different PSD techniques are illustrated in Fig Often, the range required in a single measurement is (much) narrower than the range of the technique/instrument. Concentration Range for PSD Measurement The concentration range of a method or technique describes the typical concentration required for adequate PSD measurement. Low concentration capability of a technique is required in cases where the typical particulate concentration is low. High concentration capability during PSD measurement is especially important for in-line analysis of concentrated process streams and in situations where no dilution can be applied (See also Chap. 14). Often, the particulate concentration has to be accommodated to the requirements of the technique used for the measurement, which is only possible in online and off-line measurements. In general, dilution of a process stream or dispersion of the product before PSD measurement would be allowed, provided that this does not change the PSD by breakage, swelling, dissolution, or clustering. The typical concentration range of PSD instruments can be classified as follows: - Extremely low < 0.001% (v/v) - Very low % (v/v) - Low % (v/v) - Medium 0.1 5% (v/v) - High 5 20% (v/v) - Very high > 20% (v/v)

7 3.3 Quality Aspects in Particulate Analysis 49 Fig. 3.1 Typical over-all measurement ranges of PSD techniques

8 50 3 Quality Aspects in Particulate Analysis Analysis Time High speed of analysis is especially important if the results are to be used for process control or monitoring or for a decision on acceptance or rejection of a product batch (See also Chap. 14). Typically, the characteristic time for analysis should be about an order of magnitude smaller than the characteristic time of the process studied. For example, the analysis of changes of droplet size in a spray may require times of (milli-)seconds, the analysis of stable sprays or dispersed dry powders seconds and the analysis for monitoring of production processes minutes. Also, the labor time involved is important. The analysis time depends on three factors: 1. The time required for the actual measurement of a sample in order to reach a meaningful result. This time ranges from milliseconds to hours. Long times require stable conditions in and around the equipment. Typically, only little labor is involved, sometimes with help of automation. Only this actual measurement time is mentioned in the stated quality aspects of the different techniques. 2. The time required to accommodate the sample for measurement (sample splitting, dispersion, dilution) and to set-up and clean the equipment. Typically, this part requires about 5 30 min for an off-line analysis, with usually less than 5 min labor. 3. In off-line analysis, the sample has to be transported to the laboratory and the analysis results reported to the plant. The time involved strongly depends on the distance from plant to laboratory and on the degree of automation. Noise Noise is defined as random change of detector or instrument signal around a mean value with a short time constant relative to the measurement (usually < 0.1 s). In general, noise is to be discriminated from real signals coming from particles in the sensing zone of an instrument. Thus, large noise amplitudes are related to a poor detection limit for particle size or concentration of an instrument (See Fig. 3.6) or low precision of the resulting PSD. Apart from internal instrument sources, poor electric contacts or poor grounding of the instrument, electric apparatus in the close vicinity or strong vibrations may generate noise. The latter aspects are especially relevant in industrial areas. Drift Drift is defined as change of detector or instrument response with time at large time constant relative to the measurement (usually > 1 min). It may show up as

9 3.3 Quality Aspects in Particulate Analysis 51 systematic changes of particle size (distribution) as well as of concentration in time. Drift is usually related to temperature changes of the instrument or ageing of the detector(s). For optical instruments, fouling of lenses and cell windows can cause drifting. Drift can be detected and its effect limited by instrument qualification and calibration (See Chap. 17). Precision Precision is defined as the closeness of agreement between independent measurement results of a given property by the same method (due to random errors). Usually, repeatability and reproducibility are distinguished. Repeatability is when the same operator uses the same instrument under identical (stated) conditions within a short period of time. For PSD measurements, often two different cases can be discriminated. One is for an instrument or technique. Here, the same sample aliquot (e.g., stable suspension) is being measured for a given short period of time (See also Sect. 20.2). It tests the instrument. The other case concerns repeatability of the procedure. Then, the same analyst uses the same sample batch, but for each analysis a new aliquot is prepared and dispersed before analysis. Thus, sampling and dispersion uncertainties are added to the technique uncertainties, the contribution of which may be significant (See Chaps. 4 and 5). In this type of repeatability testing, the quality of operators to execute a PSD analysis by a given procedure may also become evident. It will be clear that all contributions to the overall error need to be reduced to an acceptable level. Reproducibility is when different operators and different instruments (of the same type) in different surroundings are involved, while using the same method. The results of a series of measurements of a certain parameter usually follow a normal distribution (See Chap. 20). In relation, the measure for precision is standard deviation s or σ (absolute) or coefficient of variation v (relative). A small standard deviation or coefficient of variation corresponds to high precision (See Fig. 3.2). Adequate precision both repeatability and reproducibility is most relevant for particle size measurement. This precision should be in agreement with the precision ,0;0,1 1,0;0, ,0;0,1 1,0;0, Fig. 3.2 Illustration of precision in a normal distribution (σ 0.1 and 0.03, respectively); (left) differential curves, (right) cumulative curves

10 52 3 Quality Aspects in Particulate Analysis required for adequate quality assessment of a given product. Note that over-specification may cause increased costs. Poor precision is usually the result of application of different operational conditions, absence of an optimized standard operating procedure (SOP) or a high noise level in the instrument. Note that reproducibility is generally worse than repeatability as more uncertainties are involved. Figure 3.2 also illustrates the differences between narrower and broader PSD s. The quality for repeatability, expressed as coefficient of variation of independent measurements in the range of D 10 -D 90 (and concentration), can be classified as follows: Instrument Method Method D 10 D D D Amount in size class,% (v/v) - Excellent < 0.3% < 1% < 0.2% - Good % 1 2% % - Medium 0.5 1% 2 5% 0.5 2% - Low 1 2% 5 10% 2 4% - Very low > 2% > 10% > 4% Note that low precision in some cases may suffice for adequate characterization of product performance, viz. if product performance shows low sensitivity to changes of the PSD. Bias Bias is defined as the difference between the mean value, coming from a set of test results, and an accepted reference value. It represents the systematic measurement error. Both absolute and relative distances are used as a measure. In some techniques, drift may lead to bias. In techniques, where calibration is performed to relate signals to particle sizes, poor quality of calibration may lead to bias. An example of bias is given in Fig It will be clear that bias is more easily seen for products having a narrower PSD ,0;0,1 1,3;0, ,0;0,1 1,3;0, Fig. 3.3 Illustration of bias (modes at 1 and 1.3, respectively); (left) differential curves, (right) cumulative curves

11 3.3 Quality Aspects in Particulate Analysis 53 The quality of bias of PSD techniques/methods, expressed as percentage difference between assigned D 50 and mean D 50, coming from at least three independent measurements, including sampling, dispersion and measurement, can be classified as follows: - Excellent <1% - Good 1 2% - Medium 2 5% - Poor 5 10% - Very poor >10% Bias is, for industrial particulate mixtures, usually only relevant if different companies e.g., producer and user test product specifications with different instrument types and/or methods. The certified data of reference materials, of course, are not supposed to have a bias for the stated certification technique(s). Accuracy Accuracy is defined as closeness of agreement between a test result and an accepted reference value. Thus, it includes both random and systematic measurement errors and is a combination of repeatability and bias. Note that the idea of bias and accuracy in particle size analysis is influenced by the concept of equivalent sphere diameters, since the same non-spherical particle may show different values in different orientations when measuring with different sizing techniques (Chap. 2). Thus, reference values for non-spherical particles should come from the same technique as the value under consideration. Resolution Resolution describes the ability of a measuring device to distinguish meaningfully between closely adjacent values of size. In this book, I will use the (estimated) broadening of a monodisperse peak by a technique as a measure for its resolution. It is expressed as a coefficient of variation (v techn,%). Thus, a higher value means a lower resolution. The value given can be related to the definition of resolution R 1 as used in chromatography. For example, R = 1.5 corresponds with a relative size difference of 6v techn and gives about 2% valley height in between two symmetrical Gaussian peaks of equal height. Alternatively, the term ``peak capacity is sometimes used. It relates to 1/v techn. It will be clear that resolution is also related to the width of the size classes applied in PSD determination: at least three classes are involved if there are two 1 In chromatography, resolution is defined as: R=2 (t R,1 tr,2 )/(w 1 + w 2 ), where t R is the retention time (peak position) and w is the base width, both of component 1 and 2, respectively. The absolute base width of a Gaussian peak can be approximated by 4 σ.

12 54 3 Quality Aspects in Particulate Analysis differential cumulative differential cumulative Fig. 3.4 Illustration of resolution for two normal distributions with modes at 1 and 1.3 (σ 0.1 left and 0.05 right); differential and cumulative curves peaks with one valley in between. If the amounts in the classes are in some way correlated, then more classes may be required. An illustration of resolution is given for a 50/50 mixture in Fig This figure also shows that differential/density curves give an easier identification of the presence of two peaks than cumulative curves. Related to the above, below two values are given for the quality of resolution. The first is the coefficient of variation that is related to the peak broadening of a technique. The second value gives the minimum relative distance of two monosized peaks of equal height to show a valley in between the peaks of at most 2 % of the peak height: v techn, % relative distance, % - Excellent <0.5 % <3 % - Good % 3 10 % - Medium 2 8 % % - Low 8 30 % % - Very low >30 % >200 % The resolution of the different measurement techniques is limited by different factors: Degree of Brownian motion or particle convection during sedimentation. Spread of sieve apertures in a sieving medium. Homogeneity of electrical field in ESZ aperture. Number of pixels per particle in image processing. Difference in signals from particles of different size. Width of chosen or given size classes. Degree of smoothing applied in calculation of PSD from detector signals. For most industrial products, high resolution is not very important, as their size distribution is usually wide. Still, adequate resolution is required for sufficient sensitivity (See below) with respect to the amount of material in specified size classes. For very narrow PSD s, for example standard reference materials, resolution is highly relevant. First, higher resolution allows easier discrimination between parti-

13 3.3 Quality Aspects in Particulate Analysis differential cumulative differential cumulative Fig. 3.5 Illustration of sensitivity in a PSD (90/10 mixture at relative distance 0.3 and 95/5 mixture at relative distance 0.5; σ 0.1); differential and cumulative curves cles of different size. Secondly, it allows unbiased determination of the width of the PSD. And, finally, it enables quantitative determination of small agglomerates. Sensitivity; Limit of Detection; Lower Determination Limit Sensitivity is defined as change of (instrument) response with change in absolute analyte concentration or specific for PSD analysis in the amount or relative concentration of material in a specified size class of a size distribution (also named ``critical component, if it is regarded critical to product performance). The general definition relating to absolute concentration is important when the total particulate concentration is to be measured. When this concentration is low, the limit of detection and the lower determination limit should be known (See below). Both aspects of sensitivity are related to the ability of a method or technique to distinguish meaningfully between closely related concentrations of material. Thus, sensitivity describes the ability to discriminate between products of acceptable and non-acceptable quality. In general, sensitivity depends on noise level, degree of smoothing applied in the technique, precision and the size range of the amount of material. The latter effect is illustrated in Fig. 3.5 for a 90/10 mixture of peaks, which are not completely separated, and a 95/5 mixture of well separated peaks. The first mixture shows a shoulder on the main peak in the differential curve and some tailing in the cumulative curve. The 10% admixture can only be deduced from the cumulative curve by comparing it with Fig. 3.3 (for example, by comparing the amounts above size 1.3). In the second mixture, the 5% are clearly visible in both differential and cumulative curve and can be best quantified in the cumulative curve. The quantification of sensitivity in PSD analysis will be given here as a special case of the lower determination limit. The limit of detection (Lc ) is defined as the limiting value, where the presence of ``critical component can just be detected with a stated confidence, but with low significance for its quantity or concentration. Typically, a 95% or 99.7% confidence level is taken.

14 56 3 Quality Aspects in Particulate Analysis Fig. 3.6 Illustration of detection limit and lower determination limit in relation to blank or background signal The lower determination limit (L d ) is defined as the limiting value designating the lowest absolute concentration or relative amount for some ``critical component in a PSD that can be estimated with a stated confidence. Usually, L d is taken as mean blank signal plus four or six times the standard deviation of the analytical method at low concentration (cf. Figs. 3.5 and 3.6). This is related to a 95% and 99.7% confidence level for correct decisions about the presence of the critical component, that is: a decision present if present (with 2.5% and 0.13% chances on error of type I) and a decision not present if not present (with 2.5% and 0.13% chances on error of type II; See Chap. 20). The difference between the limit of detection L c and the lower determination limit L d is illustrated in Fig The lower determination limit of a technique for particle concentration in liquid or air can be classified as follows: Liquid Air - Extremely low < 10-4 % (v/v) < 1 g/m 3 - Very low % (v/v) 1 10 g/m 3 - Low % (v/v) g/m 3 - Medium % (v/v) g/m 3 - High 0.1 1% (v/v) g/m 3 - Very high > 1% (v/v) >104 g/m 3 Note the large difference in qualification of determination limit about a factor of 10 6 between the values given in liquid and air. This is related to the normal scale of observation, which is in the centimeter range for liquids and in the meter range for air. Also the differences in refractive index between liquids and air play a role in both the detection and the lower determination limit. The lower determination limit of a technique for a small separate peak in PSD, can be classified as follows: - Excellent <0.5% (v/v) of PSD - Good 0.5 1% (v/v) of PSD - Medium 1 3% (v/v) of PSD - Poor 3 10% (v/v) of PSD - Very poor >10% (v/v) of PSD

15 3.3 Quality Aspects in Particulate Analysis 57 The range of particulate concentrations that can be measured with a stated confidence is usually called dynamic range. Traceability Traceability is defined as property of a measurement result whereby it can be related in a quantitative way to (inter)national standards through an unbroken chain of comparisons. Thus, it relates the measured result with the standard unit of length, mass or time, with stated uncertainties involved in all steps of comparison. The importance of the above quality aspects in relation to product performance depends on the type of product. Always, the measurement range of a method or technique should cover the particle or pore size or concentration that has significant influence on the product performance. Moreover, short analysis times are usually important, especially if the results are to be used for process control. In addition, for industrial particulate products, which have fairly wide size distributions, usually good precision and sensitivity are most relevant. For standard reference materials with a narrow size distribution also bias, resolution and traceability are important. For clear process streams (e.g., parenteral fluids) and clean air (e.g., production areas for printed circuits), the lower detection limit of absolute particulate concentration plays an essential role. This contrasts with most particulate process streams, in which the particulate concentrations are high. Thus, in-line analysis is only possible for techniques with measurement capability at such high concentrations. Most often, the interest lies in the relative concentration of material over the PSD or in a specific size class. If capability for high-concentration measurement is not available, then on-line or off-line analysis has to be executed on diluted samples. General advice on quantitative requirements for the quality aspects for a method or technique cannot be given, as the requirements depend strongly on the type of product and its application. Besides these quality aspects, the choice of a measurement technique is often also governed by investment and operational costs. Moreover, the choice of a specific instrument depends usually on the quality of the supplied information, the results of relevant tests and the estimated or known quality of maintenance support. Typical figures for all aspects of particle size measurement techniques will be given in Chaps In addition, the quality of measurement results depends on two more items, viz.: Operator Competence Proven competence of operators for specified tasks is an essential part of reliable and adequate quality of measurement results. It requires adequate education and training of operators as well as regular testing of their capability with standard samples.

16 58 3 Quality Aspects in Particulate Analysis Reporting Analysis results are only as good as they are reported. Therefore, any omission in a report limits the quality of reported results. Following items can be distinguished in reporting (See also Sect. 3.4 and Annex 3A): data, considerations and conclusions of preliminary experiments and tests for relating product quality and PSD in a quantitative way as well as for selecting appropriate instrumentation and method for analysis; literature used dates and results of regular qualification tests of instruments and operators and of instrument maintenance and repair in case of off-line analysis, general (qualitative) impressions of product appearance, e.g., sample identification; type of product (granules, powder, paste); particle shape; color and tinting; occurrence of segregation and lumps or clusters; dispersion quality operator identification and date(s) of analysis reference to relevant SOP s, or identification of all instrumentation and methods applied, including full details of instrument settings all detailed PSD analysis results summary of characteristic product and/or PSD parameters required for product characterization. 3.4 Design of a PSD Analysis Procedure A decision on the choice of a particle size measurement procedure in relation to a given application has to be based on an adequate consideration of all of the above aspects by the user. It should be clear that a sound strategy has to be followed for adequate measurement of particulate performance through PSD determination. Such a strategy takes careful consideration of each of the following steps: 1. translation of product performance into characteristic parameter(s) for PSD, shape and/or porosity, preferably in a quantitative mathematical model (Chaps. 1 3) 2. quantitative definition of required analytical quality aspects in view of required discrimination in product quality 3. definition of procedures and precautions in view of environment, hygiene and safety. 4. choice of total operating procedure (Chaps. 3 14) in view of required PSD parameters, required measurement frequency, quality aspects and possibilities for dry or wet analysis and for off-line, on-line and in-line measurement 5. choice of sampling equipment and measurement instrument in view of best offer based on both quality and price 6. set-up of operating procedure, meeting the required quality, including: agreement with eventual written standards (Chap. 15) powder sampling and sample splitting, leading to representative samples for PSD analysis, while taking eventual product segregation into account (Chap. 4)

17 3.4 Design of a PSD Analysis Procedure 59 sample dispersion and dilution, leading to stable particle dispersions (Chap. 5) choice of suitable reference material (Chap. 16) instrument calibration/qualification (Chaps. 6, 17) PSD analysis data representation 7. data interpretation and assessment of precision 8. operator education and training 9. set-up of procedures for method validation, instrument qualification and quality management (Chaps. 3, 17) 10. (establishment of measurement outputs and inputs for process control; Chap. 14) 11. writing down the details of the complete method in a SOP 12. writing down the details of all above-mentioned aspects in a Protocol for adequate quality management (Chap. 17). General considerations for both the Protocol and SOP in relation to adequate characterization of particulate products are given in Annex 3A. In Steps 5 and 6, the important decisions on the total operation procedure are made. Often, equipment and procedures are used, which usually stem from previous experience. When the precision of the total method is to be improved, it is advised to have a closer statistical look at sampling, dispersion and measurement separately. In order to meet this need, several samples are taken from the same batch, each dispersed and each analyzed several times according to the optimized procedure. The resulting data can then be analyzed by the statistical approach given in Section 20.3 and used as a basis for improvement. A well-written SOP forms the heart of a measurement strategy in which quality is controlled. It originates from adequate preliminary investigations, experience and awareness of all potential error sources and it describes all relevant details in a standardized way. Potential error sources are listed below: General errors Not following the standard procedure Use of improper or deteriorated reference material Sampling errors Insufficiently representative sample for product lot Insufficient sample size Contamination Losses during sampling or in instrument Change of chemical composition Change of physical composition (size, moisture) Unintentional human errors (e.g., mislabeling) Fraud or sabotage Dispersion errors Insufficient wetting of powder Incomplete de-agglomeration Attrition or breakage of primary particles Dissolution or swelling of particles Presence of air bubbles Instability of dispersion (creaming, re-agglomeration, settling)

18 60 3 Quality Aspects in Particulate Analysis Fig. 3.7 Error indications for sampling, dispersion and size measurement in relation to particle size (position of lines depends on e.g., product properties) Error, % sampling dispersion instrument Particle size, mm Measurement errors Reporting errors Poor instrument preparation (calibration, alignment) Incorrect measurement conditions Incorrect optical model or parameters Inadequate PSD model Particles outside measurement range Inadequate instrument calibration/qualification Noise or drift of instrument Incorrect data transcription Omission of operator identification Omission of sample identification Omission of measurement date. Figure 3.7 illustrates that random errors in particle size made during sampling or dispersion are often dominant over those made in the measurement. This is especially true for sampling of segregated, non-cohesive powders with broad PSD s and for not-well dispersed or stabilized suspensions and emulsions, which show agglomeration, flocculation, creaming, or sedimentation. Of course, these errors are strongly dependent on amount of material taken in consideration, material properties, shape of the particles and width of the size distribution. A consequence of the above is also that the precision of each step in the total procedure must be (well) below the required overall precision for characterization. In this book, a distinction is made between the words: operating procedure measurement method written document describing all details of one or more operational steps in the characterization of a particulate product, from primary sampling to reporting, leading to one or more measured parameter(s) with specified quality (either or not standardized) short description of the essential points of the measurement procedure for product characterization, usually including sample preparation/administration, measurement technique and operation conditions

19 3.5 Definitions, Abbreviations and Symbols 61 measurement technique (description of) principle and specific instrument configuration applied for measurement protocol written document describing the goals of analysis, the relevant parameters, the required quality, the total operating procedure and the way of reporting for adequate characterization of a defined product property. A protocol also includes or refers to documented results of the investigations that have lead to the choice of equipment and operating procedure. A good protocol gives the possibility to check whether the quality requirements set for the different steps of the characterization are in accordance with the requirements set for the performance of a product. It also allows adequate judgment of characterization costs, including purchase of instrument, frequency of analysis and validation, and in-/on-line analysis facilities, in relation to the costs of existing or improved product. 3.5 Definitions, Abbreviations and Symbols Accuracy Analysis time Bias Critical component Drift Dynamic range Limit of detection Lower determination limit closeness of agreement between a test result and an accepted reference value (including both random and systematic errors) total time required for analysis of a sample, including sample splitting, dispersion/dilution, measurement, and reporting difference between the mean value, coming from a set of test results, and an accepted reference value (systematic error) amount or concentration of material in a specified size class of a PSD that is of special interest for product performance gradual change of detector or instrument response with time (time constant large in comparison to measurement time, typically > 1 min) range of particulate concentrations that can be measured with a stated confidence limiting value that designates the lowest concentration or quantity for some critical component that can be discriminated from noise or background with a stated confidence, but with low significance for its concentration or quantity limiting value that designates the lowest concentration or quantity for some critical component that can be determined with a stated confidence

20 62 3 Quality Aspects in Particulate Analysis Measurement method short description of the measurement procedure, including sample preparation/administration and measurement conditions and technique Measurement range range of particle sizes that can be measured by a technique with stated confidence Measurement technique principle and configuration of a measurement instrument Measurement time time required for the mere measurement of a sample in order to reach a meaningful result Noise spurious, random change of detector or instrument signal around a mean value with short time constant relative to measurement (typically <0.1 s) Operating procedure written document describing all details of one or more operational steps (primary sampling, sample splitting, dispersion, instrument calibration/qualification, measurement, method validation, and reporting) for characterization of a particulate product property, leading to one or more specified parameter(s) with specified quality, either or not standardized Particle discrete piece of material Particulate material material consisting of particles Powder Precision Protocol Qualification Repeatability (instrument) Repeatability (method) Reproducibility mixture of dry, solid particles closeness of agreement between multiple measurement results of a given property by the same method (due to random errors); See also repeatability and reproducibility written document describing the goals of analysis, the relevant parameters, the required quality, the total operating procedure and the way of reporting proof with reference material that an instrument is operating in agreement with its specifications closeness of agreement between multiple measurement results of a given property in the same dispersed sample aliquot, executed by the same operator in the same instrument under identical conditions within a short period of time closeness of agreement between multiple measurement results of a given property in different aliquots of a sample, executed by the same operator in the same instrument under identical conditions within a short period of time closeness of agreement between multiple measurement results of a given property in different aliquots of a sample, prepared and executed by different oper-

21 3.5 Definitions, Abbreviations and Symbols 63 Resolution Sensitivity Specification Traceability Validation CNC DLS DMA D 10 D 50 D 90 ESZ FFF GPC HDC IR L c L d PCA PCS PM10 PSD SEC SEM SOP s s g ators in similar instruments in different surroundings by the same method ability of a measuring device to distinguish meaningfully between closely adjacent values of size change of (instrument) response with change in absolute analyte concentration or specific for PSD analysis in the amount or relative concentration of material in a specified size class of a size distribution precise statement of a set of requirements to be satisfied by a material, product, system or service, together with procedures for determination whether each of the requirements is fulfilled property of a measurement result whereby it can be related in a quantitative way to appropriate (inter)national standards through an unbroken chain of comparisons (with stated uncertainties in all steps of comparison) proof with reference material that a procedure is acceptable for all elements of its scope condensation nucleus counter dynamic light scattering differential mobility analyzer size of 10% point of a cumulative undersize PSD median particle size; size of 50% point of a cumulative undersize PSD size of 90% point of a cumulative undersize PSD electrical sensing zone field flow fractionation gel permeation chromatography hydrodynamic chromatography infrared radiation limit of detection lower determination limit principal component analysis photon correlation spectroscopy mass of particles per unit volume, smaller than or equal to an aerodynamic diameter of 10 m particle size distribution size exclusion chromatography scanning electron microscopy standard operating procedure estimated standard deviation (absolute value for precision or PSD width) from a series of measurements geometric standard deviation (characteristic width parameter for log-normal PSD)

22 64 3 Quality Aspects in Particulate Analysis TEM UV VIS v transmission electron microscopy ultraviolet radiation visible radiation coefficient of variation (relative value for precision, standard deviation divided by related mean value, expressed as fraction or percentage) true value for standard deviation (absolute value) Bibliography 1. M. Alderliesten, Part. Part. Syst. Charact. 19 (2002) ; 21 (2004) and 22 (2005) K. Gotoh, H. Masuda, K. Higashitani, Powder Technology Handbook; 1997 CRC. 3. W.C. Hinds, Aerosol Technology; 1999 Wiley & Sons. 4. P.J. Rousseeuw, A.M. Levy, Robust Regression and Outlier Detection; 1987 Wiley & Sons. 5. K. Willeke and P.A. Baron, Aerosol Measurement; 1993 Van Nostrand Reinhold.