- Email: [email protected]
Particle size analysis on wide size distribution powders; effect of sampling and characterization technique
Advanced Powder Technology 26 (2015) 200–207
Contents lists available at ScienceDirect
Advanced Powder Technology journal homepage: www.elsevier.com/locate/apt
Original Research Paper
Particle size analysis on wide size distribution powders; effect of sampling and characterization technique G. Bahar Basim a,⇑, Mohsen Khalili b a b
Department of Mechanical Engineering, Ozyegin University, Cekmekoy, Istanbul, Turkey DuPont Central Research and Development, Wilmington, DE, USA
a r t i c l e
i n f o
Article history: Received 9 June 2014 Received in revised form 5 August 2014 Accepted 22 September 2014 Available online 6 October 2014 Keywords: Particle size distribution Particle sizing techniques Wide particle size distribution powders Sampling
a b s t r a c t Particle size distribution of powders plays a very important role in determining the critical chemical and physical properties of the particulate systems. Precise determination of particle size distribution depends on effective sampling of the powders, which is more pronounced for the particulate systems with a wide particle size distribution. Predominantly, the significant increase in the total surface area of the powders at nanometer scale particle sizes may lead to improper characterization of the bulk if the sampling technique fails to collect and represent them in the analyses. In this study, effects of sampling on the precision of particle size analysis are studied on a clay sample with a wide particle size distribution (particles ranging from nanometer to micrometer sizes) using light scattering technique in aqueous media. Three different sampling methods are applied to systematically analyze the effect of sampling on particle size measurements including; riffling the original sample into sixteen equal parts, sampling the powder after removing the very fine and very coarse size particles and riffling to sixteen parts and finally by riffling the powder to the exact feed amount of the particle size analyzer. The effectiveness of the applied methods were compared statistically by calculating the coefficient of variance (CV) values of the 10%, 50% and 90% passing particle size data of the sequential runs. The most effective sampling method was determined to be riffling the sample to the exact feed amount of the analyzer based on obtaining the minimum CV values of the measurements. In the second part of the study, results of size distribution analyses conducted by different particle size analyzers utilizing numerous characterization techniques are compared using the most effective sampling technique developed in the first part. It is observed that the use of different characterization equipment tend to result in variations in the particle size distributions of the same powder which presents another variability in classification of the wide particle size distribution powders. Ó 2014 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.
1. Introduction Particle size distribution of particulate systems is a critical variable for a number of industrial operations ranging from mining to pharmaceuticals. Precise measurement of particle size distribution of powders is important since the quality and performance of most powder-based products are closely related to the size distribution and particularly the concentration of the fine particles. Although the fine size fraction generally compromise of a small volume, the presence of fine particles leads to immense increase in the specific surface areas, creating a state of high surface energy that dominates properties of the selected system [1]. A good example is increasing moisture content in coal due to fine size content ⇑ Corresponding author. Tel.: +90 216 5649344; fax: +90 2165649057. E-mail address: [email protected] (G.B. Basim).
( 100 mesh) particles retaining more water molecules due to their large surface area that results in elevated filtration, drying and transportation costs [2]. Another critical example applies to microelectronics manufacturing, where the presence of slightly larger size particles at parts per million (ppm) levels in slurry made of nanometer size particles may result in significant surface deformation during the planarization of the wafers by the chemical mechanical planarization (CMP) process [3,4]. For pharmaceutical applications, where more than three-quarters of all tablets and capsules are manufactured using powder blends are more prone to sampling problems since the inaccuracy of sampling may directly affect the human health [5]. Food products, such as skim milk powder, are also hard to sample due to agglomeration during processing which skews the particle size distributions to larger ends and hence may not be representative [6]. Consequently, analysis of a representative powder sample is very necessary to obtain
http://dx.doi.org/10.1016/j.apt.2014.09.009 0921-8831/Ó 2014 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
the correct size distribution of a particulate system in many industrial applications, which in turn helps in enhancing the process efficiency as well as the product performance. The main objective of this study is to evaluate the effect of powder sampling on precision of the particle size distribution analyses and to delineate potential approaches for improving the sampling related variability in size distribution analysis of wide particle size distribution powders. The precision of particle size analysis strongly depends on the effectiveness of the selected sampling technique. In many industrial applications, as discussed above, sampling starts from a large amount of material, yet the collected sample powder must accurately represent this larger entity. Generally, gathering a representative sample with a very small volume is the primary challenge to particle sizing operations. It was reported that the quick and careless sampling of powder often results in an expensive study of an irrelevant or non-representative group of particles [7]. Furthermore, it was also reported on a pharmaceutical application that not only the method of sampling but also the selection of proper sampling device is critical in sampling. Use of different sizes of the thieves for sampling the same drug powder was observed to make significant differences in the strengths of the pills produced [8]. There are new technologies developed for the sampling of powders for delicate applications [9,10], methods that utilize digital imaging [11], in addition to methodologies for sampling large amount of materials (as in mining industry) [12] and many conventional instruments available for decades [13]. One of the most important factors in sampling powders for particle size analysis is whether the characterization will be conducted offline or online [14]. For most of the industrial applications, particle size analyses are conducted offline in the laboratories. After the powder sampled from the production line reaches the laboratory for particle size analysis, further sampling is required to decrease the amount of material to the suitable quantity for the selected particle size analyzer. In the case of online sampling of powders (which is conducted during processing), on the other hand, it is also important to check whether the collection of particles is required, or the information can be obtained by indirect methods. The amount of sample that must be collected should not be disturbing the production process while the cost of the sampling method in money and time, calibration and operational characteristics of the sampling system are additional critical factors to be considered [14]. Optical particle sizers provide robust characterization ability for online monitoring through their ability of covering a wide size range by changing the technique that they utilize through changing the optical configuration [15]. There are standards developed for sampling of powders for particle size analysis, which further verifies the importance of sampling to conduct accurate particle sizing analyses [16]. Determination of the most suitable sampling technique is driven by a number of different considerations including the total mass, number concentration, size distribution and chemical composition of the particles to be characterized in addition to the size distribution technique to be used [17–19]. When sampling powders, there are two types of sampling errors possible that can affect the particle size analysis [20]; (i) errors due to segregation of the bulk powder which can be prevented by suitable mixing and (ii) statistical errors that are due to random fluctuations and cannot be prevented. Both errors are more noticeable when the particle size distribution of the studied powder is wide and variable (changes due to agglomeration or due to breakage of the large size particles by attrition). While the first type of sampling errors can be minimized by suitable mixing and collecting a representative sample, statistical errors cannot be prevented since the quantitative distribution in samples of a given magnitude is not constant even for an ideal random mixture [20]. Yet, increasing the sample size helps estimate the statistical errors and reduce them. Hence it
201
is important to statistically evaluate the results of particle sizing to understand the effect of sampling on precision of the characterization. In this study, a clay sample with a significant variability in particle size was selected as a representative wide size range particulate system for offline characterization at the laboratory. Clays are well characterized for their colloidal properties in water based systems and are used in many industrial applications [21]. A detailed statistical approach was followed to evaluate the variations that sampling may create in particle size analysis. Three different sampling methodologies were implemented by riffling the main powder (i) to sixteen parts and preparing suspensions of these samples in water (ii) using the same procedure after the very fine and very coarse fraction of the powder is removed by screening to tighten the particle size distribution and, (iii) riffling the powder to the exact feed amount of the particle characterization equipment and preparing aliquots of the exact feed amount. The sampled powders were analyzed for particle size distributions at 10%, 50% and 90% passing values to determine the best sampling technique, which produces the most repeatable results in particle sizing. The particle size analyses were conducted using conventional sieve analysis, light scattering and number counting techniques. Initially, Coulter LS 230 (wet) and Microtrac Full Range Analyzer (FRA) instruments were employed to evaluate the best sampling technique. Next the most accurate sampling technique was used to compare the variability in the results of different particle size analyzers including, Coulter LS 230 (both dry and wet), Microtrac FRA (wet), Coulter Multisizer and Sympatec Helos (both dry and wet). 2. Materials and methods 2.1. Materials A broad size distribution bentonite clay sample was selected to study the effect of sampling on the precision of the particle size analysis. The main components of the clay sample were Al3Mg2, CaAl2Si2, AlMg, Al3Mg2, Al2Mg, CaAl2Si1.5, Al0.58 MgO0.42, Si, K and Mg2Si according to X-ray analysis which was consistent with calcium bentonite composition. Fig. 1 illustrates a scanning electron microscope (SEM) micrograph of the clay powder at 200 magnification on which the broad size distribution of the sample was detected clearly. The sieve analyses were also conducted on the powder, confirming the broadness of the size distribution of selected clay as demonstrated in Fig. 2. The fifty percent passing particle size (d50) was calculated as 137.6 lm based on sieve sizing.
Fig. 1. SEM micrograph of the clay powder (Magnification = 200).
202
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
Fig. 2. Sieve analysis of the clay sample.
2.2. Sampling methods Fig. 3 schematically illustrates the sample preparation procedure. Initially, the clay powder was riffled to 16 tubes and each tube’s content was stored in a separate bottle. Three bottles (# 1, 6 and 11) were selected randomly for the first part of the sampling analysis. By mixing the dry powders in the bottles with a spatula, three 0.2 g samples were collected from bottle # 1 and one sample was collected from each of the bottles # 6 and 11. The 0.2-g powders were mixed with 19.8-ml water (a total weight of 20 g, 1-wt%) in 50-ml beakers, shaken vigorously and conditioned for minimum 1-h before the sizing tests were conducted. Another vigorous conditioning was applied before the solutions were fed into the particle sizing equipment to prevent settling of the particles. These sample solutions were named as preparations (P). Three aliquots (A) were collected with a dropper from the first preparation of bottle #1 with an order of A1 was collected from the bottom, A2 was collected from the middle and A3 was collected from the top part of the bottle. Only one aliquot was taken from each of the other preparations by dipping the dropper to the bottom of the suspension in the bottle. These samples were analyzed by Coulter LS 230 (3 runs per aliquot) and Microtrac FRA (1 run per aliquot) light scattering instruments. Coulter LS 230 runs were conducted at
100 rpm stirring speed at 0.4–2000 lm detection range for 90 s, and the Microtrac FRA runs were conducted for 30 s time range. For the second sampling method, the very fine ( 38 lm) and very coarse (+212 lm) fractions of the clay sample were removed by screening from the clay powder using the content of bottle #2. The aim was to evaluate the effect of tightening the particle size distribution on sampling efficiency by removing the end tails of the particle size distribution and determining the repeatability of particle size analyses. Three preparations were made from the screened powder using the same technique (0.2 g powder + 19.8 g water) as described previously. Three aliquots were taken from preparation 1 for the size analysis and only one aliquot was taken from preparations 2 and 3 following the same procedure defined in the first part. For the third sampling method, the exact amount of powder required to conduct a run of Coulter LS 230 and Microtrac FRA were determined (0.04-g). This sampling methodology helps improve the reproducibility of the particle size analyses since the aliquots are prepared from the exact amount of dry powder that is needed to operate the analyzer rather than collecting the sample from an aliquot that has been prepared at a higher volume and hence only some of the sampled dry powder can be fed to the analyzer. The dry sample content of Bottle #8 was riffled to the predetermined exact feed amounts of these analyzers and the particle size analyses were conducted. In addition, to evaluate the effectiveness of the riffling technique, bottles # 4, 5 and 6 were sent to Beckman Coulter Corporation. Using the exact feed amount of Coulter LS 230, they made three preparations from Bottle 4 and one preparation from each of bottle # 5 and 6. Their analyses were found to be in agreement with the conclusions of our study; they reported that riffling of the exact feed amount method gives a very good precision in particle size analysis. To compare the effectiveness of the applied sampling methods in repeatability of the particle size analyses, Coulter LS 230 and Microtrac FRA light scattering instruments were employed. On Coulter LS 230 all aliquots were analyzed three times while recirculating within the equipment to compare variations between runs of the identical sample feed as it keeps on circulating in the wet module of the instrument. The Microtrac FRA runs were conducted only once as the sample cannot be recirculated in the FRA system. The sizes of 10%, 50% and 90% passing fractions of each run were noted for each run. The mean particle size, standard deviation
Fig. 3. Schematic representation of sample preparation procedure.
203
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
(SD) and coefficient of variance values (CV) were calculated to determine the variations in different aliquots, preparations and bottles. CV is the ratio of the standard deviation to the mean value of the distribution multiplied by hundred and it shows the extent of variability in relation to mean of the population. 2.3. Comparison of particle size measurements of different analyzers In the second part of the study the variations in the particle size analysis of different instruments were analyzed using the same clay powder sampled by the riffling method. Three analyzers utilizing light scattering techniques Coulter LS 230, Microtrac FRA and Sympatec Helos were used in addition to one number counting technique (Coulter Multisizer). 3. Results and discussion 3.1. Determination of the best sampling methodology for enhanced repeatability of particle size analyses 3.1.1. Particle size analysis on original clay powder by wet sampling Initially, the effect of sampling on particle size analysis was studied on the original clay powder by preparing 1 wt% suspensions from the selected bottles #1, 6 and 11. The preliminary analyses were performed on the size data obtained from the three aliquots of the same preparation (bottle # 1, preparation 1) by collecting the aliquots from the bottom, middle and top level of the suspension in the bottle named as A1, A2 and A3 consecutively. The aim of this part of the study was to find out whether the level the suspension was sampled from the bottle would affect the particle size measurements. Table 1 summarizes the statistical analysis conducted on the data obtained from the aliquots of the first preparation of bottle #1 by Coulter LS 230 including the mean particle size, standard deviation and the coefficient of variances for the three aliquots as well as the three consecutive runs conducted on the same aliquot feed. It can be seen that both the SD and CV values between the aliquots collected from the same preparation were high especially at the fine fraction reported as 10% passing. Particularly the variation between the 10% passing size of the subsequent runs of aliquot #1, which was collected from the bottom of the bottle, had a very high CV value of 54.8. This variation can be attributed to be due to various factors. First of all, it is possible
to have a larger fine size fraction at the very bottom of the bottle due to hindered settling. Secondly, the fine fraction of the particles might have agglomerated and settled to the bottom of the bottle although the sample was well shaken during the suspension preparation and before it was collected from the bottle. Lastly, larger particles, which settle faster, maybe going through attrition generating an additional fine fraction at the bottom of the suspension. It also appears that as the same feed is run in the equipment again and again, 10% passing particle size tend to decrease as can be seen for aliquot #1 for runs 1, 2 and 3 measured as 17.7, 9.1, 6.1 lm average size, respectively. All three factors discussed above, hindered settling, agglomeration or attrition maybe leading to this observation. Furthermore, it is suspected that the sequential runs conducted by keeping the particles circulating in the instruments may result in breakage of the larger size agglomerates by enhanced stabilization of the particles. This argument can further be supported by the fact that the particle suspension fed into the analyzer is further diluted in the volume of water that is in the equipment’s measurement cell, which is 200 ml for Coulter LS 230 decreasing the suspension to a 0.1 wt% concentration effectively. Aliquot #2 and #3, which were collected from the middle and the top of the bottle, had more consistent results as can be seen from the SD and CV values of Table 1. Nonetheless, the true variability of the particle size measurements at the fine fraction is only represented when the sample is collected from the bottom of the bottle since the effective particle size distribution may include both flocculated particles and aggregates as well as discrete particles [22]. This variance is reflected when the aliquots 1, 2 and 3 are compared for run 1 at 10% passing sizes reporting a CV value of 76 indicating that the position where the sample is collected within the same suspension significantly affects the size analyses for the fine size evaluations. In order to represent the potential of the fine size fraction of the
Table 2 Original clay sample statistical analysis for Microtrac FRA runs. Bottle #1, preparation 1, aliquots 1, 2, 3. Microtrac FRA
Aliquots
% Passing
1
2
3
(lm)
92.3 200.2 360.5
74.3 182.5 364.5
67.7 180.2 335.0
78.1 187.6 353.3
Run 1
10% 50% 90%
Mean
SD
CV
12.7 10.9 16.0
16.3 5.8 4.5
Table 1 Original clay sample statistical analysis for Coulter LS 230 runs. Bottle #1, preparation 1, aliquots 1, 2, 3. Coulter LS 230
Aliquots
% Passing
1
2
3
(lm)
Mean
SD
CV
Run 1
10% 50% 90%
17.7 140.2 283.7
6.4 119.1 263.2
4.3 94.1 229.7
9.5 117.8 258.9
7.2 23.1 27.3
76.0 19.6 10.5
Run 2
10% 50% 90%
9.1 139.3 281.6
5.1 118.0 259.8
4.0 91.6 227.5
6.1 116.3 256.3
2.7 23.9 27.2
44.1 20.5 10.6
Run 3
10% 50% 90%
6.1 138.9 282.1
4.6 117.1 257.6
3.8 91.0 225.3
4.8 115.7 255.0
1.2 24.0 28.5
24.2 20.7 11.2
Mean
10% 50% 90%
10.9 139.4 282.5
5.4 118.1 260.2
4.0 92.3 227.5
SD
10% 50% 90%
6.0 0.7 1.1
0.9 1.0 2.8
0.3 1.6 2.2
CV
10% 50% 90%
54.8 0.5 0.4
16.4 0.8 1.1
6.8 1.8 1.0
204
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
Table 3 Original clay sample statistical analysis for Coulter LS 230 runs. Bottle #1, preparations 1, 2 and 3 (aliquots 1). Coulter LS 230
Preparations
Mean
SD
CV
1
2
3
(lm)
Run 1
10% 50% 90%
17.7 140.2 283.7
13.3 164.6 304.2
19.6 167.5 304.8
16.8 157.4 297.6
3.2 15.0 12.0
19.1 9.5 4.0
Run 2
10% 50% 90%
9.1 139.3 281.6
9.3 165.3 303.4
11.3 168.1 301.6
9.9 157.6 295.5
1.2 15.9 12.1
12.4 10.1 4.1
Run 3
10% 50% 90%
6.1 138.9 282.1
7.3 164.4 300.0
8.8 169.9 300.5
7.4 157.7 294.2
1.3 16.5 10.5
17.9 10.5 3.6
% Passing
Table 4 Original clay sample statistical analysis for Microtrac FRA runs. Bottle #1, preparations 1, 2 and 3 (aliquots 1). Microtrac FRA
Preparations
% Passing
1
2
3
(lm)
92.3 200.2 360.5
81.3 187.2 361.4
72.0 176.7 346.5
81.9 188.0 356.1
Run 1
10% 50% 90%
Mean
SD
CV
10.1 11.8 8.4
12.4 6.3 2.3
powders to agglomerate or added into the system by attrition of larger ones, the remaining of the wet sampling analyses were conducted by collecting the aliquots from the bottom of the sample bottles after the bottle content was shaken vigorously. Table 2 shows the results of the same analysis conducted on the Microtrac FRA runs, for which the sample was discharged immediately after the run and therefore could not be rerun in the system. It can be seen that the cut sizes obtained with Microtrac FRA were higher compared to Coulter LS 230 results. However, the maximum variation was still observed on the fine fraction. These results once again confirm that regardless of the selected sizing equipment, samples collected from the same preparation but from different positions in the container might show a significant difference depending on how they are sampled. Tables 3 and 4 show the statistical analyses on the tree 1 wt% solution preparations prepared from the dry powder of bottle #1 with Coulter LS 230 and Microtrac FRA, respectively. The variability in this part comes from sampling the 0.2 g of powder from the same bottle. These results also show a high variance on the fine fraction measurements yet there is a relative decrease in the overall CV values as compared to the previous results reported in Tables 1 and 2. The main difference is that, in Table 1 aliquots of the same preparation were compared for particle size with respect to the position they were collected from the bottle. In this part of the sampling comparison, the aliquots were all collected from the bottom of the bottles for three different preparations of the dry powder from the same bottle. The reduced variability in CV values in this set of testing highlight that the collection of an aliquot from a suspension can change the particle size distribution of the sample more than how the sample suspension was prepared from a given dry powder. The observation on the decrease in the measured particle size for the sequential runs remained the same in this test as well once again supporting the discussion of enhanced stability of powder as it remained for a longer time in the sizing equipment. Nevertheless, since most of the sizing systems give a sufficient conditioning time to the particles before the characterization can be started and additionally due to the fact that not all the sizing equipment has the ability of rerunning the samples, it is suggested that the first set of measurements are taken into consideration or a consistent approach is followed by
the operator in sampling the suspensions and running them in the analyzers. Finally, the particle size analyses obtained from the three different bottle samples were compared as illustrated in Tables 5 and 6. The results of the first aliquots of the first preparations of each bottle were analyzed for each bottle. The CV values were higher as compared to the sampling from the same bottle for three times yet more consistent as compared to the aliquots collected from the same bottle but different positions. Also within the two analyzers, Coulter LS 230 was observed to be more sensitive in detecting the fine size fraction as compared to Microtrac FRA. This sampling method makes use of a very small amount of the bottle content that may account for the high variation between size data obtained from different bottles, even though the contents of the bottles may be similar. 3.1.2. Sampling analysis on the screened clay powder The results obtained on the original clay sample prepared as wet suspensions showed high variability as reported by SD and CV values in particle size analysis. The main source of the poor repeatability was suspected to be due to the wide particle size distribution of the clay powder. In order to understand the effect of controlling the width of the size distribution and decrease the variation in measurements, the very coarse and very fine fractions were screened out of the original powder forming a particulate sample contained within 38–212 lm size range. Since it was observed that Coulter LS 230 was more sensitive in detecting the variation, this set of testing was only conducted using this equipment. Three 20-g suspensions of 1 wt% were prepared using the screened powder and three aliquots were taken from the first preparation again from the bottom, middle and the top of the suspension. In addition, tree suspensions prepared from three different bottles were also compared. Table 7 summarizes the results of the statistical analyses conducted on the screened powders by Coulter LS 230. It can be seen that the sampling variation decreased significantly both among the three aliquots collected from the different positions of the same preparation (SD 0.6–0.8 and CV 0.7–1 for 10% passing) and for the aliquots sampled from the bottom of the tree different preparations (SD 5.3–6.0 and CV 6.9–7.7 for 10% passing). These results further confirm the discussion of wide size distribution creating a significant variability in sampling. Therefore, as expected, the system particle size distribution approaches to monosize, precision of the particle size analysis tends to improve significantly [3]. 3.1.3. Sampling analysis on clay powder riffled to exact feed amount Although separation of the very fine and coarse fraction of particles from the original clay powder helped improve the repeatability of the particle size analysis through facilitating better quality of sampling, the size distribution of the original clay powder was
205
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207 Table 5 Original clay sample statistical analysis for Coulter LS 230 runs. Bottles 1, 6 and 11 (preparations 1, aliquots 1). Coulter LS 230
Bottles
Mean
SD
CV
1
6
11
(lm)
Run 1
10% 50% 90%
17.7 140.2 283.7
9.4 160.9 297.2
27.9 170.4 308.9
18.3 157.2 296.6
9.3 15.4 12.6
50.5 9.8 4.3
Run 2
10% 50% 90%
9.1 139.3 281.6
7.5 162.3 297.0
17.7 168.5 305.7
11.4 156.7 294.8
5.5 15.4 12.2
47.8 9.8 4.1
Run 3
10% 50% 90%
6.1 138.9 282.1
6.3 162.8 295.5
13.3 167.1 302.9
8.6 156.3 293.5
4.1 15.2 10.5
47.6 9.7 3.6
% Passing
Table 6 Original clay sample statistical analysis for Microtrac FRA runs. Bottles 1, 6 and 11 (preparations 1, aliquots 1). Microtrac FRA
Bottles
% Passing
1
6
11
(lm)
92.3 200.2 360.5
76.8 186.6 348.2
82.7 190.9 334.1
83.9 192.6 347.6
Run 1
10% 50% 90%
Mean
SD
CV
7.8 7.0 13.2
9.3 3.6 3.8
changed knowingly. In order to analyze the size distribution of the original powder with a wide particle size distribution a third technique was developed. In the first sampling method where a wet suspension was prepared from the original powder, a high variation was observed between the aliquots of the same preparation. This lead us to develop a new technique in which the powder is riffled to the exact feed amount of a single run for the particle size analyses to be conducted. In order to determine the exact amount of dry powder required to conduct a single measurement, an aliquot with a known volume was fed to the tool until the tool’s sensor indicated that the number of particles were sufficient to make the measurement. The volume fed to the tool was calculated by measuring and subtracting the remaining sample volume from the initially prepared sample volume. The exact feed amount was determined to be 0.4-g for Coulter LS 230 through drying the exact volume of one feed to determine the solids content. The calculation was cross verified through preparing an exact feed with the calculated solids loading in one aliquot and conducting the measurements again. Once this amount was determined, the original dry clay powder was riffled to this exact amount and sixteen suspensions were prepared using these riffled powders. The contents of eight even numbered riffle tubes were used for the Coulter LS 230 analysis and the other eight were run through the Microtrac FRA. To conduct a run, all of the suspension in a tube was fed into the analyzer. Table 8a and b
shows the comparison between the first sampling method and the third sampling method on Coulter LS 230 and Microtrac FRA, respectively. The statistical analyses on the preparations of the same bottles of the wet sampling technique were compared to the analysis on the samples riffled to exact feed amount of the analyzer. The calculated SD and CV values decreased for all the 10%, 50% and 90% passing fractions by about 2–3 times when the size analysis performed by the exact feed amount of powder. It is obvious from the results given in Table 8 that the riffling technique helped improve the precision of the particle size analysis regardless of the particle sizing equipment used for characterization. A more systematic study was conducted on the effect of riffling as a sampling method using Coulter LS 230 at Beckman-Coulter Inc. The exact amount of powder required for a run was riffled out of three bottles numbered 4, 5 and 6. Three preparations were made from bottle #4 and single preparations were made from the other two bottles. Table 9 shows the results of the statistical analyses conducted on the size analysis of the three bottles and the three preparations of the bottle #4. The results indicate a significant improvement in the repeatability and precision of the size analysis. According to these results we can conclude that the riffling the sample to the exact feed amount is the most effective sampling technique that can be used to represent the size distribution of the wide particle size powders. 3.2. Comparison of different particle analyzers From the first part of the analysis detailed in the previous sections, riffling the powder to the exact feed amount needed for the particle size analyzer was determined to be the most effective way of sampling for a wide particle size distribution sample. Hence, in this part of the study, where the measurements of various particle sizing equipment are compared and their consistency in sizing was investigated, the exact feed amount of the powder was riffled and prepared for every equipment used.
Table 7 Screened clay sample statistical analysis for Coulter LS 230 runs. Comparison of aliquots and preparations. Coulter LS 230
Aliquots
% Passing
Mean (lm)
SD
CV
Mean (lm)
Preparations SD
CV
Run 1
10% 50% 90%
76.7 146.9 234.7
0.8 0.3 2.0
1.0 0.2 0.8
76.5 147.7 236.1
5.3 5.2 4.2
6.9 3.5 1.8
Run 2
10% 50% 90%
77.4 148.1 234.9
0.6 0.6 1.9
0.7 0.4 0.8
77.3 148.7 236.1
5.7 5.6 4.1
7.4 3.7 1.7
Run 3
10% 50% 90%
77.8 148.6 234.5
0.7 0.8 1.7
0.9 0.5 0.7
77.8 149.2 236.1
6.0 5.8 4.7
7.7 3.9 2.0
206
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
Table 8 Comparison of the statistical analysis for original wet suspended and riffled clay samples (a) with Coulter LS 230 and (b) Microtrac FRA. Wet Suspended Mean
Riffled to exact feed amount SD
CV
Mean
Coulter LS 230 % Passing Run 1 10% 16.8 50% 157.4 90% 297.6
SD
CV
3.2 15.0 12.0
19.1 9.5 4.0
13.8 143.6 287.1
1.4 4.5 3.9
10.1 3.1 1.4
Microtrac FRA % Passing Run 1 10% 81.9 50% 188.0 90% 356.1
10.1 11.8 8.4
12.4 6.3 2.3
55.9 163.7 313.9
3.8 2.4 17.8
6.7 1.5 5.7
Table 9 Riffled clay sample statistical analysis for Coulter LS 230 runs. Comparison of the samples riffled from three different bottles and three preparations. Coulter LS 230
Preparations
% Passing
Mean
SD
CV
Mean
Bottles SD
CV
Run 1
10% 50% 90%
14.4 142.0 303.3
0.4 4.4 4.1
2.8 3.1 1.3
14.5 141.7 302.3
0.2 2.5 1.5
1.1 1.8 0.5
Run 2
10% 50% 90%
14.2 141.7 303.0
0.4 4.9 3.6
2.5 3.5 1.2
14.3 141.7 301.7
0.1 2.5 2.1
0.4 1.8 0.7
Run 3
10% 50% 90%
14.0 141.0 301.7
0.3 5.2 4.1
2.5 3.7 1.4
14.2 141.7 301.7
0.1 2.5 1.2
0.4 1.8 0.4
There are many different measurement principles including mechanical separation (sieves), light diffraction, sedimentation, photon correlation spectroscopy, dynamic light scattering, field flow fractionation, electrophoretic mobility and induced acoustic response for particle characterization. Among these, light scattering is the method of choice as it has a robust theory (diffraction) to calculate the particle size in addition to high speed, good reliability and high reproducibility [22]. Hence, we studied three different light scattering instruments and one number counting technique in this part of the analysis. For the light scattering tests, Coulter LS 230 (wet and dry), Microtrac FRA (only wet) and Sympatec Helos (wet and dry) were utilized. In addition, Coulter Multisizer analyzer was used for the number counting analysis. Fig. 4 shows the cumulative volume percent passing versus particle
size data obtained from the selected instruments and Table 10 summarizes the numerical values for the statistical analysis. It can be seen that the average (mean) particle size, 50% passing (median) and the most repeated particle size (mode) are not in a very good agreement for different instruments. Even the dry and wet analysis of the same instrument may give different particle size distributions. This tendency was also observed by other researches, and was attributed to the diversity of the measuring principles and instrument models [23–26]. Particularly, Hayakawa and coworkers studied spherical particles, granulated particles and particles with various shapes in a systematic way and also concluded that the method of sampling affects the particle size measurements significantly [24–26]. In this analysis, we have minimized the variation given by sampling yet still observed a significant variability among different particle sizing methods and even with equipment using the same methodology in particle sizing but made by various different manufacturers. It is important to be aware of the differences between the various particle sizing techniques and select the most effective one for the purposes of the conducted analysis and maintain a consistency when analyzing particulate systems.
Table 10 Comparison of the consistency of different particle size analyzers.
Fig. 4. Cumulative particle size distribution of clay sample obtained by using different analyzers.
Sieve analysis Microtrac FRA Coulter Multisizer Coulter LS (wet) Coulter LS (dry) Sympatec Helos (wet) Sympatec Helos (dry)
Mean size (lm)
Median (lm)
Mode (lm)
185.0 178.2 144.1 144.2 150.0 147.3 112.3
137.6 161.7 148.3 133.5 138.1 134.5 97.3
100.0 209.3 181.5 161.2 213.1 174.0 174.0
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
4. Summary The results of this study showed that the employed sampling technique has an important effect on the precision of particle size analysis. The most effective sampling technique requires the protection of the original particle size distribution of the particulate system during the analysis. Otherwise, the sample is not representative of the original powder and the size analysis may display a high variation. As the broadness of the size distribution of the powder increases, sampling gets more critical since it becomes more difficult to obtain a representative sample. According to our analysis, the most effective sampling technique was found to be the riffling of the powder to the exact amount required by the analyzer. It was also shown that the results of the size analysis of a particulate system may vary based on the selected technique and instrumentation. These issues also indicate the requirement of international standards on sampling for particle size characterization, which are currently being prepared and are believed to address most of the problems in particle science and technology.
Acknowledgments The authors would like to acknowledge the support from Mr. W. Rohricht at the Particle Science and Technology Center at DuPont Central Research & Development. The valuable contributions of Beckman Coulter Inc. are also acknowledged.
References [1] E. Matljevic, The world of fine particles (ultrafine particles produced by precipitation or chemical reactions with drops), ChemTech 21 (1991) 176–181. [2] G.B. Basim, Fine Coal Dewatering, MS Thesis, Virginia Polytechnic Institute and State University, Blacksburg, VA, USA, 1997. [3] G.B. Basim, U. Mahajan, J.J. Adler, B.M. Moudgil, R.K. Singh, Effect of particle size distribution of chemical mechanical polishing slurries for enhanced polishing with minimal defects, J. Electrochem. Soc. 147–9 (2000) 3523–3528. [4] G.B. Basim, B.M. Moudgil, Effect of soft agglomerates on CMP slurry performance, J. Colloid Interface Sci. 256–1 (2002) 137–142. [5] F.J. Muzzio, C.L. Goodridge, A. Alexander, P. Arratia, H. Yang, O. Sudah, G. Mergen, Sampling and characterization of pharmaceutical powders and granular blends, Int. J. Pharm. 250 (2003) 51–64.
207
[6] C. Turchiuli, R. Smail, E. Dumoulin, Fluidized bed agglomeration of skim milk powder: analysis of sampling for the follow-up of agglomerate growth, Powder Technol. 28 (2013) 161–168. [7] B.H. Kaye, R. Trottier, The many measures of fine particles, Chem. Eng. 102 (1995) 78–86. [8] J.A. Berman, A. Schoeneman, J.T. Shelton, Unit dose sampling: a tale of two thieves, Drug Dev. Ind. Pharm. 22 (11) (1996) 1121–1132. [9] F.J. Muzzio, P. Robinson, C. Wightman, D. Brone, Sampling practices in powder blending, Int. J. Pharm. 155 (1997) 153–178. [10] L. Susana, P. Canu, A.C. Santomaso, Development and characterization of a new thief sampling device for cohesive powders, Int. J. Pharm. 416 (2011) 260–267. [11] A. Boschetto, V. Giordano, Powder sampling and characterization by digital image analysis, Measurement 45 (2012) 1023–1038. [12] H.R. Cooper, New technology provides better crushed-ore sampling results, Eng. Min. J. Press 190–6 (1989) 54–59. [13] H.A. Behre, M.D. Hassialis, Sampling and testing, in: A.F. Taggart (Ed.), Mineral Dressing Handbook, John Wiley & Sohns, New York, 1945, pp. 1–208. [14] C.H. Murphy, Handbook of Particle Sampling and Analysis Methods, Verlag Chemie International, Deerfield Beach, Florida, 1984. [15] I. Goaninoni, E. Golinelli, G. Melzi, S. Musazzi, U. Perini, F. Trespidi, Optical particle sizers for online applications in industrial plants, Opt. Lasers Eng. 39 (2003) 141–154. [16] ISO TC 24/SC4/WG11, Sample Splitting for Particle Size Characterization, 1996. [17] G. Guidarelli, F. Craciun, C. Galassi, E. Roncari, Ultrasonic characterization of solid-liquid suspensions, Ultrasonics 36 (1998) 467–470. [18] J.F. Fabries, R. Wrobel, P. Gorner, D. Bemer, P. Bonnet, H. Nunge, M. Lafontaine, S. Binet, Comparison of particle-size distributions of bitumen fumes measured by aerosizer and QCM impactor techniques, J. Aerosol Sci. 31 (2000) 1011– 1012. [19] S. Jaffari, B. Forbesa, E. Collins, D.J. Barlowa, G.P. Martina, D. Murnanec, Rapid characterisation of the inherent dispersibility of respirable powders using dry dispersion laser diffraction, Int. J. Pharm. 447 (2013) 124–131. [20] T. Allen, Powder Sampling and Particle Size Determination, Elsevier Ltd., London, UK, 2003. pp. 1–55. [21] G. Lagaly, I. Dékány, Colloid clay science, in: F. Bergaya, G. Lagaly (Eds.), Handbook of Clay Science-Fundamentals, Developments in Clay Science, vol. 5, Oxford, UK, 2013, pp. 243–345. [22] Z. Ma, H.G. Merkus, J.G.A.E. de Smet, C. Heffels, B. Scarlett, New developments in particle characterization by laser diffraction: size and shape, Powder Technol. 111 (2000) 66–78. [23] J.M. Phillips, D.E. Walling, An assessment of the effect of sample collection, storage and suspension on the representativeness of measurements of the effective particle size distribution of fluval suspended sediment, Wat. Res. 29– 11 (1995) 2498–2508. [24] O. Hayakawa, K. Nakahira, J. Tsubaki, Comparison of particle size analysers and evaluation of its measuring technique with fine ceramic powders (Part 1), J. Ceram. Soc. Jpn. 103 (1995) 392–397. [25] O. Hayakawa, K. Nakahira, J. Tsubaki, Comparison of particle size analysers and evaluation of its measuring technique with fine ceramic powders (Part 2), J. Ceram. Soc. Jpn. 103 (1995) 500–5005. [26] M. Naito, O. Hayakawa, K. Nakahira, H. Mori, J. Tsubaki, Effect of particle shape on the particle size distribution measured with commercial equipment, Powder Technol. 100 (1998) 52–60.
Contents lists available at ScienceDirect
Advanced Powder Technology journal homepage: www.elsevier.com/locate/apt
Original Research Paper
Particle size analysis on wide size distribution powders; effect of sampling and characterization technique G. Bahar Basim a,⇑, Mohsen Khalili b a b
Department of Mechanical Engineering, Ozyegin University, Cekmekoy, Istanbul, Turkey DuPont Central Research and Development, Wilmington, DE, USA
a r t i c l e
i n f o
Article history: Received 9 June 2014 Received in revised form 5 August 2014 Accepted 22 September 2014 Available online 6 October 2014 Keywords: Particle size distribution Particle sizing techniques Wide particle size distribution powders Sampling
a b s t r a c t Particle size distribution of powders plays a very important role in determining the critical chemical and physical properties of the particulate systems. Precise determination of particle size distribution depends on effective sampling of the powders, which is more pronounced for the particulate systems with a wide particle size distribution. Predominantly, the significant increase in the total surface area of the powders at nanometer scale particle sizes may lead to improper characterization of the bulk if the sampling technique fails to collect and represent them in the analyses. In this study, effects of sampling on the precision of particle size analysis are studied on a clay sample with a wide particle size distribution (particles ranging from nanometer to micrometer sizes) using light scattering technique in aqueous media. Three different sampling methods are applied to systematically analyze the effect of sampling on particle size measurements including; riffling the original sample into sixteen equal parts, sampling the powder after removing the very fine and very coarse size particles and riffling to sixteen parts and finally by riffling the powder to the exact feed amount of the particle size analyzer. The effectiveness of the applied methods were compared statistically by calculating the coefficient of variance (CV) values of the 10%, 50% and 90% passing particle size data of the sequential runs. The most effective sampling method was determined to be riffling the sample to the exact feed amount of the analyzer based on obtaining the minimum CV values of the measurements. In the second part of the study, results of size distribution analyses conducted by different particle size analyzers utilizing numerous characterization techniques are compared using the most effective sampling technique developed in the first part. It is observed that the use of different characterization equipment tend to result in variations in the particle size distributions of the same powder which presents another variability in classification of the wide particle size distribution powders. Ó 2014 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.
1. Introduction Particle size distribution of particulate systems is a critical variable for a number of industrial operations ranging from mining to pharmaceuticals. Precise measurement of particle size distribution of powders is important since the quality and performance of most powder-based products are closely related to the size distribution and particularly the concentration of the fine particles. Although the fine size fraction generally compromise of a small volume, the presence of fine particles leads to immense increase in the specific surface areas, creating a state of high surface energy that dominates properties of the selected system [1]. A good example is increasing moisture content in coal due to fine size content ⇑ Corresponding author. Tel.: +90 216 5649344; fax: +90 2165649057. E-mail address: [email protected] (G.B. Basim).
( 100 mesh) particles retaining more water molecules due to their large surface area that results in elevated filtration, drying and transportation costs [2]. Another critical example applies to microelectronics manufacturing, where the presence of slightly larger size particles at parts per million (ppm) levels in slurry made of nanometer size particles may result in significant surface deformation during the planarization of the wafers by the chemical mechanical planarization (CMP) process [3,4]. For pharmaceutical applications, where more than three-quarters of all tablets and capsules are manufactured using powder blends are more prone to sampling problems since the inaccuracy of sampling may directly affect the human health [5]. Food products, such as skim milk powder, are also hard to sample due to agglomeration during processing which skews the particle size distributions to larger ends and hence may not be representative [6]. Consequently, analysis of a representative powder sample is very necessary to obtain
http://dx.doi.org/10.1016/j.apt.2014.09.009 0921-8831/Ó 2014 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
the correct size distribution of a particulate system in many industrial applications, which in turn helps in enhancing the process efficiency as well as the product performance. The main objective of this study is to evaluate the effect of powder sampling on precision of the particle size distribution analyses and to delineate potential approaches for improving the sampling related variability in size distribution analysis of wide particle size distribution powders. The precision of particle size analysis strongly depends on the effectiveness of the selected sampling technique. In many industrial applications, as discussed above, sampling starts from a large amount of material, yet the collected sample powder must accurately represent this larger entity. Generally, gathering a representative sample with a very small volume is the primary challenge to particle sizing operations. It was reported that the quick and careless sampling of powder often results in an expensive study of an irrelevant or non-representative group of particles [7]. Furthermore, it was also reported on a pharmaceutical application that not only the method of sampling but also the selection of proper sampling device is critical in sampling. Use of different sizes of the thieves for sampling the same drug powder was observed to make significant differences in the strengths of the pills produced [8]. There are new technologies developed for the sampling of powders for delicate applications [9,10], methods that utilize digital imaging [11], in addition to methodologies for sampling large amount of materials (as in mining industry) [12] and many conventional instruments available for decades [13]. One of the most important factors in sampling powders for particle size analysis is whether the characterization will be conducted offline or online [14]. For most of the industrial applications, particle size analyses are conducted offline in the laboratories. After the powder sampled from the production line reaches the laboratory for particle size analysis, further sampling is required to decrease the amount of material to the suitable quantity for the selected particle size analyzer. In the case of online sampling of powders (which is conducted during processing), on the other hand, it is also important to check whether the collection of particles is required, or the information can be obtained by indirect methods. The amount of sample that must be collected should not be disturbing the production process while the cost of the sampling method in money and time, calibration and operational characteristics of the sampling system are additional critical factors to be considered [14]. Optical particle sizers provide robust characterization ability for online monitoring through their ability of covering a wide size range by changing the technique that they utilize through changing the optical configuration [15]. There are standards developed for sampling of powders for particle size analysis, which further verifies the importance of sampling to conduct accurate particle sizing analyses [16]. Determination of the most suitable sampling technique is driven by a number of different considerations including the total mass, number concentration, size distribution and chemical composition of the particles to be characterized in addition to the size distribution technique to be used [17–19]. When sampling powders, there are two types of sampling errors possible that can affect the particle size analysis [20]; (i) errors due to segregation of the bulk powder which can be prevented by suitable mixing and (ii) statistical errors that are due to random fluctuations and cannot be prevented. Both errors are more noticeable when the particle size distribution of the studied powder is wide and variable (changes due to agglomeration or due to breakage of the large size particles by attrition). While the first type of sampling errors can be minimized by suitable mixing and collecting a representative sample, statistical errors cannot be prevented since the quantitative distribution in samples of a given magnitude is not constant even for an ideal random mixture [20]. Yet, increasing the sample size helps estimate the statistical errors and reduce them. Hence it
201
is important to statistically evaluate the results of particle sizing to understand the effect of sampling on precision of the characterization. In this study, a clay sample with a significant variability in particle size was selected as a representative wide size range particulate system for offline characterization at the laboratory. Clays are well characterized for their colloidal properties in water based systems and are used in many industrial applications [21]. A detailed statistical approach was followed to evaluate the variations that sampling may create in particle size analysis. Three different sampling methodologies were implemented by riffling the main powder (i) to sixteen parts and preparing suspensions of these samples in water (ii) using the same procedure after the very fine and very coarse fraction of the powder is removed by screening to tighten the particle size distribution and, (iii) riffling the powder to the exact feed amount of the particle characterization equipment and preparing aliquots of the exact feed amount. The sampled powders were analyzed for particle size distributions at 10%, 50% and 90% passing values to determine the best sampling technique, which produces the most repeatable results in particle sizing. The particle size analyses were conducted using conventional sieve analysis, light scattering and number counting techniques. Initially, Coulter LS 230 (wet) and Microtrac Full Range Analyzer (FRA) instruments were employed to evaluate the best sampling technique. Next the most accurate sampling technique was used to compare the variability in the results of different particle size analyzers including, Coulter LS 230 (both dry and wet), Microtrac FRA (wet), Coulter Multisizer and Sympatec Helos (both dry and wet). 2. Materials and methods 2.1. Materials A broad size distribution bentonite clay sample was selected to study the effect of sampling on the precision of the particle size analysis. The main components of the clay sample were Al3Mg2, CaAl2Si2, AlMg, Al3Mg2, Al2Mg, CaAl2Si1.5, Al0.58 MgO0.42, Si, K and Mg2Si according to X-ray analysis which was consistent with calcium bentonite composition. Fig. 1 illustrates a scanning electron microscope (SEM) micrograph of the clay powder at 200 magnification on which the broad size distribution of the sample was detected clearly. The sieve analyses were also conducted on the powder, confirming the broadness of the size distribution of selected clay as demonstrated in Fig. 2. The fifty percent passing particle size (d50) was calculated as 137.6 lm based on sieve sizing.
Fig. 1. SEM micrograph of the clay powder (Magnification = 200).
202
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
Fig. 2. Sieve analysis of the clay sample.
2.2. Sampling methods Fig. 3 schematically illustrates the sample preparation procedure. Initially, the clay powder was riffled to 16 tubes and each tube’s content was stored in a separate bottle. Three bottles (# 1, 6 and 11) were selected randomly for the first part of the sampling analysis. By mixing the dry powders in the bottles with a spatula, three 0.2 g samples were collected from bottle # 1 and one sample was collected from each of the bottles # 6 and 11. The 0.2-g powders were mixed with 19.8-ml water (a total weight of 20 g, 1-wt%) in 50-ml beakers, shaken vigorously and conditioned for minimum 1-h before the sizing tests were conducted. Another vigorous conditioning was applied before the solutions were fed into the particle sizing equipment to prevent settling of the particles. These sample solutions were named as preparations (P). Three aliquots (A) were collected with a dropper from the first preparation of bottle #1 with an order of A1 was collected from the bottom, A2 was collected from the middle and A3 was collected from the top part of the bottle. Only one aliquot was taken from each of the other preparations by dipping the dropper to the bottom of the suspension in the bottle. These samples were analyzed by Coulter LS 230 (3 runs per aliquot) and Microtrac FRA (1 run per aliquot) light scattering instruments. Coulter LS 230 runs were conducted at
100 rpm stirring speed at 0.4–2000 lm detection range for 90 s, and the Microtrac FRA runs were conducted for 30 s time range. For the second sampling method, the very fine ( 38 lm) and very coarse (+212 lm) fractions of the clay sample were removed by screening from the clay powder using the content of bottle #2. The aim was to evaluate the effect of tightening the particle size distribution on sampling efficiency by removing the end tails of the particle size distribution and determining the repeatability of particle size analyses. Three preparations were made from the screened powder using the same technique (0.2 g powder + 19.8 g water) as described previously. Three aliquots were taken from preparation 1 for the size analysis and only one aliquot was taken from preparations 2 and 3 following the same procedure defined in the first part. For the third sampling method, the exact amount of powder required to conduct a run of Coulter LS 230 and Microtrac FRA were determined (0.04-g). This sampling methodology helps improve the reproducibility of the particle size analyses since the aliquots are prepared from the exact amount of dry powder that is needed to operate the analyzer rather than collecting the sample from an aliquot that has been prepared at a higher volume and hence only some of the sampled dry powder can be fed to the analyzer. The dry sample content of Bottle #8 was riffled to the predetermined exact feed amounts of these analyzers and the particle size analyses were conducted. In addition, to evaluate the effectiveness of the riffling technique, bottles # 4, 5 and 6 were sent to Beckman Coulter Corporation. Using the exact feed amount of Coulter LS 230, they made three preparations from Bottle 4 and one preparation from each of bottle # 5 and 6. Their analyses were found to be in agreement with the conclusions of our study; they reported that riffling of the exact feed amount method gives a very good precision in particle size analysis. To compare the effectiveness of the applied sampling methods in repeatability of the particle size analyses, Coulter LS 230 and Microtrac FRA light scattering instruments were employed. On Coulter LS 230 all aliquots were analyzed three times while recirculating within the equipment to compare variations between runs of the identical sample feed as it keeps on circulating in the wet module of the instrument. The Microtrac FRA runs were conducted only once as the sample cannot be recirculated in the FRA system. The sizes of 10%, 50% and 90% passing fractions of each run were noted for each run. The mean particle size, standard deviation
Fig. 3. Schematic representation of sample preparation procedure.
203
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
(SD) and coefficient of variance values (CV) were calculated to determine the variations in different aliquots, preparations and bottles. CV is the ratio of the standard deviation to the mean value of the distribution multiplied by hundred and it shows the extent of variability in relation to mean of the population. 2.3. Comparison of particle size measurements of different analyzers In the second part of the study the variations in the particle size analysis of different instruments were analyzed using the same clay powder sampled by the riffling method. Three analyzers utilizing light scattering techniques Coulter LS 230, Microtrac FRA and Sympatec Helos were used in addition to one number counting technique (Coulter Multisizer). 3. Results and discussion 3.1. Determination of the best sampling methodology for enhanced repeatability of particle size analyses 3.1.1. Particle size analysis on original clay powder by wet sampling Initially, the effect of sampling on particle size analysis was studied on the original clay powder by preparing 1 wt% suspensions from the selected bottles #1, 6 and 11. The preliminary analyses were performed on the size data obtained from the three aliquots of the same preparation (bottle # 1, preparation 1) by collecting the aliquots from the bottom, middle and top level of the suspension in the bottle named as A1, A2 and A3 consecutively. The aim of this part of the study was to find out whether the level the suspension was sampled from the bottle would affect the particle size measurements. Table 1 summarizes the statistical analysis conducted on the data obtained from the aliquots of the first preparation of bottle #1 by Coulter LS 230 including the mean particle size, standard deviation and the coefficient of variances for the three aliquots as well as the three consecutive runs conducted on the same aliquot feed. It can be seen that both the SD and CV values between the aliquots collected from the same preparation were high especially at the fine fraction reported as 10% passing. Particularly the variation between the 10% passing size of the subsequent runs of aliquot #1, which was collected from the bottom of the bottle, had a very high CV value of 54.8. This variation can be attributed to be due to various factors. First of all, it is possible
to have a larger fine size fraction at the very bottom of the bottle due to hindered settling. Secondly, the fine fraction of the particles might have agglomerated and settled to the bottom of the bottle although the sample was well shaken during the suspension preparation and before it was collected from the bottle. Lastly, larger particles, which settle faster, maybe going through attrition generating an additional fine fraction at the bottom of the suspension. It also appears that as the same feed is run in the equipment again and again, 10% passing particle size tend to decrease as can be seen for aliquot #1 for runs 1, 2 and 3 measured as 17.7, 9.1, 6.1 lm average size, respectively. All three factors discussed above, hindered settling, agglomeration or attrition maybe leading to this observation. Furthermore, it is suspected that the sequential runs conducted by keeping the particles circulating in the instruments may result in breakage of the larger size agglomerates by enhanced stabilization of the particles. This argument can further be supported by the fact that the particle suspension fed into the analyzer is further diluted in the volume of water that is in the equipment’s measurement cell, which is 200 ml for Coulter LS 230 decreasing the suspension to a 0.1 wt% concentration effectively. Aliquot #2 and #3, which were collected from the middle and the top of the bottle, had more consistent results as can be seen from the SD and CV values of Table 1. Nonetheless, the true variability of the particle size measurements at the fine fraction is only represented when the sample is collected from the bottom of the bottle since the effective particle size distribution may include both flocculated particles and aggregates as well as discrete particles [22]. This variance is reflected when the aliquots 1, 2 and 3 are compared for run 1 at 10% passing sizes reporting a CV value of 76 indicating that the position where the sample is collected within the same suspension significantly affects the size analyses for the fine size evaluations. In order to represent the potential of the fine size fraction of the
Table 2 Original clay sample statistical analysis for Microtrac FRA runs. Bottle #1, preparation 1, aliquots 1, 2, 3. Microtrac FRA
Aliquots
% Passing
1
2
3
(lm)
92.3 200.2 360.5
74.3 182.5 364.5
67.7 180.2 335.0
78.1 187.6 353.3
Run 1
10% 50% 90%
Mean
SD
CV
12.7 10.9 16.0
16.3 5.8 4.5
Table 1 Original clay sample statistical analysis for Coulter LS 230 runs. Bottle #1, preparation 1, aliquots 1, 2, 3. Coulter LS 230
Aliquots
% Passing
1
2
3
(lm)
Mean
SD
CV
Run 1
10% 50% 90%
17.7 140.2 283.7
6.4 119.1 263.2
4.3 94.1 229.7
9.5 117.8 258.9
7.2 23.1 27.3
76.0 19.6 10.5
Run 2
10% 50% 90%
9.1 139.3 281.6
5.1 118.0 259.8
4.0 91.6 227.5
6.1 116.3 256.3
2.7 23.9 27.2
44.1 20.5 10.6
Run 3
10% 50% 90%
6.1 138.9 282.1
4.6 117.1 257.6
3.8 91.0 225.3
4.8 115.7 255.0
1.2 24.0 28.5
24.2 20.7 11.2
Mean
10% 50% 90%
10.9 139.4 282.5
5.4 118.1 260.2
4.0 92.3 227.5
SD
10% 50% 90%
6.0 0.7 1.1
0.9 1.0 2.8
0.3 1.6 2.2
CV
10% 50% 90%
54.8 0.5 0.4
16.4 0.8 1.1
6.8 1.8 1.0
204
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
Table 3 Original clay sample statistical analysis for Coulter LS 230 runs. Bottle #1, preparations 1, 2 and 3 (aliquots 1). Coulter LS 230
Preparations
Mean
SD
CV
1
2
3
(lm)
Run 1
10% 50% 90%
17.7 140.2 283.7
13.3 164.6 304.2
19.6 167.5 304.8
16.8 157.4 297.6
3.2 15.0 12.0
19.1 9.5 4.0
Run 2
10% 50% 90%
9.1 139.3 281.6
9.3 165.3 303.4
11.3 168.1 301.6
9.9 157.6 295.5
1.2 15.9 12.1
12.4 10.1 4.1
Run 3
10% 50% 90%
6.1 138.9 282.1
7.3 164.4 300.0
8.8 169.9 300.5
7.4 157.7 294.2
1.3 16.5 10.5
17.9 10.5 3.6
% Passing
Table 4 Original clay sample statistical analysis for Microtrac FRA runs. Bottle #1, preparations 1, 2 and 3 (aliquots 1). Microtrac FRA
Preparations
% Passing
1
2
3
(lm)
92.3 200.2 360.5
81.3 187.2 361.4
72.0 176.7 346.5
81.9 188.0 356.1
Run 1
10% 50% 90%
Mean
SD
CV
10.1 11.8 8.4
12.4 6.3 2.3
powders to agglomerate or added into the system by attrition of larger ones, the remaining of the wet sampling analyses were conducted by collecting the aliquots from the bottom of the sample bottles after the bottle content was shaken vigorously. Table 2 shows the results of the same analysis conducted on the Microtrac FRA runs, for which the sample was discharged immediately after the run and therefore could not be rerun in the system. It can be seen that the cut sizes obtained with Microtrac FRA were higher compared to Coulter LS 230 results. However, the maximum variation was still observed on the fine fraction. These results once again confirm that regardless of the selected sizing equipment, samples collected from the same preparation but from different positions in the container might show a significant difference depending on how they are sampled. Tables 3 and 4 show the statistical analyses on the tree 1 wt% solution preparations prepared from the dry powder of bottle #1 with Coulter LS 230 and Microtrac FRA, respectively. The variability in this part comes from sampling the 0.2 g of powder from the same bottle. These results also show a high variance on the fine fraction measurements yet there is a relative decrease in the overall CV values as compared to the previous results reported in Tables 1 and 2. The main difference is that, in Table 1 aliquots of the same preparation were compared for particle size with respect to the position they were collected from the bottle. In this part of the sampling comparison, the aliquots were all collected from the bottom of the bottles for three different preparations of the dry powder from the same bottle. The reduced variability in CV values in this set of testing highlight that the collection of an aliquot from a suspension can change the particle size distribution of the sample more than how the sample suspension was prepared from a given dry powder. The observation on the decrease in the measured particle size for the sequential runs remained the same in this test as well once again supporting the discussion of enhanced stability of powder as it remained for a longer time in the sizing equipment. Nevertheless, since most of the sizing systems give a sufficient conditioning time to the particles before the characterization can be started and additionally due to the fact that not all the sizing equipment has the ability of rerunning the samples, it is suggested that the first set of measurements are taken into consideration or a consistent approach is followed by
the operator in sampling the suspensions and running them in the analyzers. Finally, the particle size analyses obtained from the three different bottle samples were compared as illustrated in Tables 5 and 6. The results of the first aliquots of the first preparations of each bottle were analyzed for each bottle. The CV values were higher as compared to the sampling from the same bottle for three times yet more consistent as compared to the aliquots collected from the same bottle but different positions. Also within the two analyzers, Coulter LS 230 was observed to be more sensitive in detecting the fine size fraction as compared to Microtrac FRA. This sampling method makes use of a very small amount of the bottle content that may account for the high variation between size data obtained from different bottles, even though the contents of the bottles may be similar. 3.1.2. Sampling analysis on the screened clay powder The results obtained on the original clay sample prepared as wet suspensions showed high variability as reported by SD and CV values in particle size analysis. The main source of the poor repeatability was suspected to be due to the wide particle size distribution of the clay powder. In order to understand the effect of controlling the width of the size distribution and decrease the variation in measurements, the very coarse and very fine fractions were screened out of the original powder forming a particulate sample contained within 38–212 lm size range. Since it was observed that Coulter LS 230 was more sensitive in detecting the variation, this set of testing was only conducted using this equipment. Three 20-g suspensions of 1 wt% were prepared using the screened powder and three aliquots were taken from the first preparation again from the bottom, middle and the top of the suspension. In addition, tree suspensions prepared from three different bottles were also compared. Table 7 summarizes the results of the statistical analyses conducted on the screened powders by Coulter LS 230. It can be seen that the sampling variation decreased significantly both among the three aliquots collected from the different positions of the same preparation (SD 0.6–0.8 and CV 0.7–1 for 10% passing) and for the aliquots sampled from the bottom of the tree different preparations (SD 5.3–6.0 and CV 6.9–7.7 for 10% passing). These results further confirm the discussion of wide size distribution creating a significant variability in sampling. Therefore, as expected, the system particle size distribution approaches to monosize, precision of the particle size analysis tends to improve significantly [3]. 3.1.3. Sampling analysis on clay powder riffled to exact feed amount Although separation of the very fine and coarse fraction of particles from the original clay powder helped improve the repeatability of the particle size analysis through facilitating better quality of sampling, the size distribution of the original clay powder was
205
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207 Table 5 Original clay sample statistical analysis for Coulter LS 230 runs. Bottles 1, 6 and 11 (preparations 1, aliquots 1). Coulter LS 230
Bottles
Mean
SD
CV
1
6
11
(lm)
Run 1
10% 50% 90%
17.7 140.2 283.7
9.4 160.9 297.2
27.9 170.4 308.9
18.3 157.2 296.6
9.3 15.4 12.6
50.5 9.8 4.3
Run 2
10% 50% 90%
9.1 139.3 281.6
7.5 162.3 297.0
17.7 168.5 305.7
11.4 156.7 294.8
5.5 15.4 12.2
47.8 9.8 4.1
Run 3
10% 50% 90%
6.1 138.9 282.1
6.3 162.8 295.5
13.3 167.1 302.9
8.6 156.3 293.5
4.1 15.2 10.5
47.6 9.7 3.6
% Passing
Table 6 Original clay sample statistical analysis for Microtrac FRA runs. Bottles 1, 6 and 11 (preparations 1, aliquots 1). Microtrac FRA
Bottles
% Passing
1
6
11
(lm)
92.3 200.2 360.5
76.8 186.6 348.2
82.7 190.9 334.1
83.9 192.6 347.6
Run 1
10% 50% 90%
Mean
SD
CV
7.8 7.0 13.2
9.3 3.6 3.8
changed knowingly. In order to analyze the size distribution of the original powder with a wide particle size distribution a third technique was developed. In the first sampling method where a wet suspension was prepared from the original powder, a high variation was observed between the aliquots of the same preparation. This lead us to develop a new technique in which the powder is riffled to the exact feed amount of a single run for the particle size analyses to be conducted. In order to determine the exact amount of dry powder required to conduct a single measurement, an aliquot with a known volume was fed to the tool until the tool’s sensor indicated that the number of particles were sufficient to make the measurement. The volume fed to the tool was calculated by measuring and subtracting the remaining sample volume from the initially prepared sample volume. The exact feed amount was determined to be 0.4-g for Coulter LS 230 through drying the exact volume of one feed to determine the solids content. The calculation was cross verified through preparing an exact feed with the calculated solids loading in one aliquot and conducting the measurements again. Once this amount was determined, the original dry clay powder was riffled to this exact amount and sixteen suspensions were prepared using these riffled powders. The contents of eight even numbered riffle tubes were used for the Coulter LS 230 analysis and the other eight were run through the Microtrac FRA. To conduct a run, all of the suspension in a tube was fed into the analyzer. Table 8a and b
shows the comparison between the first sampling method and the third sampling method on Coulter LS 230 and Microtrac FRA, respectively. The statistical analyses on the preparations of the same bottles of the wet sampling technique were compared to the analysis on the samples riffled to exact feed amount of the analyzer. The calculated SD and CV values decreased for all the 10%, 50% and 90% passing fractions by about 2–3 times when the size analysis performed by the exact feed amount of powder. It is obvious from the results given in Table 8 that the riffling technique helped improve the precision of the particle size analysis regardless of the particle sizing equipment used for characterization. A more systematic study was conducted on the effect of riffling as a sampling method using Coulter LS 230 at Beckman-Coulter Inc. The exact amount of powder required for a run was riffled out of three bottles numbered 4, 5 and 6. Three preparations were made from bottle #4 and single preparations were made from the other two bottles. Table 9 shows the results of the statistical analyses conducted on the size analysis of the three bottles and the three preparations of the bottle #4. The results indicate a significant improvement in the repeatability and precision of the size analysis. According to these results we can conclude that the riffling the sample to the exact feed amount is the most effective sampling technique that can be used to represent the size distribution of the wide particle size powders. 3.2. Comparison of different particle analyzers From the first part of the analysis detailed in the previous sections, riffling the powder to the exact feed amount needed for the particle size analyzer was determined to be the most effective way of sampling for a wide particle size distribution sample. Hence, in this part of the study, where the measurements of various particle sizing equipment are compared and their consistency in sizing was investigated, the exact feed amount of the powder was riffled and prepared for every equipment used.
Table 7 Screened clay sample statistical analysis for Coulter LS 230 runs. Comparison of aliquots and preparations. Coulter LS 230
Aliquots
% Passing
Mean (lm)
SD
CV
Mean (lm)
Preparations SD
CV
Run 1
10% 50% 90%
76.7 146.9 234.7
0.8 0.3 2.0
1.0 0.2 0.8
76.5 147.7 236.1
5.3 5.2 4.2
6.9 3.5 1.8
Run 2
10% 50% 90%
77.4 148.1 234.9
0.6 0.6 1.9
0.7 0.4 0.8
77.3 148.7 236.1
5.7 5.6 4.1
7.4 3.7 1.7
Run 3
10% 50% 90%
77.8 148.6 234.5
0.7 0.8 1.7
0.9 0.5 0.7
77.8 149.2 236.1
6.0 5.8 4.7
7.7 3.9 2.0
206
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
Table 8 Comparison of the statistical analysis for original wet suspended and riffled clay samples (a) with Coulter LS 230 and (b) Microtrac FRA. Wet Suspended Mean
Riffled to exact feed amount SD
CV
Mean
Coulter LS 230 % Passing Run 1 10% 16.8 50% 157.4 90% 297.6
SD
CV
3.2 15.0 12.0
19.1 9.5 4.0
13.8 143.6 287.1
1.4 4.5 3.9
10.1 3.1 1.4
Microtrac FRA % Passing Run 1 10% 81.9 50% 188.0 90% 356.1
10.1 11.8 8.4
12.4 6.3 2.3
55.9 163.7 313.9
3.8 2.4 17.8
6.7 1.5 5.7
Table 9 Riffled clay sample statistical analysis for Coulter LS 230 runs. Comparison of the samples riffled from three different bottles and three preparations. Coulter LS 230
Preparations
% Passing
Mean
SD
CV
Mean
Bottles SD
CV
Run 1
10% 50% 90%
14.4 142.0 303.3
0.4 4.4 4.1
2.8 3.1 1.3
14.5 141.7 302.3
0.2 2.5 1.5
1.1 1.8 0.5
Run 2
10% 50% 90%
14.2 141.7 303.0
0.4 4.9 3.6
2.5 3.5 1.2
14.3 141.7 301.7
0.1 2.5 2.1
0.4 1.8 0.7
Run 3
10% 50% 90%
14.0 141.0 301.7
0.3 5.2 4.1
2.5 3.7 1.4
14.2 141.7 301.7
0.1 2.5 1.2
0.4 1.8 0.4
There are many different measurement principles including mechanical separation (sieves), light diffraction, sedimentation, photon correlation spectroscopy, dynamic light scattering, field flow fractionation, electrophoretic mobility and induced acoustic response for particle characterization. Among these, light scattering is the method of choice as it has a robust theory (diffraction) to calculate the particle size in addition to high speed, good reliability and high reproducibility [22]. Hence, we studied three different light scattering instruments and one number counting technique in this part of the analysis. For the light scattering tests, Coulter LS 230 (wet and dry), Microtrac FRA (only wet) and Sympatec Helos (wet and dry) were utilized. In addition, Coulter Multisizer analyzer was used for the number counting analysis. Fig. 4 shows the cumulative volume percent passing versus particle
size data obtained from the selected instruments and Table 10 summarizes the numerical values for the statistical analysis. It can be seen that the average (mean) particle size, 50% passing (median) and the most repeated particle size (mode) are not in a very good agreement for different instruments. Even the dry and wet analysis of the same instrument may give different particle size distributions. This tendency was also observed by other researches, and was attributed to the diversity of the measuring principles and instrument models [23–26]. Particularly, Hayakawa and coworkers studied spherical particles, granulated particles and particles with various shapes in a systematic way and also concluded that the method of sampling affects the particle size measurements significantly [24–26]. In this analysis, we have minimized the variation given by sampling yet still observed a significant variability among different particle sizing methods and even with equipment using the same methodology in particle sizing but made by various different manufacturers. It is important to be aware of the differences between the various particle sizing techniques and select the most effective one for the purposes of the conducted analysis and maintain a consistency when analyzing particulate systems.
Table 10 Comparison of the consistency of different particle size analyzers.
Fig. 4. Cumulative particle size distribution of clay sample obtained by using different analyzers.
Sieve analysis Microtrac FRA Coulter Multisizer Coulter LS (wet) Coulter LS (dry) Sympatec Helos (wet) Sympatec Helos (dry)
Mean size (lm)
Median (lm)
Mode (lm)
185.0 178.2 144.1 144.2 150.0 147.3 112.3
137.6 161.7 148.3 133.5 138.1 134.5 97.3
100.0 209.3 181.5 161.2 213.1 174.0 174.0
G.B. Basim, M. Khalili / Advanced Powder Technology 26 (2015) 200–207
4. Summary The results of this study showed that the employed sampling technique has an important effect on the precision of particle size analysis. The most effective sampling technique requires the protection of the original particle size distribution of the particulate system during the analysis. Otherwise, the sample is not representative of the original powder and the size analysis may display a high variation. As the broadness of the size distribution of the powder increases, sampling gets more critical since it becomes more difficult to obtain a representative sample. According to our analysis, the most effective sampling technique was found to be the riffling of the powder to the exact amount required by the analyzer. It was also shown that the results of the size analysis of a particulate system may vary based on the selected technique and instrumentation. These issues also indicate the requirement of international standards on sampling for particle size characterization, which are currently being prepared and are believed to address most of the problems in particle science and technology.
Acknowledgments The authors would like to acknowledge the support from Mr. W. Rohricht at the Particle Science and Technology Center at DuPont Central Research & Development. The valuable contributions of Beckman Coulter Inc. are also acknowledged.
References [1] E. Matljevic, The world of fine particles (ultrafine particles produced by precipitation or chemical reactions with drops), ChemTech 21 (1991) 176–181. [2] G.B. Basim, Fine Coal Dewatering, MS Thesis, Virginia Polytechnic Institute and State University, Blacksburg, VA, USA, 1997. [3] G.B. Basim, U. Mahajan, J.J. Adler, B.M. Moudgil, R.K. Singh, Effect of particle size distribution of chemical mechanical polishing slurries for enhanced polishing with minimal defects, J. Electrochem. Soc. 147–9 (2000) 3523–3528. [4] G.B. Basim, B.M. Moudgil, Effect of soft agglomerates on CMP slurry performance, J. Colloid Interface Sci. 256–1 (2002) 137–142. [5] F.J. Muzzio, C.L. Goodridge, A. Alexander, P. Arratia, H. Yang, O. Sudah, G. Mergen, Sampling and characterization of pharmaceutical powders and granular blends, Int. J. Pharm. 250 (2003) 51–64.
207
[6] C. Turchiuli, R. Smail, E. Dumoulin, Fluidized bed agglomeration of skim milk powder: analysis of sampling for the follow-up of agglomerate growth, Powder Technol. 28 (2013) 161–168. [7] B.H. Kaye, R. Trottier, The many measures of fine particles, Chem. Eng. 102 (1995) 78–86. [8] J.A. Berman, A. Schoeneman, J.T. Shelton, Unit dose sampling: a tale of two thieves, Drug Dev. Ind. Pharm. 22 (11) (1996) 1121–1132. [9] F.J. Muzzio, P. Robinson, C. Wightman, D. Brone, Sampling practices in powder blending, Int. J. Pharm. 155 (1997) 153–178. [10] L. Susana, P. Canu, A.C. Santomaso, Development and characterization of a new thief sampling device for cohesive powders, Int. J. Pharm. 416 (2011) 260–267. [11] A. Boschetto, V. Giordano, Powder sampling and characterization by digital image analysis, Measurement 45 (2012) 1023–1038. [12] H.R. Cooper, New technology provides better crushed-ore sampling results, Eng. Min. J. Press 190–6 (1989) 54–59. [13] H.A. Behre, M.D. Hassialis, Sampling and testing, in: A.F. Taggart (Ed.), Mineral Dressing Handbook, John Wiley & Sohns, New York, 1945, pp. 1–208. [14] C.H. Murphy, Handbook of Particle Sampling and Analysis Methods, Verlag Chemie International, Deerfield Beach, Florida, 1984. [15] I. Goaninoni, E. Golinelli, G. Melzi, S. Musazzi, U. Perini, F. Trespidi, Optical particle sizers for online applications in industrial plants, Opt. Lasers Eng. 39 (2003) 141–154. [16] ISO TC 24/SC4/WG11, Sample Splitting for Particle Size Characterization, 1996. [17] G. Guidarelli, F. Craciun, C. Galassi, E. Roncari, Ultrasonic characterization of solid-liquid suspensions, Ultrasonics 36 (1998) 467–470. [18] J.F. Fabries, R. Wrobel, P. Gorner, D. Bemer, P. Bonnet, H. Nunge, M. Lafontaine, S. Binet, Comparison of particle-size distributions of bitumen fumes measured by aerosizer and QCM impactor techniques, J. Aerosol Sci. 31 (2000) 1011– 1012. [19] S. Jaffari, B. Forbesa, E. Collins, D.J. Barlowa, G.P. Martina, D. Murnanec, Rapid characterisation of the inherent dispersibility of respirable powders using dry dispersion laser diffraction, Int. J. Pharm. 447 (2013) 124–131. [20] T. Allen, Powder Sampling and Particle Size Determination, Elsevier Ltd., London, UK, 2003. pp. 1–55. [21] G. Lagaly, I. Dékány, Colloid clay science, in: F. Bergaya, G. Lagaly (Eds.), Handbook of Clay Science-Fundamentals, Developments in Clay Science, vol. 5, Oxford, UK, 2013, pp. 243–345. [22] Z. Ma, H.G. Merkus, J.G.A.E. de Smet, C. Heffels, B. Scarlett, New developments in particle characterization by laser diffraction: size and shape, Powder Technol. 111 (2000) 66–78. [23] J.M. Phillips, D.E. Walling, An assessment of the effect of sample collection, storage and suspension on the representativeness of measurements of the effective particle size distribution of fluval suspended sediment, Wat. Res. 29– 11 (1995) 2498–2508. [24] O. Hayakawa, K. Nakahira, J. Tsubaki, Comparison of particle size analysers and evaluation of its measuring technique with fine ceramic powders (Part 1), J. Ceram. Soc. Jpn. 103 (1995) 392–397. [25] O. Hayakawa, K. Nakahira, J. Tsubaki, Comparison of particle size analysers and evaluation of its measuring technique with fine ceramic powders (Part 2), J. Ceram. Soc. Jpn. 103 (1995) 500–5005. [26] M. Naito, O. Hayakawa, K. Nakahira, H. Mori, J. Tsubaki, Effect of particle shape on the particle size distribution measured with commercial equipment, Powder Technol. 100 (1998) 52–60.