Comparison of Optical Methods for Fines and Filler Characterization

(1)

Comparison of optical methods for fines and filler characterization

Farnaz Farahani

December 2015

Master Thesis Report in Chemical Engineering

(2)

Acknowledgements

The support and supervision of Lars Wågberg, Kari Hyll, support and management of Hannes Vomhoff, Elisabeth Björk, contribution of lab assistants in the chemical lab of Innventia AB, especially Åsa Blademo, sample contribution of Specialty Minerals, and support by the Vinnova EucaBraz project is gratefully acknowledged.

(3)

1

1 Summary

This thesis presents an evaluation study of three different image analysis instruments (FiberTester, FiberTester+ and ImageStream) and a laser diffraction (LD) instrument, for the analysis of the fine fraction of a stock. The instruments had different spatial resolutions and measurement ranges.

Measurements were made on three different samples; pulp fines, paper filler

(precipitated calcium carbonate) and the mix of them. Two comparisons were made;

one with only data from image-based analysers, and one where LD data was also included. In the first comparison, the data was area-weighted, while in the second comparison, it was volume-weighted. To conduct a meaningful data comparison between the imaging and LD techniques, the multi-parameter imaging data was transformed to Equivalent Sphere Diameter.

For the pulp fines sample, LD did not exhibit particles with Equivalent Sphere Diameter smaller than 6 μm and showed poor correlation with Image-based analysers. However, the LD results correlated better with Image-based methods for PCC particles. The ImageStream was the most capable instrument for detecting smallest particles among all the analysers. Among the image-based analysers, FT+ delivered the best results for relatively large particles.

(6)

4

2 Introduction

Fines play an important role in behaviour of the wet web and quality of the final paper.

Many studies have revealed the significance of fines and their impact on retention, drainage, paper making chemistry, optical properties like light scattering and

mechanical properties of the paper like tensile index, tear index, roughness, and finally density of the paper sheets. Since fines have a large surface area compared to their mass, they play an important role in drainage and bonding ability. Despite the significance of fines, limited knowledge of their characteristics and morphology is available, as their small size makes them difficult to characterize. The importance of particle

characterization in general arises from the fact that many of physical, mechanical and in some cases chemical properties of dispersed materials depend on their particle size and shape characteristics. Not only the size of these particles is critical to understand, but also the shape of these particles plays an important role in behaviour of the dispersed material. The shape of the particles can also affect particle process in addition to the properties of the final products.

Spherical particles will behave differently from needle-like particles. In powder particle processes spheres flow easily, but needle-likes do not, the viscosity increases by

increasing aspect ratio. Blending time is different for different shapes of particles and mobility of the particles depends on their shape to a large extend and it causes

segregation. Product performance is largely influenced by shape of particles as well. For some products like glass beads for highway paint or propend, a better performance is achieved by spherical particles, but for some other products like abrasives, non- spherical particles have better performance [2].

In the pulp and paper industry, image-based fibre analysers are common for the

characterization of pulp, and here primarily for the characterisation of the morphology of the fibres. These analysers give both the size and shape of particles. Although image analysis is an affordable and powerful technique, the method has limitations. Image analysis provides measurements on a 2-D image of a 3-D particle. In addition, particles can only be detected if they are larger than the resolution of the instrument. Given the fact that a large share of the fines and filler particles are considerable smaller than the resolution of the instruments, a large part of the fine fraction cannot be characterized.

This has prompted researchers to use other methods, such as laser diffraction, in studies of fines and fillers.

From laser diffraction, highly reproducible results can be provided in a short amount of time. On the other hand, laser diffraction data is difficult to interpret and to correlate

(7)

5

with results from other methods. The interpretation is especially challenging for non- spherical shaped particles since it considers particles as perfect spheres. Additionally, it is sensitive to the optical properties of the particles. These limitations have only rarely been addressed in the studies.

Recent development of image-based analysers has increased the resolution of image- based analysers. Thus, the purpose of this study was to evaluate the potential of image- based instruments for the size and shape analysis of fines and fillers, and to further investigate the performance of laser diffraction. This was done by analysing results from the use and comparison of image analysis (IA) and laser diffraction (LD) particle size analysis (PSA).

(8)

6

3 Background

3.1.1 The fine fraction of a stock

The main components in the stock used in the papermaking machine include fibers, cellulosic fines, fillers, polymer additives to enhance retention and fixation and finally dissolved colloidal substances (DCS). The fraction of the stock which passes through holes with a diameter of 75 μm (200-mesh screen) is commonly called the fine fraction.

Dependent on the composition of the stock, the fine fraction can contain many different particle types.

3.1.2 Definition of fines

Fines are the components of the fine fraction which originate from the pulp. Fines may contain everything from bundles of elementary fibrils 20 nm wide and 1 µm long, to ray cells and fibre fragments several hundreds of micrometers in length. Thus, the size range of fines is very wide. There is an unclear boundary between DCS (dissolved and colloidal substances) and fines, and also between fines and cellulose micro/nano-fibrils.

According to Luukko et al [3] fines are visible in a light microscope but the size boundary between “colloidal” and “microscopic” is not clear [4].

In order to understand the properties of fines and their influence on paper sheet properties, knowledge of both chemical composition, such as lignin and carbohydrate content, and physical properties, such as shape and size, are desirable.

3.1.3 Classification and origin of fines

Mechanical fines have traditionally been categorized as flaky or fibrillar fines. Chemical fines have been categorized as primary and secondary fines. Primary fines are those, which are a result of the chemical pulping process, and already exist in the pulp before refining, a mechanical treatment process. Secondary fines are produced during the refining. Primary fines are mostly flake-like and secondary fines are fibril-like. During the refining, the surface (primary wall) of the fibers is peeled off, exposing the

secondary layer. The microfibrils of the secondary wall are raised to the surface as external fibrillation, and may be rubbed off the fibre. Thus, secondary fines are

produced and may origin both from the primary wall and the S-layers. Since the surface of the fibers contains more lignin than the bulk fibers, the secondary fines lignin content is higher than that in fibers but still less than that in primary fines. Primary fines

composition is mainly from ray cells, parenchyma cells and middle lamella lignin. The primary fines consist of higher lignin and slightly higher xylose than those in fibers but there is no significant difference in their carbohydrate compositions. The carbohydrate composition in secondary fines and fibers is almost similar. The primary and secondary fines each have different impact on mechanical properties of the sheets. The secondary fines contribute to the tensile strength more than the primary fines [6].

(9)

7

Figure 1. Mechanical Fines produced after first refining stage, flake-like (left) and fines produced after last refining stage, fibril-like (right) [4].

For mechanical fines, the type is related to the mechanical treatment. Rundlöf [5]

studies on mechanical fines reported that flake-like fines mainly come from middle lamella and fibril-like fines from the secondary wall. Luukko et al [3] also added that low total specific energy consumption (SEC) in the refining process will produce mainly flake fines and high total specific energy consumption gives flake fines in the early stages and fibril-like fines in the finishing stages of refining process.

Figure 2. Different types of fines and their origin in the fibre [27].

(10)

8

3.1.4 Impact of the size and shape of fines

Fines and fibers are greatly different in their physical properties and their impact on processes and products are different. Due to their small size, large ability to swell, and big specific surface area, the fines affect paper properties in many ways. Several of these fines-unique properties are dependent on their size and shape. A few examples are given below.

Due to their small size, fines can migrate into and fill inter-fiber spaces during the dewatering. Therefore, for the same dewatering condition, the sheet dryness will decrease by the existence of fines.

Chemical fines give the highest increase in sheet strength properties of the pulp;

especially the secondary fines. This has been linked to the higher degree of fibrils of secondary chemical fines, compared to mechanical fines [6]. The high aspect ratio and flexibility of the fibrillar fines are believed to be especially important.

In mechanical pulp, flake-like fines contribute more to the scattering of light in the sheet due to the low density and their shape. Here, the flaky shape may act as a mirror for the light. Fibril like fines are more flexible and can be densely packed and therefore create less light scattering [7].

Lignin swells less than carbohydrates. Since flake-like fines contain a higher amount of lignin than fibril-like fines, they are expected to swell less. On the other hand, charged groups present in fines play an important role in swelling of fines. Flake-like fines originate from the pectin rich polysaccharides, which contain more total charged groups but less surface charged groups than fibril-like fines. The specific surface area of the fibril-like fines is higher than that in flake-like fines, which can be the reason of higher amount of surface charge of fibril-like fines [4]. Finally, it should be mentioned that the large swelling of the fines may affect their optical properties, as the optical properties change with the amount of water.

The chemical composition of flake- and fibril-like fines in mechanical pulp is different.

The lignin content of flake-like fines in mechanical pulp has been reported to between 35-39%, while it was 31-33% for the fibril-like fines [7-9].

As the specific surface area (surface area per unit mass) is much larger for fines than for fibers (in the range of ten times), they are potentially able to adsorb larger amounts of contaminant in the process water [4]. Moreover the flake-like fines adsorb more colloidal wood resin than the fibril-like fines, and it is probably due to higher hydrophobicity in flake-like fines [4].

In conclusion, previous studies on both chemistry and physical properties of fines illustrate their important role for the quality of the paper. The shape-dependent categorization into flaky and fibrillar fines is important, which is in turn linked to the origin of the fines in the fibre or in the pulp production process. Summarizing the points

(11)

9

described above, a paper sheet, which contains fibril-like fines, has a higher tensile index, lower light scattering coefficient, and a higher density than a sheet containing flake-like fines.

3.1.5 Fillers

Fillers are pigment powders that are produced from natural mineral products. They have a size of 0.1 to 5 𝜇m. Fillers provide desirable optical paper properties such as high light scattering, brightness and opacity. They also improve surface smoothness and

dimensional stability of the paper. Due to reduced amount of biomaterial mass per unit weight of paper when using paper fillers, the energy demand in pulp and paper making process is lower [10]. Fillers are also used in the paper making industry due to the scarcity of cellulose raw materials, increasing costs and for forest resources preservation purposes. Fillers are five to seven times cheaper than bleached chemical pulps [11].

While there are benefits with applying fillers in papermaking, there are still some challenges associated with the fillers. Firstly, fillers have poor binding, considering the fact that both fibres and fillers are negatively charged and they repel each other and that affects the strength of the paper. Therefore, they do not easily retain on the sheets.

Fillers can also disturb fibre-fibre bonding and negatively influence tensile strength of the sheets. There are a number of commercially-available fillers from which precipitated calcium carbonate (PCC), ground calcium carbonate (GCC), and kaolin are most

common. Other pigments include talc, precipitated silica and silicates (PSS) and TiO2. GCC and PCC are used in neutral or alkaline paper production whereas PSS, talc, and TiO2 are used in alkaline/neutral or acid papermaking processes. The demand for calcium carbonate pigments is currently growing while there is a marked decline in kaolin demand. The main reason for this is the conversion of acid papermaking to neutral/alkaline and the demand for brighter and bulkier papers [12].

Characterizing the size and shape of mineral filler particles will help us to understand how they can affect paper properties. Bulkiness and porosity of the paper sheet can be strongly influenced by the filler particle size and shapes.

Previous studies show how different sizes of fillers can affect the paper properties. The figures below show some of the optical and mechanical properties of paper sheets are influenced by the size of the fillers see Figure 3.

(12)

10

Figure 3. Effect of paper filler size on some of the paper sheet properties [28].

Mineral fillers are produced in different shapes, such as plate-like, irregular, blocky or rounded. PCC is produced in different crystal shapes, whereas GCC particles have generally a blocky shape, see figure 4 and figure 5.

Figure 4. Micrograph of rhombohedral (left) and aragonite PCC (right) [28].

(13)

11

Figure 5. SEM of hydrous Kaolin (left) and Micrograph of GCC (right) [28].

Precipitated calcium carbonate (PCC) is used as filler and sealants, in food and pharmaceuticals, paints, inks, and paper. PCC provides high opacity and brightness to the final paper and is the mineral that is mostly used in paper products. The demand for PCC has been growing from 275,000 metric tons in 1986 to more than 4 million metric tons today.

Since PCC is one of the most common fillers used in paper industry, and allows a flexibility regarding the location of the manufacturing site and the produced shapes of PCC particles, it would be interesting to characterize its size and shape and therefore was chosen as the second sample of the study in this work.

3.1.6 Impact of the size and shape of fillers

PCC is manufactured and not mined. Therefore, chemical and physical properties of PCC particles are controllable [13]. Depending on the desirable final physical properties, which are affected by the shape of the particle, a formulator can choose a shape which gives the best performance to the final product. The shape of PCC crystals depends on the products and the production process, see example of different shapes and sizes in figure 3. As by increasing the degree of PCC structure, a bulkier product with a better optical performance is achieved. Degree of structure or “structuring”

defines the aggregation of primary particles of PCC to a larger secondary structure [14].

3.1.7 Mixes

In several situations, fillers and fines are mixed in a stock. This is especially common for recycled pulp, but also in white water and in pulp containing broke. As the

properties of fines and fillers are very different, a stock containing much filler could need addition of fines to compensate for the loss in paper strength. A stock containing a higher share of fines could allow for the addition of more fillers to save raw material cost and improve the optical properties. However, there are currently no efficient

(14)

12

automatomatized image-based methods which can identify if a particle is filler or a pulp fine.

3.2 Stock characterization methods

Given that the size and shape of fine material is of great importance for product properties, surprisingly few methods are dedicated to their characterization. The wide size range of the fine material is a great challenge, as both the smallest and the largest particles needs to be detected at the same time. This requires a measurement instrument to have both a high resolution and a large dynamic size rang.

Optical Characterization methods

Optical fiber length analysis is a common method for obtaining fiber length

distributions. As they are image-based, they can give both size and shape data for every particle. However, there are some limitations with current optical fiber length analyzers.

First, the image resolution of the systems is limited. Single-camera fibre analyzers have had a resolution around 10 µm. Three to five pixels are required to build up a particle, so the minimum particle size has in practice been 30-50 µm. Newer, two-camera systems have added a camera with resolution around 1.5µm, but these have been expensive. With a significant share of the fine fraction being smaller than this, many fines particles will not be detected.

Moreover, even if all the fines would be detected, the weight fraction of fines cannot be obtained from the results. Although optical analyzers provide a length-weighted

percentage of particles smaller than 0.2 mm length, this result cannot be reliable, as these results would only be correct if fibers and fines had the same mass per unit length (coarseness). Thus, not only all fines are not detected but also the fine content value obtained from these methods is not reliable [15].

Some of the optical fiber length analyzers detect the fibers using polarized light. Since cellulose is birefringent and lignin is not, only wood particles, which contain cellulose, would be detected. Therefore lignin rich particles including fines from mechanical pulps cannot be detected. Even wood particles containing small amount of cellulose may not be detectable because of providing insufficient birefringence. Therefore the

polarization-based analysers are not able to detect all fine particles [3].

Laser diffraction

The laser diffraction (LD) method is based on the fact that the spatial distribution of scattered light is a function of the size of particles. In a laser diffraction instrument, the sample stream is illuminated with a laser beam. The laser beam has two light sources, He and Ne, with different wavelengths. The laser light is diffracted by the particles the homogeneous sample stream and a diffraction pattern is produced. The pattern is then measured by detectors and is transferred to the particle size. The smaller particles create

(15)

13

more diffuse diffraction rings than the larger particles. Some laser diffraction

instruments are equipped with a blue and a red laser source of light. The blue laser is used for measuring relatively small particles and the red laser for large particles. The hardware of LD is schematically shown in the figure below.

Figure 6. Schematic hardware of laser diffraction instrument.

There are two main theories by which scattering and adsorption of particles can be predicted, Fraunhofer and Mie models. Fraunhofer model basically can predict scattering pattern of light when a solid, opaque particle with a known size is passing through a laser beam. It is especially used for relatively large particles. Considering that the particles of most experiments are transparent and this theory is not able to describe the scattering pattern for them, Mie theory is a more accepted model to predict the scattering pattern of particles, which was used in LD method in this work. Mie theory provides the pattern, of which the light is scattered, passes through or adsorbed by the spherical particles. Although this theory seems to be more correct than Fraunhofer model, it does require the user to know the refractive index in advance. So the user estimates the refractive index of the particle before analysis by Mastersizer, then it captures the actual scattering pattern backward! Refractive estimation is explained in detail in the next part. Since Mie theory measures particles based on the assumption that they are perfect spheres. As most particles are not ideal spheres, the particle size

measurement becomes complicated. It will be difficult to define size characterizations, especially for irregularly shaped particles.

Mastersizer measures size of the particles by their volume. Since the shape of spheres can be defined by only one unique value, then an equivalent parameter of the irregular shape particles can be described to measure the size of an imaginary spherical particle.

Therefore equivalent sphere diameter (ESD) can be calculated for all the particles, of which most particles are irregular shaped [22].

(16)

14

Volume of the cylinder:

V= 𝜋𝑙 (𝐷)²/4 (6)

Therefore, the equivalent sphere diameter for the cylindrical particle can be calculated as follows:

d =∛(6𝑉/𝜋) (7)

Outside the pulp and paper industry, laser diffraction is probably the most widely used non-microscopic technique for particle size analysis due to ease of use, repeatability, speed of measurements and a wide measurement range. However, there are many articles addressing critical issues concerning this method. Benefits and issues associated with laser diffraction are:

-Ease of use: laser diffraction method requires no calibration, and can be verified easily by available NIST-traceable standards.

-Range of applicability: Laser diffraction is able to characterize different types of sample, such as dry powders, sprays and suspensions.

-Measurement range: a wide dynamic range of size of particles from 20 nm to 2-3 millimetres can be measured so that both well dispersed and agglomerated particles are detected.

- Speed of measurement: Down to 400μs/measurement.

- Measurement repeatability: rapid measurements allow the user to repeat the results many times and improve the accuracy of the measurement [16].

D

l d

Figure 7. ESD of a cylindrical particle.

(17)

15

However there are arguments that this method suffers from some pitfalls and inaccuracies, which are described below.

One of the basic assumptions of the laser diffraction method is random particle orientation. This is especially crucial for extremely non-spherical particles. This assumption is questionable. Here, problems arise from the fact that the flow in laser diffraction is not turbulent but rather laminar. Therefore the particles orient themselves in the flow direction according to the “flow alignment” phenomenon. As a result, particles pass the laser at 90° (perpendicular to the beam) due to “flow alignment” and the projected area is therefore measured instead of actual size, this is especially

important for particles with an aspect ratio larger than 5:1 [17]. According to Kelly RN [18] the assumption of random orientation for non-spherical particles larger than 1 μm with a high aspect ratio is false.

Another questionable assumption in laser diffraction method is that results are volume based and reflect a volume distribution. Kelly RN et al [19] showed that laser

diffraction is not able to differentiate between plates and cubes having similar linear dimensions but different volumes. Therefore “volume probability” results of laser diffraction experiments for irregular particles are based on the projected surface area provided by image analysis. Recent studies on comparison of laser diffraction results with other particle-sizing techniques indicate that there is a poor correlation between image analysis and laser diffraction results for needle-like, rod or plate particles.

Considering the fact that error functions in different methods are different, the image- based and direct observation methods gain more value in particle sizing than results from complicated calculations. Laser diffraction shows 31% error in size for plates and up to 70%, compared with direct measurement [16].

3.2.1 Flow cytometry

Flow cytometry refers to the particle analysis technology, which measures light scatter and fluorescence intensity of particles in suspension. The technique has been used for a long time in the biomedical field. In the pulp and paper industry, it has been used in some studies to characterize mechanical fines [3]. It is especially used to identify different types of particles, which can then be separately analysed.

Advantage of flow cytometry are that a large number of particles in a suspension are analysed in a short period of time, that both optical and chemical information is obtained, and that data is obtained for each individual particle. The limitations of flow cytometry are that the measurement range is limited (usually between 0.2 and 120 µm) and that size information may only be obtained by calibration measurements on

reference particles. Shape data is difficult to obtain.

(18)

16

3.2.2 Combination techniques

With traditional particle characterization analysis, one has to choose between analysis of large population of particles without images (such as laser diffraction method or flow cytometry) or visualization of a limited number of particles by microscope without obtaining representative quantitative results (such as optical microscopes). Microscopy provides information about the fine structure and qualification; on the other hand flow cytometry is able to provide statistical information. The ImageStream is an instrument that combines these two features. In this study physical properties of fines (size and shape) are investigated with different image based characterization methods,

ImageStream, Fiber Tester, and Fiber Tester +, and a light diffraction method, Master Sizer.

(19)

17

4 Materials and Methods

4.1 Sample preparation 4.1.1 Stocks

Unbleached Kraft softwood pulp (UKSW from Frövi, BillerudKorsnäs, refined in the mill) and precipitated calcium carbonate (Albacar LO, Specialty Minerals) were used in experiments. A blend of these two samples (90 wt% UKSW and 10 wt% PCC) was also prepared for measurements. A previous study reported that the Albacar LO PCC was rather coarse, with a mean size of 2.2 µm [6].

4.1.2 BDDJ screening

UKSW pulp was diluted in water with the concentration of 2.06 (gram of dry mass of pulp per volume of water). In order to obtain a fines fraction, a Britt Dynamic Drainage Jar (BDDJ) was used in which fines passed through a mesh with 75 μm holes. BDDJ is a fractionator, which is commonly used to separate fines from fibers. According to the standard of BDDJ, 0.5 of dry pulp was diluted in 2500ml of water.

In order to estimate the amount of fines collected in each of BDDJ trials, BDDJ dilution was filtered by vacuum filtration. The filter paper was dried in an oven over night at 110°C. For the 0.5 g of the pulp, 0.0117 g of fines was obtained. A relatively high sample concentration was needed for the ImageStream measurements in order to analyse a sufficient number of particles. Since the concentration of fines in BDDJ dilution is relatively low (4.68×10^-3 g/l), the BDDJ screening was modified and only 1000 ml of water was used instead of 2500 ml. Therefore, a concentration of 1.17×10^-2 g/ l was prepared.

The details of sample preparation for each characterization method are described in the corresponding section.

(20)

18

4.2 Measurement instruments 4.2.1 Overview

An overview on the specifications of the different measurement methods is given in Table 1.

Table1: Overview of measurement properties of different methods used in this work [21].

LD IS20X IS40X IS60X FT+ FT

Resolution ~ 20 nm 0.33 µm/pix

0.5 µm/pix

1.0 µm/pix

~ 3.5 µm/pix

~ 10 µm/pix Maximum

particle size 2000 µm ~120 µm ~ 9000 µm.

Standard

weighting Volume Number Length

Standard

output SED Length, Width, Area, CED,

etc. Length, Width

Measurement

time 10 minutes 3 minutes 10-20 minutes

Sample

volume 20 or 60 ml 200 µl 200 ml

4.2.2 ImageStream^X MarkII

An ImageStream ^XMarkII from Merck Millipore was used. Each suspension was run in three different objective microscopes with different magnifications (IS20X, IS40X and IS60X). Each magnification provides different resolution. The specifications for the different magnifications are given in Table 2.

Table 2: Properties of different magnifications in the ImageStream instrument [21].

Magnification 60X 40X 20X

Resolution 0.33 µm/pixel 0.5 µm/pixel 1.0 µm/pixel Field-of-view 40 x 170 µm 60 x 256 µm 120 x 512 µm

Depth-of-focus 2.5 µm 4.0 µm 8.0 µm

In the figure 9 below, an overview over the ImageStream platform is schematically shown. Particles in a suspension are hydrodynamically focused and illuminated by a red bright field and a laser with a wavelength of 488nm. A microscope objective collects

(21)

19

fluorescence, transmitted and scattered light, and then a dichroic filter split them into six images (see Figure 9) in different ranges of wavelengths. These images are directed onto the surface of a charge coupled device (CCD) camera, and saved for further analysis by the IDEAS software [20]. In this study, only the visual image (Ch01) was used.

Figure 8. An example of sub-images of sample from the IS60X in different wavelengths, as they appeared in the IDEAS software.

Figure 9. Schematic hardware of ImageStream.

In the measurement of the stock samples, a sample of 100 μl of the suspension was added into a plastic tube, and was placed in the system. To allow for estimation of the measurement uncertainty, each measurement was repeated five times. Each

(22)

20

measurement was done in about 3 minute, with 10000 images of separate particles being captured in each run. The magnification was then changed, and the procedure was repeated until all three magnifications had been used.

All the images were collected for analysis by the IDEAS software (Image Data

Exploration and Analysis Software, Amnis Corporation, USA). The five measurements from the same suspension, which were run with the same magnification, were then merged into a single file. Automatic image corrections such as background subtraction were made by the analysis software.

In each run, SpeedBeads which are particles that test and calibrate the instrument’s illumination, optical, camera and fluidic systems, are mixed into the sample. In addition SpeedBeads provide run-time information, which helps continuous synchronization between the camera and the sample flow rate. But images containing these SpeadBeads were not continent to our experiments, and were therefore deleted using a template defined in analysis software, IDEAS. A template is simply a set of instructions for further analysis. The definition for features, graphs, regions and populations are instructed in the default template. Figure 10 shows an example of a template.

Figure 10. Regions containing particles with different estimated types can be defined in template.

(23)

21

Although the template defines the region of our interest, it still contains some images of the SpeadBeads. It is almost impossible to adjust a perfect region, which contains only the particles we want to characterize. Besides, there are sometimes multi particles in one single image, or poorly focused images which make the measurements complicated.

These images can be “tagged” and be collected as a separate population folder.

This population can then be omitted, when defining the final analysis population

(Figure 10). When all the images were ready for characterization, the desired properties to measure (length, width, area and perimeter) were featured for the final population.

The result data was collected as a text file.

Figure 11 Tagged images are collected in a folder and removed from the final analysis population.

4.2.3 FiberTester and FiberTester+

The FiberTester (FT) and FiberTester+ (FT+) from L&W are image-based fibre analysers. The FT+ provides a slightly different result than the FT since it captures images of detected particles with a higher resolution. Because of the higher resolution, it is possible to detect both fibrils and fibers with the FT+. In addition, fines can be

classified in P (primary) and S (secondary) fines using the FT+.

Another technical difference between the two instruments is that the images are captured in a shorter time in the FT+ so that the fibres won’t get enough time to move during the capturing time. This will lead to slightly higher values of size presented in FT in comparison with FT+.

During the FT+ measurements, 100-200 ml of each sample was added to a beaker. All the beakers containing sample suspensions were placed on the sample-changer carousel.

Each test was done by approximately 10 minutes, including dilution and flushing and measurement, depending on whether maximum number of particles (set to 100000

(24)

22

particles) was analysed or the maximum running time was reached. The same procedure was applied when using the FT.

FT and FT+ data results usually length-weighted. However, when raw data is obtained for each particle, as in this study, it is number-weighted.

For each particle detected in the FT and FT+, a value for A (area) and P (perimeter) is derived in the hardware of the camera. From these parameters the rectangle-equivalent length and width is calculated as follows:

𝐴 = 𝐿𝑊 (1) 𝑃 = 2𝑊 + 2𝐿 (2) Since𝐿 ≫ 𝑊, then the equations above can be modified as follows (ER stands for equivalent rectangle):

𝐿ER= 𝑃/2 (3) 𝑊ER= 2𝐴/𝑃 (4)

Equivalent rectangle length and width are the parameter of a rectangle whose area is the same as the particles.

Equivalent rectangle length and width is then reported for each detected particle as a text file.

Figure 12. Values of width and length of an equivalent rectangle whose area matches that of the particle’s image was measured.

(25)

23

4.2.4 Mastersizer 2000 (laser diffraction)

The laser diffraction method was applied using a Mastersizer 2000 from Malvern Instruments, using the universal liquid module. Three batches of suspensions as explained before were prepared as measuring samples. Measuring conditions were as follows: Sample refractive index 1.38 to 1.56 (estimation of refractive index is explained below), running time was 1 minute. Between each sample measurement, a sample of water was run in order to clean the fluid container. For each sample, the measurement was repeated and an average of the two measurements was given. The data was extracted and then analysed using the Mastersizer2000 analysis software.

As mentioned before, in order to interpret results from the Mastersizer, the refractive index of the particles needed to be estimated in advance. Choosing correct values of refractive index is an important key to obtain accurate result from laser diffraction particle size measurements.

Refractive index is most crucial for spherical or/and transparent particles or when the sizes of the particles are close to wavelength of the light and when RI of the particle is close to that of the fluid, and least crucial for non-spherical or/and opaque particles or when the sizes of the particles are relatively larger than wavelength of the light and when RI of the particle is larger than that of fluid.

The refractive index of PCC was obtained from tabled values, based on the assumption that it did not change when the particle was suspended in water. For the UKSW and mixed samples, the refractive index was calculated based on a mixing rule [21]. The chemical composition and water content of the fines was accounted for.

Table 3. Estimated refractive index of the stock samples. In all cases, the extinction coefficient was estimated to be negligible.

Stock sample UKSW PCC Mix

Refractive index 1.38 1.56 1.47

In the analysis software, the refractive index was set for each sample, and the particle size distribution was generated for 50 logarithmically spaced sizes ranging from 0.02 µm to 1000 µm. The data was then exported as a text file.

Figure 13 below shows that the refractive index plays an important role when applying the Mie theory. It shows how the size distribution of the mix sample varies with

different values of refractive index.

(26)

24

Figure 13 Influence of refractive index on a result of a particle size distribution obtained using the LD method.

LD gives only a single size descriptor, the equivalent sphere diameter, while the ImageStream, FT and FT+ give several. This makes the comparison of data complicated. Therefore, in order to meaningfully conduct the quantitative data

comparison between LD and other methods, an additional Matlab analysis was required.

5 MATLAB Analysis

As discussed before, laser diffraction reports values of equivalent sphere diameter (ESD) for the particles as the only size descriptor. In order to conduct a meaningful data comparison between LD and other methods, one has to estimate ESD for the same samples from all the other methods (FT, FT+ and ImageStream). They also needed to be volume-weighted. Another set of comparison between image-based analysis instruments was also done. Then, the data from the imaging analysis methods was area-weighted.

The values of equivalent rectangle width and length for ImageStream data were calculated directly from area and perimeter according to Equations (1) and (2). The aspect ratio (length/width) was calculated as equivalent rectangle length per equivalent rectangle width. Particles with aspect ratio less than 1.33 were assumed spherical and ESD for them was considered equal to the equivalent circle diameter (ECD) which is the diameter of a circle whose area is equal to the area of the particle. For other particles with aspect ratio higher than 1.33, ESD was considered as the diameter of a sphere whose volume is equal to the volume of a cylinder with WER and LER dimensions.

0 1 2 3 4 5 6 7 8 9 10 11

0,0 0,1 1,0 10,0 100,0 1000,0

V olum e in b in %

Equivalent Sphere Diameter

UKSW RI : 1,38 Mix RI : 1,4 Mix RI : 1,42 Mix RI : 1,44 Mix RI : 1,46 Mix RI : 1,48 Mix RI : 1,5 Mix RI : 1,52 Mix RI:1,54 PCC RI:1,56

(27)

25

In order to be able to compare the results for the different methods, the histograms of volume-weighted ESD data obtained from image-based analysers with bin-sizes and range equal to the logarithmic x-axis of graph of Mastersizer was produced. For the image-based comparisons, area-weighted values of length and width of particles were plotted.

A=L+W P=2L+2W

ImageStream(A,P)

LER= P/2 WER= 2A/L

AR=LER/W

ER

AR<1.33 AR>1.33

ESD=ECD= (4A/π)^1/2 ESD= ((3/2)WER

2 LER)^1/3

Figure 14. Algorithm of obtaining ESD values from FT, FT+ and ImageStream.

FT,FT+(L,W)

w L

E S D

E C D

Figure 15. Equivalent sphere diameter was calculated from WER and LER for AR higher than 1.33 and from ECD for AR lower than 1.33.

WER

LER ESD ECD ESD

(28)

26

6 Results and discussion

6.1 Image-based and LD methods comparison

As described before, the only characterization parameter that can be defined for all the instruments, is Equivalent Sphere Diameter (ESD). The values for this parameter was obtained directly from LD, and calculated from other parameters in other methods. The figures below show the ESD vs volume for all three samples by various instruments.

The values are volume-weighted.

The data in figure 16 demonstrates clear differences in ESD for UKSW between

different analysers. Fine particles with ESD values lower than 6 μm are not seen in the Mastersizer (LD method) data, despite the common expectation that Mastersizer should be able to detect particles with ESD≥ 0.2 μm. However, Mastersizer seems to be in better agreement with other methods in detecting small PCC particles compared to the UKSW result, since the minimum size detected for ESD of PCC by Mastersizer is shifted to lower than 1.5μm. Possible explanation is that PCC particles have higher light scattering than fines and therefore Mastersizer suits better for PCC particles than

cellulose fines.

As mentioned before refractive index is most crucial for spherical or/and transparent particles or when the sizes of the particles are close to wavelength of the light and when RI of the particle is close to that of fluid, and least crucial for non-spherical or/and opaque particles or when the sizes of the particles are relatively larger than wavelength of the light and when RI of the particle is larger than that of fluid. Considering this important point that RI of UKSW is relatively close to RI of the fluid, which is water (1.33) might explain why LD performed poorer for UKSW. Also UKSW are likely to be more transparent than PCC particles since cellulose fibres generally absorb water and swell.

Aside from that, since in LD particles are considered spheres, one could expect better compatibility of results of LD with other methods for particles with lower aspect ratio.

PCC particles are produced and are supposed to have a known (regular) structure compared to wood fines that are extremely irregular; therefore a narrower size distribution for PCC is supposed to be obtained compared to UKSW. Comparing the results for PCC and the other two samples containing UKSW shows better compatibility for PCC results as expected.

For the LD results of the mix sample two peaks for PCC and fines were obtained. The graph shows that the results are more dominated by fines rather than PCC particles. This was expected since only 10wt% of the mix consists of PCC. In addition, cellulosic fine particles are larger than PCC particles, and also PCC particles tend to deposit on cellulosic fines, therefore they might not be under-counted in the measurements. Also

(29)

27

considering the fact that the measurements of LD are set up to account for the particles as a volumetric weighting, and since PCC particles are denser than fine PCC particles are expected to be partly disappear in the mixture measurements with LD.

Comparing FT and FT+ for all three samples, demonstrates that FT+ is able to detect smaller particles and that is because FT+ has a higher resolution than FT. Apart from that FT shows statically higher values than FT+ as if FT observes the same particles bigger than what they are observed by FT+. The possible explanation is that since lighting time is shorter in FT+ than in FT, particles have less time to move while they are captured by the camera. So, particles are observed smaller in size in FT+. FT results show an overestimation for all three samples with significantly large ESD. As particle size increases, fewer particles are measured, and statistically a lower number of large particles can influence particle size distribution, since the data is volume weighted.

Completely different results were obtained from LD and ImageStream for UKSW and also the mixture sample with a very small overlap. Looking at the graphs and comparing ImageStream results with other methods implies a major shift towards smaller size. It indicates that ImageStream presents data for small particles, which are not visible to LD or FT+. The 20X magnification of ImageStream results show a strong peak for all three samples. The stronger influence of larger particles on volume-based data, as the number of them is reduced (since 20X detects larger particles than 40X and 60X) could be an explanation to this phenomenon.

(30)

28 Figure 16. ESD for UKSW by various instruments.

Figure 17. ESD for PCC by various instruments.

(31)

29

Figure 18. ESD for the mixture of PCC and UKSW sample by various instruments.

Figure 19. Laser diffraction results for all three samples.

(32)

30

Figure 20. ImageStream 20X Length distribution for all the samples

(33)

31

Figure 23. FT length distribution for all the samples.

Figure 24. FT+ length distribution for all the samples.

(34)

32

6.2 Image-based methods comparison

Area-weighted width and length data from all three methods were plotted. The data in figures below indicates that ImageStream performed best for the smallest particles. On the other hand, only larger particles are detected by FT as expected due to the lower resolution compared to FT+ and ImageStream. The minimum and maximum values of particles detected by various methods are reported in the tables below.

FT FT⁺ IS20X IS40X IS60X PCC max 3363 59 53 60 43.667 PCC min 15 4 4 3 2 UKSW max 4131 792 106 92 114.67 UKSW min 15 4 4 2 1.8333 Table 5. Min and max values of Equivalent rectangle length (𝜇𝑚) obtained from various methods.

FT FT+ IS20X IS40X IS60X PCC max 78.9 59.3 6.6327 9.8333 5.5196 PCC min 11.3 4.8 0.25 0.25 0.16667 UKSW max 78.2 68.9 13.978 6.6566 10.623 UKSW min 9.1 3.5 0.25 0.125 0.16667

Table 6. Min and max values of Equivalent rectangle width (μm) obtained from various methods.

(35)

33

Figure 25. Equivalent rectangle length distribution for UKSW by image-based methods.

Figure 26. Equivalent rectangle width distribution for UKSW by Image-based methods.

(36)

34

Figure 27. Equivalent rectangle length distribution for PCC by image-based methods.

Figure 28. Equivalent rectangle width distribution for PCC by Image-based methods.

(37)

35

7 Conclusions

Laser diffraction has been widely used in size characterization studies of fines, as

image-based instruments have had insufficient resolution and dynamic range. This study

Figure 29. Equivalent rectangle length distribution for mixture sample by Image-based methods.

Figure 30. Equivalent rectangle width distribution for mixture sample by Image-based methods.

(38)

36

compared laser diffraction measurements to measurements with image-based instruments with different resolutions.

Differences between the measurements principles and also hardware of different characterization methods made the data evaluation between them challenging. Image analysis typically report more than one size descriptor, while LD measures equivalent sphere diameter as the only size descriptor. However a correlation between the results can be made by recalculating the length and width data from image analysis into volume-weighted equivalent sphere diameter.

The results from image-based methods were transferred to ESD and were plotted together with LD results. The effect of the differences in resolution and dynamic ranges of the image analysis instruments was clear. The highest-resolution instrument, the ImageStream, performed well for small particles but could not simultaneously measure the largest due to its limited dynamic range. The FT and FT+ was better for relatively large particles, and the FT+ could also measure some of the smaller material.

LD was clearly unsatisfactory for measuring UKSW fines and also the mixture of fines and filler. While LD claims a resolution high enough to detect nanofibril aggregates (20nm), no fines particles smaller than ~3µm were seen in the data. However a more reasonable size distribution for PCC was obtained from LD.

In summary, LD was not found to be better for fines characterization than the high- resolution image based instruments. However, as flow imaging is limited by the

diffraction limit, resolutions better than ~200nm are unlikely to be achieved. Thus, none of the investigated instruments has the potential to detect cellulose micro/nano-fibrils; a particle type where there is a great demand for a quantitative and statistical analysis method.

In future work non-optical characterization methods could be a subject of research for characterization of fines and fillers, since these methods are not influenced by optical properties of the particles. One such method is Tunable Resistive Pulse Sensing (TRPS). And finally caution needs to be considered regarding refractive index when using LD method for a mixture of two different materials with very different range of size and chemical composition. Dedicated measurements of the refractive index of wet fibres, fines, and fillers would be valuable future work.

(39)

37

8 References

[1] Kelly, RN., Disante, K., Stranzl, E,. Kazanjian, J,. Bowen, P,. Matsuyama, T,.

Gabas, N,. Graphical Comparison of Image Analysis and Laser Diffraction Particle Size Analysis Data Obtained From the Measurements of Nonspherical Particle Systems.

AAPS PharmSciTech 2006; Article 69.

[2] M. Bumiler: The importance of particle shape, HORIBA Scientific 2013.

[3] Luukko, K. and Nurminen, I. (1999): Fines generation in the first-stage refiner of thermomechanical pulping, an interpretation by image analysis, Pap. Puu 81(4), 311.

[4] Mosbye, J., Colloidal wood resin: Analyses and interactions, PhD thesis, Norwegian University of Science and Technology; July 2003.

[5] Rundlöf, M. (2002): Interaction of dissolved and colloidal substances with fines of mechanical pulp-influence on sheet properties and basic aspects of adhesion, PhD thesis, Royal Institute of Technology, Stockholm, Sweden.

[6] R.S. Seth: The Measurement and Significance of Fines, Pulp and paper Canada 2003.

[7] K. Luukko, H. Paulapuro: Mechanical pulp fines: effect of particle size and shape, TAPPI JOURNAL, February 1999, Vol. 82(2)

[8] Sundberg, A., Pranovich, A. V. and Holmbom, B. (2003): Chemical characterisation of various types of mechanical pulp fines, J. Pulp Paper Sci. 29(5), 173.

[9] Heikkurinen, A. and Hattula, T. (1993): Mechanical pulp fines - characterization and Implications for defibration mechanisms, 18th Int. Mech. Pulping Conf., Oslo,

294.

[10] Vispul, S. Chauhan, Nishi K. Bhardwaj : Effect of Particle Size of Talc Filler on Structural and Optical Properties of Paper, Lignocellulose 1(3), 241-259, 2012.

[11] J,Grönfors. Use of fillers in paper and paperboard grades. Tampere University of Applied Sciences, International Pulp and Paper Technology, 2010.

[12] Wilsen, I. Filler and Coating Pigments for Papermakers. Industrial Minerals and Rocks, Page 1287-1300, 2006.

[13] Chen, X., Qian, X., and An X: Using Calcium Carbonate Whiskers as Papermaking Filler, BioResource, 6(3) 2435-2447, 2011.

(40)

38

[14] Thad C. Maloney, José Ataide, Juha Kekkonen, Henrik Fordsmand, Henrik Hoeg- Petersen.(2006?): Changes to PCC structure in papermaking.

[15] Seth, R.S., Jang, H.F., Chan, B.K., WU, C.B: Transverse Dimensions of Wood Pulp Fibers and Their Implications for End Use. Cambrige, Pira International, Leatherhead, UK, 473-503, 1997.

[16] Richard N. Kelly, Frank M. Etzler. What is wrong with laser diffraction?

[17] Xu R, Di Guida A. Size and Shape Characterization of Small Particles. Powder Technol, 132:145-153, 2003.

[18] Kelly RN. False Assumptions: Laser Difraction PSA Systems Exposed. NJPhAST Meeting, May 13, 2004

[19] Kelly RN, Kazanjian J. Use of LGC Promochem Shape Standards AEA1001- AEA1003 in the Study of the Effects of Particle Shape on the Results from Current Generation Laser Diffraction-Based Particle Size Analysis Systems. Personal Communication, in preparation, 2005.

[20] Inspire, ImageStream System Software, User’s Manual, Aminis Corporation, April 2006.

[21] K. Hyll : Size and shape characterization of fines and fillers method overview and Literature review. Page 1-98, February 2015.

[22] Horiba Scientific, A Guidebook to Particle Size Analysis. Horiba Instrument Inc.

Irvine, USA, 2014.

[23]Wood, J. R. Grondin, M, and Karins, A: Characterization of Mechanical Pulp Fines with a Small Hydrocyclone. Part 1: The principle and nature of the separation, J. Pulp.

Pap. Sci. 17(1), 1-5.

[24] Bergström, L., Stemme, S., Dahlfors, T., Arwin, H. and Ödberg, L: Spectroscopic ellipsometry characterisation and estimation of the Hamaker constant of cellulose, Cellulose, 6, 1-13, 1996.

[25] Hollertz, R. : Dieletric properties of wood fibre components relevant for electrical insulation applications, Thesis from KTH Royal Institute of Technology, Stockholm, 2014.

[26] Segelstein, D. : The complex refractive index of water, Thesis from University of Missouri, Kansas City (MI) 1981.

[27] Rundlöf, M., Htun, M., Höglund, H. and Wågberg, L. (2000): Mechanical pulp fines of poor quality – characteristics and influence of white water, J. Pulp Paper Sci.

26(9), 308.

(41)

39

[28] Hubbe, Martin A.:Filler Particle Shape vs. Paper Properties-A review, TAPPI 2004 Spring Tech. conf., Altlanta, Paper 4-3.

Comparison of Optical Methods for Fines and Filler Characterization